fix(route/ieee): fix journal.ts, author.ts and earlyaccess.ts #15120
+229
−108
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Successfully generated as following: http://localhost:1200/ieee/journal/6287639/vol-only-seq - Failed ❌ |
github-actions
bot
added
the
Auto: Route Test Complete
|
Successfully generated as following: http://localhost:1200/ieee/journal/6287639 - Failed ❌ |
|
Successfully generated as following: http://localhost:1200/ieee/journal/6287639 - Failed ❌ |
Derekmini
changed the title
/routes/ieee/journal.ts and the /routes/ieee/recent.ts in category journalgot to ofetch in /routes/ieee/journal.ts
Derekmini
changed the title
got to ofetch in /routes/ieee/journal.tsgot to ofetch in /ieee/journal.ts
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
Derekmini
changed the title
got to ofetch in /ieee/journal.tsgot to ofetch in /ieee/journal.ts
|
Successfully generated as following: http://localhost:1200/ieee/journal/8782710 - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Open Journal of Signal Processing</title>
<link>https://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=8782710</link>
<atom:link href="http://localhost:1200/ieee/journal/8782710" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Open Journal of Signal Processing - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Tue, 09 Apr 2024 08:21:03 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Synthbuster: Towards Detection of Diffusion Model Generated Images</title>
<description><p><span><big>Synthbuster: Towards Detection of Diffusion Model Generated Images</big></span><br></p><p><span><small><i>Quentin Bammey</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337714</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Synthetically-generated images are getting increasingly popular. Diffusion models have advanced to the stage where even non-experts can generate photo-realistic images from a simple text prompt. They expand creative horizons but also open a Pandora's box of potential disinformation risks. In this context, the present corpus of synthetic image detection techniques, primarily focusing on older generative models like Generative Adversarial Networks, finds itself ill-equipped to deal with this emerging trend. Recognizing this challenge, we introduce a method specifically designed to detect synthetic images produced by diffusion models. Our approach capitalizes on the inherent frequency artefacts left behind during the diffusion process. Spectral analysis is used to highlight the artefacts in the Fourier transform of a residual image, which are used to distinguish real from fake images. The proposed method can detect diffusion-model-generated images even under mild
jpeg
compression, and generalizes relatively well to unknown models. By pioneering this novel approach, we aim to fortify forensic methodologies and ignite further research into the detection of AI-generated images.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334046/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334046/</guid>
</item>
<item>
<title>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</title>
<description><p><span><big>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</big></span><br></p><p><span><small><i>Yanbin Zou; Jingna Fan; Zekai Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we consider using time-of-arrival (TOA) measurements from a single moving receiver to locate a moving target at constant velocity that emits a periodic signal with unknown signal period. First, we give the TOA measurement model and deduce the Cram
$\acute{\text{e}}$
r-Rao lower bounds (CRLB). Then, we formulate a nonlinear least squares (NLS) problem to estimate the unknown parameters. We use semidefinite programming (SDP) techniques to relax the nonconvex NLS problem. However, it is shown that the SDP localization algorithm cannot provide a high-quality solution. Subsequently, we develop a fixed point iteration (FPI) method to improve the performance of the SDP algorithm. In addition, we also consider the presence of receiver position errors and develop the corresponding localization algorithm. Numerical simulations are conducted to demonstrate the localization performance of the proposed algorithms by comparing them with the CRLB.
Index Term
-Fixed point iteration (FPI), semidefinite programming (SDP), single moving receiver, target localization, time-of-arrival (TOA).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336384/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336384/</guid>
</item>
<item>
<title>The Neural-SRP Method for Universal Robust Multi-Source Tracking</title>
<description><p><span><big>The Neural-SRP Method for Universal Robust Multi-Source Tracking</big></span><br></p><p><span><small><i>Eric Grinstein; Christopher M. Hicks; Toon van Waterschoot; Mike Brookes; Patrick A. Naylor</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340057</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Neural networks have achieved state-of-the-art performance on the task of acoustic Direction-of-Arrival (DOA) estimation using microphone arrays. Neural models can be classified as end-to-end or hybrid, each class showing advantages and disadvantages. This work introduces Neural-SRP, an end-to-end neural network architecture for DOA estimation inspired by the classical Steered Response Power (SRP) method, which overcomes limitations of current neural models. We evaluate the architecture on multiple scenarios, namely, multi-source DOA tracking and single-source DOA tracking under the presence of directional and diffuse noise. The experiments demonstrate that our proposed method compares favourably in terms of computational and localization performance with established neural methods on various recorded and simulated benchmark datasets.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345765/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345765/</guid>
</item>
<item>
<title>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</title>
<description><p><span><big>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</big></span><br></p><p><span><small><i>Alexander Venus; Erik Leitinger; Stefan Tertinek; Klaus Witrisal</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper presents a neural-enhanced probabilistic model and corresponding factor graph-based sum-product algorithm for robust localization and tracking in multipath-prone environments. The introduced hybrid probabilistic model consists of physics-based and data-driven measurement models capturing the information contained in both, the line-of-sight (LOS) component as well as in multipath components (NLOS components). The physics-based and data-driven models are embedded in a joint Bayesian framework allowing to derive from first principles a factor graph-based algorithm that fuses the information of these models. The proposed algorithm uses radio signal measurements from multiple base stations to robustly estimate the mobile agent's position together with all model parameters. It provides high localization accuracy by exploiting the position-related information of the LOS component via the physics-based model and robustness by exploiting the geometric imprint of multipath components independent of the propagation channel via the data-driven model. In a challenging numerical experiment involving obstructed LOS situations to all anchors, we show that the proposed sequential algorithm significantly outperforms state-of-the-art methods and attains the posterior Cramér-Rao lower bound even with training data limited to local regions.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336409/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336409/</guid>
</item>
<item>
<title>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</title>
<description><p><span><big>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</big></span><br></p><p><span><small><i>AMIT EFRAIM; Joseph M. Francos</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In solving a model fitting problem, the existence of outliers in the set of measurements can have a devastating effect on the solution accuracy. Traditionally, in order to overcome this problem, robust point cloud registration algorithms are composed of transformation hypothesis generation, followed by hypothesis evaluation aimed at selecting the best hypothesized estimate. Hypotheses evaluation is commonly performed using the sample consensus criterion. However, since this criterion accounts only for the consensus size, it fails when the maximal sample consensus is incorrect. We propose a new hypothesis evaluation approach, generalizing the sample consensus approach, where instead of the sample consensus, the transformation that maximizes the point clouds feature correlation is selected as the best hypothesis. The feature vector at each point contains information such as on local geometry and semantic context. Utilizing this information in the hypotheses evaluation and selection procedure allows for a correct decision even when the hypothesis yielding the maximal sample consensus is false. Consequently, the probability of selecting the correct model increases. We show both mathematically and empirically that substituting the sample consensus criterion with the proposed point cloud feature correlation hypothesis test (PC-FCHT) lowers the probability of large registration errors, compared to using the special case of sample consensus. The proposed PC-FCHT is applicable to any algorithm that follows the hypothesis generation and evaluation scheme, potentially improving the success rates of a wide family of point cloud registration algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345750/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345750/</guid>
</item>
<item>
<title>Joint PAPR and OBP Reduction for NC-OFDM Systems</title>
<description><p><span><big>Joint PAPR and OBP Reduction for NC-OFDM Systems</big></span><br></p><p><span><small><i>Hsuan-Fu Wang; Fang-Biau Ueng; Bo-Heng Yeh</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3329757</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The spectrum resource is always a critical issue for wireless communications since it directly impacts the data rate and capacity. However, the problem of spectrum resource scarcity always exists. Moreover, spectrum resource scarcity becomes more severe as new communication technologies and wireless applications sprout. Noncontiguous orthogonal frequency division multiplexing (NC-OFDM) is a multicarrier method for bandwidth utilization. Unfortunately, this system has two fatal defects: high peak-to-average power ratio (PAPR) and considerable out-of-band power (OBP), which are detrimental to the system's performance. To solve these two problems, we propose a convex optimization-based method for joint PAPR and OBP reduction in NC-OFDM Systems. The strategy is to permit the secondary user to utilize the unoccupied spectrum of the primary user with dynamic spectrum sharing (DSS) based on a cognitive radio network (CRN). To this end, a flexible system operating over noncontiguous bands and DSS scenarios is necessary. The simulation results have shown that our method could effectively improve the overall performance and outperform other schemes, i.e., projections onto convex sets (POCS) and alternating projections onto convex and non-convex sets (APOCNCS), without harming the transmission of the primary system. The collaboration between secondary and primary systems is viable with the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10305259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10305259/</guid>
</item>
<item>
<title>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</title>
<description><p><span><big>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</big></span><br></p><p><span><small><i>Gal Itzhak; Israel Cohen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339433</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article presents a Kronecker-product (KP) beamforming approach incorporating sparse concentric circular arrays (SCCAs). The locations of the microphones on the SCCA are optimized concerning the broadband array directivity over a wide range of direction-of-arrival (DOA) deviations of a desired signal. A maximum directivity factor (MDF) sub-beamformer is derived accordingly with the optimal locations. Then, we propose two global beamformers obtained as a Kronecker product of a uniform linear array (ULA) and the SCCA sub-beamformer. The global beamformers differ by the type of the ULA, which is designed either as an MDF sub-beamformer along the
$\mathsf {x}$
-axis or as a maximum white noise gain sub-beamformer along the
$\mathsf {y}$
-axis. We analyze the performance of the proposed beamformers in terms of the directivity factor, the white noise gain, and their spatial beampatterns. Compared to traditional beamformers, the proposed beamformers exhibit considerably larger tolerance to DOA deviations concerning both the azimuth and elevation angles. Experimental results with speech signals in noisy and reverberant environments demonstrate that the proposed approach outperforms traditional beamformers regarding the perceptual evaluation of speech quality (PESQ) and short-time objective intelligibility (STOI) scores when the desired speech signals deviate from the nominal DOA.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342869/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342869/</guid>
</item>
<item>
<title>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</title>
<description><p><span><big>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</big></span><br></p><p><span><small><i>Karn N. Watcharasupat; Chih-Wei Wu; Yiwei Ding; Iroro Orife; Aaron J. Hipple; Phillip A. Williams; Scott Kramer; Alexander Lerch; William Wolcott</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339428</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Cinematic audio source separation is a relatively new subtask of audio source separation, with the aim of extracting the dialogue, music, and effects stems from their mixture. In this work, we developed a model generalizing the Bandsplit RNN for any complete or overcomplete partitions of the frequency axis. Psychoacoustically motivated frequency scales were used to inform the band definitions which are now defined with redundancy for more reliable feature extraction. A loss function motivated by the signal-to-noise ratio and the sparsity-promoting property of the 1-norm was proposed. We additionally exploit the information-sharing property of a common-encoder setup to reduce computational complexity during both training and inference, improve separation performance for hard-to-generalize classes of sounds, and allow flexibility during inference time with detachable decoders. Our best model sets the state of the art on the Divide and Remaster dataset with performance above the ideal ratio mask for the dialogue stem.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342812/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342812/</guid>
</item>
<item>
<title>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</title>
<description><p><span><big>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</big></span><br></p><p><span><small><i>Miguel Ferrer; María de Diego; Alberto Gonzalez</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340106</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The LMS algorithm is widely employed in adaptive systems due to its robustness, simplicity, and reasonable performance. However, it is well known that this algorithm suffers from a slow convergence speed when dealing with colored reference signals. Numerous variants and alternative algorithms have been proposed to address this issue, though all of them entail an increase in computational cost. Among the proposed alternatives, the affine projection algorithm stands out. This algorithm has the peculiarity of starting from
$N$
data vectors of the reference signal. It transforms these vectors into as many data vectors suitably normalized in energy and mutually orthogonal. In this work, we propose a version of the LMS algorithm that, similar to the affine projection algorithm, starts from
$N$
data vectors of the reference signal but corrects them by using only a scalar factor that functions as a convergence step. Our goal is to align the behavior of this algorithm with the behavior of the affine projection algorithm without significantly increasing the computational cost of the LMS.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345730/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345730/</guid>
</item>
<item>
<title>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</title>
<description><p><span><big>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</big></span><br></p><p><span><small><i>Zongyao Li; Ren Togo; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340616</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article proposes a method for transferring knowledge of semantic segmentation from a labeled source domain to an unlabeled target domain without using the source-domain data. Such a problem is called source-data-free domain adaptation, in which a pre-trained source-domain model and the unlabeled target-domain data are used to transfer the label knowledge across the domains. Like most previous methods, our method uses pseudo labels for distilling and transferring the source-domain knowledge. On the basis of the pseudo-label learning, our method improves the domain adaptation performance in two innovative ways: 1) reducing the domain differences by source-data-free style transfer and 2) exploring the style diversity within the target domain by style modification. To this end, we introduce two additional modules: 1) an inter-domain style transfer module which aligns the feature statistics of the source and target domains before producing the pseudo labels thereby improving the pseudo labels' accuracy, and 2) an intra-domain style modification module which modifies the image styles within the target domain for learning intra-domain style-invariant features. Our method with the two modules outperforms previous source-data-free domain adaptation methods in two commonly used benchmarks. Moreover, our method is well compatible with the previous methods for further improvement.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10356748/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10356748/</guid>
</item>
<item>
<title>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</title>
<description><p><span><big>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</big></span><br></p><p><span><small><i>Tobias Kabzinski; Peter Jax</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337721</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Quasi-continuous system identification of time-variant linear acoustic systems can be applied in various audio signal processing applications when numerous acoustic transfer functions must be measured. A prominent application is measuring head-related transfer functions. We treat the underlying multiple-input-multiple-output (MIMO) system identification problem in a state-space model as a joint estimation problem for states, representing impulse responses, and state-space model parameters using the expectation maximization (EM) algorithm. We address limitations of prior work by imposing different model structures, especially for dependencies within a (transformed) state vector. This results in block diagonal matrix structures, for which we derive M-step update rules. Making assumptions about this model structure and choosing a block size for a given application define the computational complexity. In examples, we found that applying this framework yields improvements of up to 10 dB in relative system distance in comparison to a conventional method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334061/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334061/</guid>
</item>
<item>
<title>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</title>
<description><p><span><big>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</big></span><br></p><p><span><small><i>Damir Rakhimov; Martin Haardt</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337729</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this paper, we present an analytical performance assessment of 2-D Tensor ESPRIT in terms of physical parameters. We show that the error in the
$r$
-mode depends only on two components, irrespective of the dimensionality of the problem. We obtain analytical expressions in closed form for the mean squared error (MSE) in each dimension as a function of the signal-to-noise (SNR) ratio, the array steering matrices, the number of antennas, the number of snapshots, the selection matrices, and the signal correlation. The derived expressions allow a better understanding of the difference in performance between the tensor and the matrix versions of the ESPRIT algorithm. The simulation results confirm the coincidence between the presented analytical expression and the curves obtained via Monte Carlo trials. We analyze the behavior of each of the two error components in different scenarios.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334446/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334446/</guid>
</item>
<item>
<title>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</title>
<description><p><span><big>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</big></span><br></p><p><span><small><i>Yaman Kındap; Simon Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343341</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this work we study linear vector stochastic differential equation (SDE) models driven by the generalised hyperbolic (GH) Lévy process for inference in continuous-time non-Gaussian filtering problems. The GH family of stochastic processes offers a flexible framework for modelling of non-Gaussian, heavy-tailed characteristics and includes the normal inverse-Gaussian, variance-gamma and Student-t processes as special cases. We present continuous-time simulation methods for the solution of vector SDE models driven by GH processes and novel inference methodologies using a variant of sequential Markov chain Monte Carlo (MCMC). As an example a particular formulation of Langevin dynamics is studied within this framework. The model is applied to both a synthetically generated data set and a real-world financial series to demonstrate its capabilities.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360268/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360268/</guid>
</item>
<item>
<title>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</title>
<description><p><span><big>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</big></span><br></p><p><span><small><i>James M. Cozens; Simon J. Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344048</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper proposes a probabilistic approach for extracting time-varying and irregular time signature information from polyphonic audio extracts, subsequently providing beat and bar line positions given inferred time signature divisions. This is achieved via dynamically evaluating the beat tempo as a function of time through finding an optimal compromise in beat and bar alignment in the time and tempo domains. Time signature divisions are determined based on a new representation, termed the Metrogram, that presents time-varying information regarding rhythmic and metric periodicities in the Tempogram. Our methodology is characterised by its ability to provide a distribution over metric interpretations, offering insights into the diverse ways music can be rhythmically perceived. Results indicate high-level accuracy for a variety of polyphonic extracts containing irregular, complex, irrational, and time-varying time signatures. Accuracy rivalling state-of-the-art methodologies is also reported in a beat tracking task performed on the standard Ballroom Dataset. The paper offers insights into the field of dynamic time signature recognition and beat tracking, offering a valuable and versatile resource for the analysis, composition, and performance of music.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363392/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363392/</guid>
</item>
<item>
<title>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</title>
<description><p><span><big>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</big></span><br></p><p><span><small><i>Yuto Watanabe; Ren Togo; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343335</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Although text-guided image manipulation approaches have demonstrated highly accurate performance for editing the appearance of images in a virtual or simple scenario, their real-world applications face significant challenges. The primary cause of these challenges is the misalignment in the distribution of training and real-world data, which leads to unstable text-guided image manipulation. In this work, we propose a novel framework called TolerantGAN and tackle the new task of real-world text-guided image manipulation independent of the training data. To achieve this, we introduce two key concepts of a border smoothly connection module (BSCM) and a manipulation direction-based attention module (MDAM). BSCM smoothens the misalignment in the distribution of training and real-world data. MDAM extracts only regions highly relevant for image manipulation and assists in reconstructing unobserved objects in the training data. For in-the-wild input images of various classes, TolerantGAN robustly outperforms the state-of-the-art methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360283/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360283/</guid>
</item>
<item>
<title>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</title>
<description><p><span><big>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</big></span><br></p><p><span><small><i>Anastasia Avdeeva; Aleksei Gusev; Tseren Andzhukaev; Artem Ivanov</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343342</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Whispered speech is a quiet voice without vocalization. One of the common cases of using whispered speech is a technique that can help overcome stuttering. But whispered speech can be uncomfortable and difficult to understand in everyday communication. To address these problems, we propose a method of low-delayed whisper-to-speech voice conversion, which can be useful in real life communication of people with disordered speech. As part of our research, we study the impact of streaming Automatic Speech Recognition models on the quality of voice conversion, comparing different streaming models and methods for model adaptation to streaming settings, and showing the importance of using such models in cases of low-delayed voice conversion.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360259/</guid>
</item>
<item>
<title>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</title>
<description><p><span><big>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</big></span><br></p><p><span><small><i>Denis C. Ilie-Ablachim; Andra Băltoiu; Bogdan Dumitrescu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344313</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We propose sparse representation and dictionary learning algorithms for dictionaries whose atoms are characterized by Gaussian probability distributions around some central atoms. This extension of the space covered by the atoms permits a better characterization of localized features of the signals. The representations are computed by trading-off representation error and the probability of the actual atoms used in the representation. We propose two types of algorithms: a greedy one, similar in spirit with Orthogonal Matching Pursuit, and one based on L1-regularization. We apply our new algorithms to unsupervised anomaly detection, where the representation error is used as anomaly score. The flexibility of our approach appears to improve more the representations of the many normal signals than those of the few outliers, at least for an anomaly type called dependency, thus improving the detection quality. Comparison with both standard dictionary learning algorithms and established anomaly detection methods is favorable.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365333/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365333/</guid>
</item>
<item>
<title>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</title>
<description><p><span><big>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</big></span><br></p><p><span><small><i>Yuya Moroto; Yingrui Ye; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344079</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>There are various sentiment theories for categorizing human sentiments into several discrete sentiment categories, which means that the theory used for training sentiment prediction methods does not always match that used in the test phase. As a solution to this problem, zero-shot visual sentiment prediction methods have been proposed to predict unseen sentiments for which no images are available in the training phase. However, the training of these previous zero-shot methods relies on a single sentiment theory, which limits their ability to handle sentiments from other theories. Thus, this article proposes a more robust zero-shot visual sentiment prediction method that can handle cross-domain sentiments defined in different sentiment theories. Specifically, by focusing on the fact that sentiments are abstract concepts common to humans regardless of whether their theories are different, we incorporate knowledge distillation into our method to construct a teacher–student model that can train the implicit relationships between sentiments defined in different sentiment theories. Furthermore, to enhance sentiment discrimination capability and strengthen the implicit relationships between sentiments, we introduce a novel sentiment loss between the teacher and student models. In this way, our model becomes robust to unseen sentiments by exploiting the implicit relationships between sentiments. The contributions of this article are the introduction of knowledge distillation and a novel sentiment loss between the teacher and student models for zero-shot visual sentiment prediction, and improved performance of zero-shot visual sentiment prediction. Experiments on several open datasets demonstrate the effectiveness of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363382/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363382/</guid>
</item>
<item>
<title>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</title>
<description><p><span><big>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</big></span><br></p><p><span><small><i>Chengen Liu; Geert Leus; Elvin Isufi</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339376</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The edge flow reconstruction task consists of retreiving edge flow signals from corrupted or incomplete measurements. This is typically solved by a regularized optimization problem on higher-order networks such as simplicial complexes and the corresponding regularizers are chosen based on prior knowledge. Tailoring this prior to the setting of interest can be challenging or it may not even be possible. Thus, we consider to learn this prior knowledge via a model-based deep learning approach. We propose a new regularized optimization problem for the simplicial edge flow reconstruction task, the simplicial ElasticNet, which combines the advantages of the
$\ell _{1}$
and
$\ell _{2}$
norms. We solve the simplicial ElasticNet problem via the multi-block alternating direction method of multipliers (ADMM) algorithm and provide conditions on its convergence. By unrolling the ADMM iterative steps, we develop a model-based neural network with a low requirement on the number of training data. This unrolling network replaces the fixed parameters in the iterative algorithm by learnable weights, thus exploiting the neural network's learning capability while preserving the iterative algorithm's interpretability. We enhance this unrolling network via simplicial convolutional filters to aggregate information from the edge flow neighbors, ultimately, improving the network learning expressivity. Extensive experiments on real-world and synthetic datasets validate the proposed approaches and show considerable improvements over both baselines and traditional non-model-based neural networks.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342735/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342735/</guid>
</item>
<item>
<title>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</title>
<description><p><span><big>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</big></span><br></p><p><span><small><i>Jim Beckers; Bart Van Erp; Ziyue Zhao; Kirill Kondrashov; Bert De Vries</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337718</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Bayesian model reduction provides an efficient approach for comparing the performance of all nested sub-models of a model, without re-evaluating any of these sub-models. Until now, Bayesian model reduction has been applied mainly in the computational neuroscience community on simple models. In this paper, we formulate and apply Bayesian model reduction to perform principled pruning of Bayesian neural networks, based on variational free energy minimization. Direct application of Bayesian model reduction, however, gives rise to approximation errors. Therefore, a novel iterative pruning algorithm is presented to alleviate the problems arising with naive Bayesian model reduction, as supported experimentally on the publicly available UCI datasets for different inference algorithms. This novel parameter pruning scheme solves the shortcomings of current state-of-the-art pruning methods that are used by the signal processing community. The proposed approach has a clear stopping criterion and minimizes the same objective that is used during training. Next to these benefits, our experiments indicate better model performance in comparison to state-of-the-art pruning schemes.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334001/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334001/</guid>
</item>
<item>
<title>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</title>
<description><p><span><big>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</big></span><br></p><p><span><small><i>Anh Minh Truong; Wilfried Philips; Peter Veelaert</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340064</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Many image-based rendering (IBR) methods rely on depth estimates obtained from structured light or time-of-flight depth sensors to synthesize novel views from sparse camera networks. However, these estimates often contain missing or noisy regions, resulting in an incorrect mapping between source and target views. This situation makes the fusion process more challenging, as the visual information is misaligned, inconsistent, or missing. In this work, we first implement a lightweight network based on the transformer, which is well-known for its capability to model long-range relationships within input data, to extract spatial features from color images. These features are then used to enhance the quality of completed depth maps. Furthermore, we combine a sequential deep neural network with a spatial attention mechanism to effectively fuse the projected features from multiple source viewpoints. This approach enables us to integrate information from an arbitrary number of source viewpoints as well as improve accuracy in synthesized views. Experimental results on challenging datasets demonstrate that our method achieves superior synthesized image quality compared to state-of-the-art (SOTA) methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345792/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345792/</guid>
</item>
<item>
<title>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</title>
<description><p><span><big>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</big></span><br></p><p><span><small><i>Reza Mirzaeifard; Naveen K. D. Venkategowda; Vinay Chakravarthi Gogineni; Stefan Werner</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344395</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper investigates quantile regression in the presence of non-convex and non-smooth sparse penalties, such as the minimax concave penalty (MCP) and smoothly clipped absolute deviation (SCAD). The non-smooth and non-convex nature of these problems often leads to convergence difficulties for many algorithms. While iterative techniques such as coordinate descent and local linear approximation can facilitate convergence, the process is often slow. This sluggish pace is primarily due to the need to run these approximation techniques until full convergence at each step, a requirement we term as a
secondary convergence iteration
. To accelerate the convergence speed, we employ the alternating direction method of multipliers (ADMM) and introduce a novel single-loop smoothing ADMM algorithm with an increasing penalty parameter, named SIAD, specifically tailored for sparse-penalized quantile regression. We first delve into the convergence properties of the proposed SIAD algorithm and establish the necessary conditions for convergence. Theoretically, we confirm a convergence rate of
$o({k^{-\frac{1}{4}}})$
for the sub-gradient bound of the augmented Lagrangian, where
$k$
denotes the number of iterations. Subsequently, we provide numerical results to showcase the effectiveness of the SIAD algorithm. Our findings highlight that the SIAD method outperforms existing approaches, providing a faster and more stable solution for sparse-penalized quantile regression.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365338/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365338/</guid>
</item>
<item>
<title>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</title>
<description><p><span><big>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</big></span><br></p><p><span><small><i>Oliver Lang; Christian Hofbauer; Reinhard Feger; Mario Huemer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343308</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>A promising waveform candidate for future joint sensing and communication systems is orthogonal frequency-division multiplexing (OFDM). For such systems, supporting multiple transmit antennas requires multiplexing methods for the generation of orthogonal transmit signals, where equidistant subcarrier interleaving (ESI) is the most popular multiplexing method. In this work, we analyze a multiplexing method called Doppler-division multiplexing (DDM). This method applies a phase shift from OFDM symbol to OFDM symbol to separate signals transmitted by different Tx antennas along the velocity axis of the range-Doppler map. The main focus of this work lies on the implications of DDM on the communication task. It will be shown that for DDM, the channels observed in the communication receiver are heavily time-varying, preventing any meaningful transmission of data when not taken into account. In this work, a communication system designed to combat these time-varying channels is proposed, which includes methods for data estimation, synchronization, and channel estimation. Bit error ratio (BER) simulations demonstrate the superiority of this communications system compared to ESI-based systems.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360223/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360223/</guid>
</item>
<item>
<title>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</title>
<description><p><span><big>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</big></span><br></p><p><span><small><i>Nils L. Westhausen; Bernd T. Meyer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343320</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we introduce a causal low-latency low-complexity approach for binaural multichannel blind speaker separation in noisy reverberant conditions. The model, referred to as Group Communication Binaural Filter and Sum Network (GCBFSnet) predicts complex filters for filter-and-sum beamforming in the time-frequency domain. We apply Group Communication (GC), i.e., latent model variables are split into groups and processed with a shared sequence model with the aim of reducing the complexity of a simple model only containing one convolutional and one recurrent module. With GC we are able to reduce the size of the model by up to 83% and the complexity up to 73% compared to the model without GC, while mostly retaining performance. Even for the smallest model configuration, GCBFSnet matches the performance of a low-complexity TasNet baseline in most metrics despite the larger size and higher number of required operations of the baseline.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360275/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360275/</guid>
</item>
<item>
<title>Reverse Ordering Techniques for Attention-Based Channel Prediction</title>
<description><p><span><big>Reverse Ordering Techniques for Attention-Based Channel Prediction</big></span><br></p><p><span><small><i>Valentina Rizzello; Benedikt Böck; Michael Joham; Wolfgang Utschick</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344024</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Channel state information (CSI) is crucial for enhancing the performance of wireless systems by allowing to adjust the transmission strategies based on the current channel conditions. However, obtaining precise CSI is difficult because of the fast-changing channel conditions caused by multi-path fading. An inaccurate CSI hinders the performance of various adaptive wireless systems, highlighting the need for channel prediction techniques to effectively mitigate the drawbacks of outdated CSI. Conventional methods typically depend on assumptions regarding user velocity or require knowledge of the Doppler frequency. In contrast to existing approaches, we aim for a more robust and practical solution by training neural networks without making any assumptions about user velocity, relying solely on noisy channel observations during training. Specifically, we adapt both the sequence-to-sequence with attention (Seq2Seq-attn) and transformer models for channel prediction. Additionally, a new technique called reverse positional encoding is introduced in the transformer model to improve the robustness of the model against varying sequence lengths. Similarly, the encoder outputs of the Seq2Seq-attn model are reversed prior to the application of attention mechanisms. By means of simulations, we show that these proposed techniques enable the models to effectively capture relationships within sequences of channel snapshots without increasing the complexity. Importantly, this capability remains robust across varying sequence lengths, representing a substantial improvement over existing methodologies.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363354/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363354/</guid>
</item>
<item>
<title>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</title>
<description><p><span><big>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</big></span><br></p><p><span><small><i>Sarina Meyer; Xiaoxiao Miao; Ngoc Thang Vu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344375</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Speaker anonymization is the task of modifying a speech recording such that the original speaker cannot be identified anymore. Since the first Voice Privacy Challenge in 2020, along with the release of a framework, the popularity of this research topic is continually increasing. However, the comparison and combination of different anonymization approaches remains challenging due to the complexity of evaluation and the absence of user-friendly research frameworks. We therefore propose an efficient speaker anonymization and evaluation framework based on a modular and easily extendable structure, almost fully in Python. The framework facilitates the orchestration of several anonymization approaches in parallel and allows for interfacing between different techniques. Furthermore, we propose modifications to common evaluation methods which improves the quality of the evaluation and reduces their computation time by 65 to 95%, depending on the metric. Our code is fully open source.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365329/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365329/</guid>
</item>
<item>
<title>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</title>
<description><p><span><big>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</big></span><br></p><p><span><small><i>Alessandro Ilic Mezza; Matteo Amerena; Alberto Bernardini; Augusto Sarti</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343318</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Real-time audio communications over IP have become essential to our daily lives. Packet-switched networks, however, are inherently prone to jitter and data losses, thus creating a strong need for effective packet loss concealment (PLC) techniques. Though solutions based on deep learning have made significant progress in that direction as far as speech is concerned, extending the use of such methods to applications of Networked Music Performance (NMP) presents significant challenges, including high fidelity requirements, higher sampling rates, and stringent temporal constraints associated to the simultaneous interaction between remote musicians. In this article, we present PARCnet, a hybrid PLC method that utilizes a feed-forward neural network to estimate the time-domain residual signal of a parallel linear autoregressive model. Objective metrics and a listening test show that PARCnet provides state-of-the-art results while enabling real-time operation on CPU.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360264/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360264/</guid>
</item>
<item>
<title>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</title>
<description><p><span><big>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</big></span><br></p><p><span><small><i>Katerina Zmolikova; Michael Syskind Pedersen; Jesper Jensen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Supervised learning-based speech enhancement methods often work remarkably well in acoustic situations represented in the training corpus but generalize poorly to out-of-domain situations, i.e. situations not seen during training. This stands in the way of further improvement of these methods in realistic scenarios, as collecting paired noisy-clean recordings in the target application domain is typically not feasible. Recording noisy-only in-domain data is, though, much more practical. In this article, we tackle the problem of unsupervised domain adaptation in speech enhancement. Specifically, we propose a way to use in-domain noisy-only data in the training of a neural network to improve upon a model trained solely on out-of-domain paired data. For this, we make use of masked spectrogram prediction, a technique from self-supervised learning that aims to interpolate masked regions of a spectrogram. We hypothesize that masked spectrogram prediction encourages learning of features that represent well both speech and noise components of the noisy signals. These features can then be used to train a more robust speech enhancement system. We evaluate the proposed method on the VoiceBank-DEMAND and LibriFSD50k databases, with WSJ0-CHiME3 serving as the out-of-domain database. We show that the proposed method outperforms both the out-of-domain system and the baseline approaches, i.e. RemixIT and noisy-target training, and also combines well with the previously proposed RemixIT method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360251/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360251/</guid>
</item>
<item>
<title>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</title>
<description><p><span><big>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</big></span><br></p><p><span><small><i>Alexander Karataev; Christian Forsch; Laura Cottatellucci</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3348343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We consider a cell-free massive multiple-input multiple-output (CF-MaMIMO) communication system in the uplink transmission and propose a novel algorithm for blind or semi-blind joint channel estimation and data detection (JCD). We formulate the problem in the framework of bilinear inference and develop a solution based on the expectation propagation (EP) method for both channel estimation and data detection. We propose a new approximation of the joint a posteriori distribution of the channel and data whose representation as a factor graph enables the application of the EP approach using the message-passing technique, local low-complexity computations at the nodes, and an effective modeling of channel-data interplay. The derived algorithm, called bilinear-EP JCD, allows for a distributed implementation among access points (APs) and the central processing unit (CPU) and has polynomial complexity. Our simulation results show that it outperforms other EP-based state-of-the-art polynomial time algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10378663/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10378663/</guid>
</item>
<item>
<title>Adversarial Representation Learning for Robust Privacy Preservation in Audio</title>
<description><p><span><big>Adversarial Representation Learning for Robust Privacy Preservation in Audio</big></span><br></p><p><span><small><i>Shayan Gharib; Minh Tran; Diep Luong; Konstantinos Drossos; Tuomas Virtanen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Sound event detection systems are widely used in various applications such as surveillance and environmental monitoring where data is automatically collected, processed, and sent to a cloud for sound recognition. However, this process may inadvertently reveal sensitive information about users or their surroundings, hence raising privacy concerns. In this study, we propose a novel adversarial training method for learning representations of audio recordings that effectively prevents the detection of speech activity from the latent features of the recordings. The proposed method trains a model to generate invariant latent representations of speech-containing audio recordings that cannot be distinguished from non-speech recordings by a speech classifier. The novelty of our work is in the optimization algorithm, where the speech classifier's weights are regularly replaced with the weights of classifiers trained in a supervised manner. This increases the discrimination power of the speech classifier constantly during the adversarial training, motivating the model to generate latent representations in which speech is not distinguishable, even using new speech classifiers trained outside the adversarial training loop. The proposed method is evaluated against a baseline approach with no privacy measures and a prior adversarial training method, demonstrating a significant reduction in privacy violations compared to the baseline approach. Additionally, we show that the prior adversarial method is practically ineffective for this purpose.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10379095/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10379095/</guid>
</item>
<item>
<title>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</title>
<description><p><span><big>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</big></span><br></p><p><span><small><i>Yasaman Parhizkar; Gene Cheung; Andrew W. Eckford</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>It is a popular hypothesis in neuroscience that ganglion cells in the retina are activated by selectively detecting visual features in an observed scene. While ganglion cell firings can be predicted via data-trained deep neural nets, the n |
|
Successfully generated as following: http://localhost:1200/ieee/journal/8782710 - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Open Journal of Signal Processing</title>
<link>https://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=8782710</link>
<atom:link href="http://localhost:1200/ieee/journal/8782710" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Open Journal of Signal Processing - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Tue, 09 Apr 2024 08:29:50 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Synthbuster: Towards Detection of Diffusion Model Generated Images</title>
<description><p><span><big>Synthbuster: Towards Detection of Diffusion Model Generated Images</big></span><br></p><p><span><small><i>Quentin Bammey</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337714</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Synthetically-generated images are getting increasingly popular. Diffusion models have advanced to the stage where even non-experts can generate photo-realistic images from a simple text prompt. They expand creative horizons but also open a Pandora's box of potential disinformation risks. In this context, the present corpus of synthetic image detection techniques, primarily focusing on older generative models like Generative Adversarial Networks, finds itself ill-equipped to deal with this emerging trend. Recognizing this challenge, we introduce a method specifically designed to detect synthetic images produced by diffusion models. Our approach capitalizes on the inherent frequency artefacts left behind during the diffusion process. Spectral analysis is used to highlight the artefacts in the Fourier transform of a residual image, which are used to distinguish real from fake images. The proposed method can detect diffusion-model-generated images even under mild
jpeg
compression, and generalizes relatively well to unknown models. By pioneering this novel approach, we aim to fortify forensic methodologies and ignite further research into the detection of AI-generated images.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334046/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334046/</guid>
</item>
<item>
<title>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</title>
<description><p><span><big>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</big></span><br></p><p><span><small><i>Yanbin Zou; Jingna Fan; Zekai Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we consider using time-of-arrival (TOA) measurements from a single moving receiver to locate a moving target at constant velocity that emits a periodic signal with unknown signal period. First, we give the TOA measurement model and deduce the Cram
$\acute{\text{e}}$
r-Rao lower bounds (CRLB). Then, we formulate a nonlinear least squares (NLS) problem to estimate the unknown parameters. We use semidefinite programming (SDP) techniques to relax the nonconvex NLS problem. However, it is shown that the SDP localization algorithm cannot provide a high-quality solution. Subsequently, we develop a fixed point iteration (FPI) method to improve the performance of the SDP algorithm. In addition, we also consider the presence of receiver position errors and develop the corresponding localization algorithm. Numerical simulations are conducted to demonstrate the localization performance of the proposed algorithms by comparing them with the CRLB.
Index Term
-Fixed point iteration (FPI), semidefinite programming (SDP), single moving receiver, target localization, time-of-arrival (TOA).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336384/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336384/</guid>
</item>
<item>
<title>The Neural-SRP Method for Universal Robust Multi-Source Tracking</title>
<description><p><span><big>The Neural-SRP Method for Universal Robust Multi-Source Tracking</big></span><br></p><p><span><small><i>Eric Grinstein; Christopher M. Hicks; Toon van Waterschoot; Mike Brookes; Patrick A. Naylor</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340057</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Neural networks have achieved state-of-the-art performance on the task of acoustic Direction-of-Arrival (DOA) estimation using microphone arrays. Neural models can be classified as end-to-end or hybrid, each class showing advantages and disadvantages. This work introduces Neural-SRP, an end-to-end neural network architecture for DOA estimation inspired by the classical Steered Response Power (SRP) method, which overcomes limitations of current neural models. We evaluate the architecture on multiple scenarios, namely, multi-source DOA tracking and single-source DOA tracking under the presence of directional and diffuse noise. The experiments demonstrate that our proposed method compares favourably in terms of computational and localization performance with established neural methods on various recorded and simulated benchmark datasets.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345765/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345765/</guid>
</item>
<item>
<title>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</title>
<description><p><span><big>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</big></span><br></p><p><span><small><i>Alexander Venus; Erik Leitinger; Stefan Tertinek; Klaus Witrisal</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper presents a neural-enhanced probabilistic model and corresponding factor graph-based sum-product algorithm for robust localization and tracking in multipath-prone environments. The introduced hybrid probabilistic model consists of physics-based and data-driven measurement models capturing the information contained in both, the line-of-sight (LOS) component as well as in multipath components (NLOS components). The physics-based and data-driven models are embedded in a joint Bayesian framework allowing to derive from first principles a factor graph-based algorithm that fuses the information of these models. The proposed algorithm uses radio signal measurements from multiple base stations to robustly estimate the mobile agent's position together with all model parameters. It provides high localization accuracy by exploiting the position-related information of the LOS component via the physics-based model and robustness by exploiting the geometric imprint of multipath components independent of the propagation channel via the data-driven model. In a challenging numerical experiment involving obstructed LOS situations to all anchors, we show that the proposed sequential algorithm significantly outperforms state-of-the-art methods and attains the posterior Cramér-Rao lower bound even with training data limited to local regions.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336409/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336409/</guid>
</item>
<item>
<title>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</title>
<description><p><span><big>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</big></span><br></p><p><span><small><i>AMIT EFRAIM; Joseph M. Francos</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In solving a model fitting problem, the existence of outliers in the set of measurements can have a devastating effect on the solution accuracy. Traditionally, in order to overcome this problem, robust point cloud registration algorithms are composed of transformation hypothesis generation, followed by hypothesis evaluation aimed at selecting the best hypothesized estimate. Hypotheses evaluation is commonly performed using the sample consensus criterion. However, since this criterion accounts only for the consensus size, it fails when the maximal sample consensus is incorrect. We propose a new hypothesis evaluation approach, generalizing the sample consensus approach, where instead of the sample consensus, the transformation that maximizes the point clouds feature correlation is selected as the best hypothesis. The feature vector at each point contains information such as on local geometry and semantic context. Utilizing this information in the hypotheses evaluation and selection procedure allows for a correct decision even when the hypothesis yielding the maximal sample consensus is false. Consequently, the probability of selecting the correct model increases. We show both mathematically and empirically that substituting the sample consensus criterion with the proposed point cloud feature correlation hypothesis test (PC-FCHT) lowers the probability of large registration errors, compared to using the special case of sample consensus. The proposed PC-FCHT is applicable to any algorithm that follows the hypothesis generation and evaluation scheme, potentially improving the success rates of a wide family of point cloud registration algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345750/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345750/</guid>
</item>
<item>
<title>Joint PAPR and OBP Reduction for NC-OFDM Systems</title>
<description><p><span><big>Joint PAPR and OBP Reduction for NC-OFDM Systems</big></span><br></p><p><span><small><i>Hsuan-Fu Wang; Fang-Biau Ueng; Bo-Heng Yeh</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3329757</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The spectrum resource is always a critical issue for wireless communications since it directly impacts the data rate and capacity. However, the problem of spectrum resource scarcity always exists. Moreover, spectrum resource scarcity becomes more severe as new communication technologies and wireless applications sprout. Noncontiguous orthogonal frequency division multiplexing (NC-OFDM) is a multicarrier method for bandwidth utilization. Unfortunately, this system has two fatal defects: high peak-to-average power ratio (PAPR) and considerable out-of-band power (OBP), which are detrimental to the system's performance. To solve these two problems, we propose a convex optimization-based method for joint PAPR and OBP reduction in NC-OFDM Systems. The strategy is to permit the secondary user to utilize the unoccupied spectrum of the primary user with dynamic spectrum sharing (DSS) based on a cognitive radio network (CRN). To this end, a flexible system operating over noncontiguous bands and DSS scenarios is necessary. The simulation results have shown that our method could effectively improve the overall performance and outperform other schemes, i.e., projections onto convex sets (POCS) and alternating projections onto convex and non-convex sets (APOCNCS), without harming the transmission of the primary system. The collaboration between secondary and primary systems is viable with the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10305259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10305259/</guid>
</item>
<item>
<title>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</title>
<description><p><span><big>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</big></span><br></p><p><span><small><i>Gal Itzhak; Israel Cohen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339433</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article presents a Kronecker-product (KP) beamforming approach incorporating sparse concentric circular arrays (SCCAs). The locations of the microphones on the SCCA are optimized concerning the broadband array directivity over a wide range of direction-of-arrival (DOA) deviations of a desired signal. A maximum directivity factor (MDF) sub-beamformer is derived accordingly with the optimal locations. Then, we propose two global beamformers obtained as a Kronecker product of a uniform linear array (ULA) and the SCCA sub-beamformer. The global beamformers differ by the type of the ULA, which is designed either as an MDF sub-beamformer along the
$\mathsf {x}$
-axis or as a maximum white noise gain sub-beamformer along the
$\mathsf {y}$
-axis. We analyze the performance of the proposed beamformers in terms of the directivity factor, the white noise gain, and their spatial beampatterns. Compared to traditional beamformers, the proposed beamformers exhibit considerably larger tolerance to DOA deviations concerning both the azimuth and elevation angles. Experimental results with speech signals in noisy and reverberant environments demonstrate that the proposed approach outperforms traditional beamformers regarding the perceptual evaluation of speech quality (PESQ) and short-time objective intelligibility (STOI) scores when the desired speech signals deviate from the nominal DOA.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342869/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342869/</guid>
</item>
<item>
<title>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</title>
<description><p><span><big>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</big></span><br></p><p><span><small><i>Karn N. Watcharasupat; Chih-Wei Wu; Yiwei Ding; Iroro Orife; Aaron J. Hipple; Phillip A. Williams; Scott Kramer; Alexander Lerch; William Wolcott</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339428</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Cinematic audio source separation is a relatively new subtask of audio source separation, with the aim of extracting the dialogue, music, and effects stems from their mixture. In this work, we developed a model generalizing the Bandsplit RNN for any complete or overcomplete partitions of the frequency axis. Psychoacoustically motivated frequency scales were used to inform the band definitions which are now defined with redundancy for more reliable feature extraction. A loss function motivated by the signal-to-noise ratio and the sparsity-promoting property of the 1-norm was proposed. We additionally exploit the information-sharing property of a common-encoder setup to reduce computational complexity during both training and inference, improve separation performance for hard-to-generalize classes of sounds, and allow flexibility during inference time with detachable decoders. Our best model sets the state of the art on the Divide and Remaster dataset with performance above the ideal ratio mask for the dialogue stem.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342812/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342812/</guid>
</item>
<item>
<title>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</title>
<description><p><span><big>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</big></span><br></p><p><span><small><i>Miguel Ferrer; María de Diego; Alberto Gonzalez</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340106</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The LMS algorithm is widely employed in adaptive systems due to its robustness, simplicity, and reasonable performance. However, it is well known that this algorithm suffers from a slow convergence speed when dealing with colored reference signals. Numerous variants and alternative algorithms have been proposed to address this issue, though all of them entail an increase in computational cost. Among the proposed alternatives, the affine projection algorithm stands out. This algorithm has the peculiarity of starting from
$N$
data vectors of the reference signal. It transforms these vectors into as many data vectors suitably normalized in energy and mutually orthogonal. In this work, we propose a version of the LMS algorithm that, similar to the affine projection algorithm, starts from
$N$
data vectors of the reference signal but corrects them by using only a scalar factor that functions as a convergence step. Our goal is to align the behavior of this algorithm with the behavior of the affine projection algorithm without significantly increasing the computational cost of the LMS.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345730/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345730/</guid>
</item>
<item>
<title>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</title>
<description><p><span><big>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</big></span><br></p><p><span><small><i>Zongyao Li; Ren Togo; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340616</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article proposes a method for transferring knowledge of semantic segmentation from a labeled source domain to an unlabeled target domain without using the source-domain data. Such a problem is called source-data-free domain adaptation, in which a pre-trained source-domain model and the unlabeled target-domain data are used to transfer the label knowledge across the domains. Like most previous methods, our method uses pseudo labels for distilling and transferring the source-domain knowledge. On the basis of the pseudo-label learning, our method improves the domain adaptation performance in two innovative ways: 1) reducing the domain differences by source-data-free style transfer and 2) exploring the style diversity within the target domain by style modification. To this end, we introduce two additional modules: 1) an inter-domain style transfer module which aligns the feature statistics of the source and target domains before producing the pseudo labels thereby improving the pseudo labels' accuracy, and 2) an intra-domain style modification module which modifies the image styles within the target domain for learning intra-domain style-invariant features. Our method with the two modules outperforms previous source-data-free domain adaptation methods in two commonly used benchmarks. Moreover, our method is well compatible with the previous methods for further improvement.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10356748/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10356748/</guid>
</item>
<item>
<title>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</title>
<description><p><span><big>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</big></span><br></p><p><span><small><i>Tobias Kabzinski; Peter Jax</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337721</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Quasi-continuous system identification of time-variant linear acoustic systems can be applied in various audio signal processing applications when numerous acoustic transfer functions must be measured. A prominent application is measuring head-related transfer functions. We treat the underlying multiple-input-multiple-output (MIMO) system identification problem in a state-space model as a joint estimation problem for states, representing impulse responses, and state-space model parameters using the expectation maximization (EM) algorithm. We address limitations of prior work by imposing different model structures, especially for dependencies within a (transformed) state vector. This results in block diagonal matrix structures, for which we derive M-step update rules. Making assumptions about this model structure and choosing a block size for a given application define the computational complexity. In examples, we found that applying this framework yields improvements of up to 10 dB in relative system distance in comparison to a conventional method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334061/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334061/</guid>
</item>
<item>
<title>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</title>
<description><p><span><big>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</big></span><br></p><p><span><small><i>Damir Rakhimov; Martin Haardt</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337729</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this paper, we present an analytical performance assessment of 2-D Tensor ESPRIT in terms of physical parameters. We show that the error in the
$r$
-mode depends only on two components, irrespective of the dimensionality of the problem. We obtain analytical expressions in closed form for the mean squared error (MSE) in each dimension as a function of the signal-to-noise (SNR) ratio, the array steering matrices, the number of antennas, the number of snapshots, the selection matrices, and the signal correlation. The derived expressions allow a better understanding of the difference in performance between the tensor and the matrix versions of the ESPRIT algorithm. The simulation results confirm the coincidence between the presented analytical expression and the curves obtained via Monte Carlo trials. We analyze the behavior of each of the two error components in different scenarios.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334446/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334446/</guid>
</item>
<item>
<title>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</title>
<description><p><span><big>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</big></span><br></p><p><span><small><i>Yaman Kındap; Simon Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343341</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this work we study linear vector stochastic differential equation (SDE) models driven by the generalised hyperbolic (GH) Lévy process for inference in continuous-time non-Gaussian filtering problems. The GH family of stochastic processes offers a flexible framework for modelling of non-Gaussian, heavy-tailed characteristics and includes the normal inverse-Gaussian, variance-gamma and Student-t processes as special cases. We present continuous-time simulation methods for the solution of vector SDE models driven by GH processes and novel inference methodologies using a variant of sequential Markov chain Monte Carlo (MCMC). As an example a particular formulation of Langevin dynamics is studied within this framework. The model is applied to both a synthetically generated data set and a real-world financial series to demonstrate its capabilities.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360268/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360268/</guid>
</item>
<item>
<title>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</title>
<description><p><span><big>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</big></span><br></p><p><span><small><i>James M. Cozens; Simon J. Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344048</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper proposes a probabilistic approach for extracting time-varying and irregular time signature information from polyphonic audio extracts, subsequently providing beat and bar line positions given inferred time signature divisions. This is achieved via dynamically evaluating the beat tempo as a function of time through finding an optimal compromise in beat and bar alignment in the time and tempo domains. Time signature divisions are determined based on a new representation, termed the Metrogram, that presents time-varying information regarding rhythmic and metric periodicities in the Tempogram. Our methodology is characterised by its ability to provide a distribution over metric interpretations, offering insights into the diverse ways music can be rhythmically perceived. Results indicate high-level accuracy for a variety of polyphonic extracts containing irregular, complex, irrational, and time-varying time signatures. Accuracy rivalling state-of-the-art methodologies is also reported in a beat tracking task performed on the standard Ballroom Dataset. The paper offers insights into the field of dynamic time signature recognition and beat tracking, offering a valuable and versatile resource for the analysis, composition, and performance of music.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363392/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363392/</guid>
</item>
<item>
<title>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</title>
<description><p><span><big>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</big></span><br></p><p><span><small><i>Yuto Watanabe; Ren Togo; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343335</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Although text-guided image manipulation approaches have demonstrated highly accurate performance for editing the appearance of images in a virtual or simple scenario, their real-world applications face significant challenges. The primary cause of these challenges is the misalignment in the distribution of training and real-world data, which leads to unstable text-guided image manipulation. In this work, we propose a novel framework called TolerantGAN and tackle the new task of real-world text-guided image manipulation independent of the training data. To achieve this, we introduce two key concepts of a border smoothly connection module (BSCM) and a manipulation direction-based attention module (MDAM). BSCM smoothens the misalignment in the distribution of training and real-world data. MDAM extracts only regions highly relevant for image manipulation and assists in reconstructing unobserved objects in the training data. For in-the-wild input images of various classes, TolerantGAN robustly outperforms the state-of-the-art methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360283/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360283/</guid>
</item>
<item>
<title>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</title>
<description><p><span><big>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</big></span><br></p><p><span><small><i>Anastasia Avdeeva; Aleksei Gusev; Tseren Andzhukaev; Artem Ivanov</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343342</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Whispered speech is a quiet voice without vocalization. One of the common cases of using whispered speech is a technique that can help overcome stuttering. But whispered speech can be uncomfortable and difficult to understand in everyday communication. To address these problems, we propose a method of low-delayed whisper-to-speech voice conversion, which can be useful in real life communication of people with disordered speech. As part of our research, we study the impact of streaming Automatic Speech Recognition models on the quality of voice conversion, comparing different streaming models and methods for model adaptation to streaming settings, and showing the importance of using such models in cases of low-delayed voice conversion.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360259/</guid>
</item>
<item>
<title>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</title>
<description><p><span><big>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</big></span><br></p><p><span><small><i>Denis C. Ilie-Ablachim; Andra Băltoiu; Bogdan Dumitrescu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344313</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We propose sparse representation and dictionary learning algorithms for dictionaries whose atoms are characterized by Gaussian probability distributions around some central atoms. This extension of the space covered by the atoms permits a better characterization of localized features of the signals. The representations are computed by trading-off representation error and the probability of the actual atoms used in the representation. We propose two types of algorithms: a greedy one, similar in spirit with Orthogonal Matching Pursuit, and one based on L1-regularization. We apply our new algorithms to unsupervised anomaly detection, where the representation error is used as anomaly score. The flexibility of our approach appears to improve more the representations of the many normal signals than those of the few outliers, at least for an anomaly type called dependency, thus improving the detection quality. Comparison with both standard dictionary learning algorithms and established anomaly detection methods is favorable.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365333/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365333/</guid>
</item>
<item>
<title>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</title>
<description><p><span><big>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</big></span><br></p><p><span><small><i>Yuya Moroto; Yingrui Ye; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344079</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>There are various sentiment theories for categorizing human sentiments into several discrete sentiment categories, which means that the theory used for training sentiment prediction methods does not always match that used in the test phase. As a solution to this problem, zero-shot visual sentiment prediction methods have been proposed to predict unseen sentiments for which no images are available in the training phase. However, the training of these previous zero-shot methods relies on a single sentiment theory, which limits their ability to handle sentiments from other theories. Thus, this article proposes a more robust zero-shot visual sentiment prediction method that can handle cross-domain sentiments defined in different sentiment theories. Specifically, by focusing on the fact that sentiments are abstract concepts common to humans regardless of whether their theories are different, we incorporate knowledge distillation into our method to construct a teacher–student model that can train the implicit relationships between sentiments defined in different sentiment theories. Furthermore, to enhance sentiment discrimination capability and strengthen the implicit relationships between sentiments, we introduce a novel sentiment loss between the teacher and student models. In this way, our model becomes robust to unseen sentiments by exploiting the implicit relationships between sentiments. The contributions of this article are the introduction of knowledge distillation and a novel sentiment loss between the teacher and student models for zero-shot visual sentiment prediction, and improved performance of zero-shot visual sentiment prediction. Experiments on several open datasets demonstrate the effectiveness of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363382/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363382/</guid>
</item>
<item>
<title>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</title>
<description><p><span><big>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</big></span><br></p><p><span><small><i>Chengen Liu; Geert Leus; Elvin Isufi</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339376</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The edge flow reconstruction task consists of retreiving edge flow signals from corrupted or incomplete measurements. This is typically solved by a regularized optimization problem on higher-order networks such as simplicial complexes and the corresponding regularizers are chosen based on prior knowledge. Tailoring this prior to the setting of interest can be challenging or it may not even be possible. Thus, we consider to learn this prior knowledge via a model-based deep learning approach. We propose a new regularized optimization problem for the simplicial edge flow reconstruction task, the simplicial ElasticNet, which combines the advantages of the
$\ell _{1}$
and
$\ell _{2}$
norms. We solve the simplicial ElasticNet problem via the multi-block alternating direction method of multipliers (ADMM) algorithm and provide conditions on its convergence. By unrolling the ADMM iterative steps, we develop a model-based neural network with a low requirement on the number of training data. This unrolling network replaces the fixed parameters in the iterative algorithm by learnable weights, thus exploiting the neural network's learning capability while preserving the iterative algorithm's interpretability. We enhance this unrolling network via simplicial convolutional filters to aggregate information from the edge flow neighbors, ultimately, improving the network learning expressivity. Extensive experiments on real-world and synthetic datasets validate the proposed approaches and show considerable improvements over both baselines and traditional non-model-based neural networks.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342735/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342735/</guid>
</item>
<item>
<title>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</title>
<description><p><span><big>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</big></span><br></p><p><span><small><i>Jim Beckers; Bart Van Erp; Ziyue Zhao; Kirill Kondrashov; Bert De Vries</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337718</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Bayesian model reduction provides an efficient approach for comparing the performance of all nested sub-models of a model, without re-evaluating any of these sub-models. Until now, Bayesian model reduction has been applied mainly in the computational neuroscience community on simple models. In this paper, we formulate and apply Bayesian model reduction to perform principled pruning of Bayesian neural networks, based on variational free energy minimization. Direct application of Bayesian model reduction, however, gives rise to approximation errors. Therefore, a novel iterative pruning algorithm is presented to alleviate the problems arising with naive Bayesian model reduction, as supported experimentally on the publicly available UCI datasets for different inference algorithms. This novel parameter pruning scheme solves the shortcomings of current state-of-the-art pruning methods that are used by the signal processing community. The proposed approach has a clear stopping criterion and minimizes the same objective that is used during training. Next to these benefits, our experiments indicate better model performance in comparison to state-of-the-art pruning schemes.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334001/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334001/</guid>
</item>
<item>
<title>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</title>
<description><p><span><big>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</big></span><br></p><p><span><small><i>Anh Minh Truong; Wilfried Philips; Peter Veelaert</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340064</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Many image-based rendering (IBR) methods rely on depth estimates obtained from structured light or time-of-flight depth sensors to synthesize novel views from sparse camera networks. However, these estimates often contain missing or noisy regions, resulting in an incorrect mapping between source and target views. This situation makes the fusion process more challenging, as the visual information is misaligned, inconsistent, or missing. In this work, we first implement a lightweight network based on the transformer, which is well-known for its capability to model long-range relationships within input data, to extract spatial features from color images. These features are then used to enhance the quality of completed depth maps. Furthermore, we combine a sequential deep neural network with a spatial attention mechanism to effectively fuse the projected features from multiple source viewpoints. This approach enables us to integrate information from an arbitrary number of source viewpoints as well as improve accuracy in synthesized views. Experimental results on challenging datasets demonstrate that our method achieves superior synthesized image quality compared to state-of-the-art (SOTA) methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345792/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345792/</guid>
</item>
<item>
<title>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</title>
<description><p><span><big>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</big></span><br></p><p><span><small><i>Reza Mirzaeifard; Naveen K. D. Venkategowda; Vinay Chakravarthi Gogineni; Stefan Werner</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344395</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper investigates quantile regression in the presence of non-convex and non-smooth sparse penalties, such as the minimax concave penalty (MCP) and smoothly clipped absolute deviation (SCAD). The non-smooth and non-convex nature of these problems often leads to convergence difficulties for many algorithms. While iterative techniques such as coordinate descent and local linear approximation can facilitate convergence, the process is often slow. This sluggish pace is primarily due to the need to run these approximation techniques until full convergence at each step, a requirement we term as a
secondary convergence iteration
. To accelerate the convergence speed, we employ the alternating direction method of multipliers (ADMM) and introduce a novel single-loop smoothing ADMM algorithm with an increasing penalty parameter, named SIAD, specifically tailored for sparse-penalized quantile regression. We first delve into the convergence properties of the proposed SIAD algorithm and establish the necessary conditions for convergence. Theoretically, we confirm a convergence rate of
$o({k^{-\frac{1}{4}}})$
for the sub-gradient bound of the augmented Lagrangian, where
$k$
denotes the number of iterations. Subsequently, we provide numerical results to showcase the effectiveness of the SIAD algorithm. Our findings highlight that the SIAD method outperforms existing approaches, providing a faster and more stable solution for sparse-penalized quantile regression.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365338/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365338/</guid>
</item>
<item>
<title>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</title>
<description><p><span><big>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</big></span><br></p><p><span><small><i>Oliver Lang; Christian Hofbauer; Reinhard Feger; Mario Huemer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343308</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>A promising waveform candidate for future joint sensing and communication systems is orthogonal frequency-division multiplexing (OFDM). For such systems, supporting multiple transmit antennas requires multiplexing methods for the generation of orthogonal transmit signals, where equidistant subcarrier interleaving (ESI) is the most popular multiplexing method. In this work, we analyze a multiplexing method called Doppler-division multiplexing (DDM). This method applies a phase shift from OFDM symbol to OFDM symbol to separate signals transmitted by different Tx antennas along the velocity axis of the range-Doppler map. The main focus of this work lies on the implications of DDM on the communication task. It will be shown that for DDM, the channels observed in the communication receiver are heavily time-varying, preventing any meaningful transmission of data when not taken into account. In this work, a communication system designed to combat these time-varying channels is proposed, which includes methods for data estimation, synchronization, and channel estimation. Bit error ratio (BER) simulations demonstrate the superiority of this communications system compared to ESI-based systems.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360223/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360223/</guid>
</item>
<item>
<title>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</title>
<description><p><span><big>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</big></span><br></p><p><span><small><i>Nils L. Westhausen; Bernd T. Meyer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343320</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we introduce a causal low-latency low-complexity approach for binaural multichannel blind speaker separation in noisy reverberant conditions. The model, referred to as Group Communication Binaural Filter and Sum Network (GCBFSnet) predicts complex filters for filter-and-sum beamforming in the time-frequency domain. We apply Group Communication (GC), i.e., latent model variables are split into groups and processed with a shared sequence model with the aim of reducing the complexity of a simple model only containing one convolutional and one recurrent module. With GC we are able to reduce the size of the model by up to 83% and the complexity up to 73% compared to the model without GC, while mostly retaining performance. Even for the smallest model configuration, GCBFSnet matches the performance of a low-complexity TasNet baseline in most metrics despite the larger size and higher number of required operations of the baseline.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360275/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360275/</guid>
</item>
<item>
<title>Reverse Ordering Techniques for Attention-Based Channel Prediction</title>
<description><p><span><big>Reverse Ordering Techniques for Attention-Based Channel Prediction</big></span><br></p><p><span><small><i>Valentina Rizzello; Benedikt Böck; Michael Joham; Wolfgang Utschick</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344024</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Channel state information (CSI) is crucial for enhancing the performance of wireless systems by allowing to adjust the transmission strategies based on the current channel conditions. However, obtaining precise CSI is difficult because of the fast-changing channel conditions caused by multi-path fading. An inaccurate CSI hinders the performance of various adaptive wireless systems, highlighting the need for channel prediction techniques to effectively mitigate the drawbacks of outdated CSI. Conventional methods typically depend on assumptions regarding user velocity or require knowledge of the Doppler frequency. In contrast to existing approaches, we aim for a more robust and practical solution by training neural networks without making any assumptions about user velocity, relying solely on noisy channel observations during training. Specifically, we adapt both the sequence-to-sequence with attention (Seq2Seq-attn) and transformer models for channel prediction. Additionally, a new technique called reverse positional encoding is introduced in the transformer model to improve the robustness of the model against varying sequence lengths. Similarly, the encoder outputs of the Seq2Seq-attn model are reversed prior to the application of attention mechanisms. By means of simulations, we show that these proposed techniques enable the models to effectively capture relationships within sequences of channel snapshots without increasing the complexity. Importantly, this capability remains robust across varying sequence lengths, representing a substantial improvement over existing methodologies.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363354/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363354/</guid>
</item>
<item>
<title>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</title>
<description><p><span><big>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</big></span><br></p><p><span><small><i>Sarina Meyer; Xiaoxiao Miao; Ngoc Thang Vu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344375</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Speaker anonymization is the task of modifying a speech recording such that the original speaker cannot be identified anymore. Since the first Voice Privacy Challenge in 2020, along with the release of a framework, the popularity of this research topic is continually increasing. However, the comparison and combination of different anonymization approaches remains challenging due to the complexity of evaluation and the absence of user-friendly research frameworks. We therefore propose an efficient speaker anonymization and evaluation framework based on a modular and easily extendable structure, almost fully in Python. The framework facilitates the orchestration of several anonymization approaches in parallel and allows for interfacing between different techniques. Furthermore, we propose modifications to common evaluation methods which improves the quality of the evaluation and reduces their computation time by 65 to 95%, depending on the metric. Our code is fully open source.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365329/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365329/</guid>
</item>
<item>
<title>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</title>
<description><p><span><big>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</big></span><br></p><p><span><small><i>Alessandro Ilic Mezza; Matteo Amerena; Alberto Bernardini; Augusto Sarti</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343318</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Real-time audio communications over IP have become essential to our daily lives. Packet-switched networks, however, are inherently prone to jitter and data losses, thus creating a strong need for effective packet loss concealment (PLC) techniques. Though solutions based on deep learning have made significant progress in that direction as far as speech is concerned, extending the use of such methods to applications of Networked Music Performance (NMP) presents significant challenges, including high fidelity requirements, higher sampling rates, and stringent temporal constraints associated to the simultaneous interaction between remote musicians. In this article, we present PARCnet, a hybrid PLC method that utilizes a feed-forward neural network to estimate the time-domain residual signal of a parallel linear autoregressive model. Objective metrics and a listening test show that PARCnet provides state-of-the-art results while enabling real-time operation on CPU.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360264/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360264/</guid>
</item>
<item>
<title>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</title>
<description><p><span><big>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</big></span><br></p><p><span><small><i>Katerina Zmolikova; Michael Syskind Pedersen; Jesper Jensen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Supervised learning-based speech enhancement methods often work remarkably well in acoustic situations represented in the training corpus but generalize poorly to out-of-domain situations, i.e. situations not seen during training. This stands in the way of further improvement of these methods in realistic scenarios, as collecting paired noisy-clean recordings in the target application domain is typically not feasible. Recording noisy-only in-domain data is, though, much more practical. In this article, we tackle the problem of unsupervised domain adaptation in speech enhancement. Specifically, we propose a way to use in-domain noisy-only data in the training of a neural network to improve upon a model trained solely on out-of-domain paired data. For this, we make use of masked spectrogram prediction, a technique from self-supervised learning that aims to interpolate masked regions of a spectrogram. We hypothesize that masked spectrogram prediction encourages learning of features that represent well both speech and noise components of the noisy signals. These features can then be used to train a more robust speech enhancement system. We evaluate the proposed method on the VoiceBank-DEMAND and LibriFSD50k databases, with WSJ0-CHiME3 serving as the out-of-domain database. We show that the proposed method outperforms both the out-of-domain system and the baseline approaches, i.e. RemixIT and noisy-target training, and also combines well with the previously proposed RemixIT method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360251/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360251/</guid>
</item>
<item>
<title>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</title>
<description><p><span><big>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</big></span><br></p><p><span><small><i>Alexander Karataev; Christian Forsch; Laura Cottatellucci</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3348343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We consider a cell-free massive multiple-input multiple-output (CF-MaMIMO) communication system in the uplink transmission and propose a novel algorithm for blind or semi-blind joint channel estimation and data detection (JCD). We formulate the problem in the framework of bilinear inference and develop a solution based on the expectation propagation (EP) method for both channel estimation and data detection. We propose a new approximation of the joint a posteriori distribution of the channel and data whose representation as a factor graph enables the application of the EP approach using the message-passing technique, local low-complexity computations at the nodes, and an effective modeling of channel-data interplay. The derived algorithm, called bilinear-EP JCD, allows for a distributed implementation among access points (APs) and the central processing unit (CPU) and has polynomial complexity. Our simulation results show that it outperforms other EP-based state-of-the-art polynomial time algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10378663/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10378663/</guid>
</item>
<item>
<title>Adversarial Representation Learning for Robust Privacy Preservation in Audio</title>
<description><p><span><big>Adversarial Representation Learning for Robust Privacy Preservation in Audio</big></span><br></p><p><span><small><i>Shayan Gharib; Minh Tran; Diep Luong; Konstantinos Drossos; Tuomas Virtanen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Sound event detection systems are widely used in various applications such as surveillance and environmental monitoring where data is automatically collected, processed, and sent to a cloud for sound recognition. However, this process may inadvertently reveal sensitive information about users or their surroundings, hence raising privacy concerns. In this study, we propose a novel adversarial training method for learning representations of audio recordings that effectively prevents the detection of speech activity from the latent features of the recordings. The proposed method trains a model to generate invariant latent representations of speech-containing audio recordings that cannot be distinguished from non-speech recordings by a speech classifier. The novelty of our work is in the optimization algorithm, where the speech classifier's weights are regularly replaced with the weights of classifiers trained in a supervised manner. This increases the discrimination power of the speech classifier constantly during the adversarial training, motivating the model to generate latent representations in which speech is not distinguishable, even using new speech classifiers trained outside the adversarial training loop. The proposed method is evaluated against a baseline approach with no privacy measures and a prior adversarial training method, demonstrating a significant reduction in privacy violations compared to the baseline approach. Additionally, we show that the prior adversarial method is practically ineffective for this purpose.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10379095/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10379095/</guid>
</item>
<item>
<title>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</title>
<description><p><span><big>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</big></span><br></p><p><span><small><i>Yasaman Parhizkar; Gene Cheung; Andrew W. Eckford</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>It is a popular hypothesis in neuroscience that ganglion cells in the retina are activated by selectively detecting visual features in an observed scene. While ganglion cell firings can be predicted via data-trained deep neural nets, the n |
|
Successfully generated as following: http://localhost:1200/ieee/journal/8782710 - Failed ❌ |
|
Successfully generated as following: http://localhost:1200/ieee/journal/8782710 - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Open Journal of Signal Processing</title>
<link>https://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=8782710</link>
<atom:link href="http://localhost:1200/ieee/journal/8782710" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Open Journal of Signal Processing - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Tue, 09 Apr 2024 08:42:59 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Synthbuster: Towards Detection of Diffusion Model Generated Images</title>
<description><p><span><big>Synthbuster: Towards Detection of Diffusion Model Generated Images</big></span><br></p><p><span><small><i>Quentin Bammey</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337714</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Synthetically-generated images are getting increasingly popular. Diffusion models have advanced to the stage where even non-experts can generate photo-realistic images from a simple text prompt. They expand creative horizons but also open a Pandora's box of potential disinformation risks. In this context, the present corpus of synthetic image detection techniques, primarily focusing on older generative models like Generative Adversarial Networks, finds itself ill-equipped to deal with this emerging trend. Recognizing this challenge, we introduce a method specifically designed to detect synthetic images produced by diffusion models. Our approach capitalizes on the inherent frequency artefacts left behind during the diffusion process. Spectral analysis is used to highlight the artefacts in the Fourier transform of a residual image, which are used to distinguish real from fake images. The proposed method can detect diffusion-model-generated images even under mild
jpeg
compression, and generalizes relatively well to unknown models. By pioneering this novel approach, we aim to fortify forensic methodologies and ignite further research into the detection of AI-generated images.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334046/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334046/</guid>
</item>
<item>
<title>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</title>
<description><p><span><big>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</big></span><br></p><p><span><small><i>Yanbin Zou; Jingna Fan; Zekai Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we consider using time-of-arrival (TOA) measurements from a single moving receiver to locate a moving target at constant velocity that emits a periodic signal with unknown signal period. First, we give the TOA measurement model and deduce the Cram
$\acute{\text{e}}$
r-Rao lower bounds (CRLB). Then, we formulate a nonlinear least squares (NLS) problem to estimate the unknown parameters. We use semidefinite programming (SDP) techniques to relax the nonconvex NLS problem. However, it is shown that the SDP localization algorithm cannot provide a high-quality solution. Subsequently, we develop a fixed point iteration (FPI) method to improve the performance of the SDP algorithm. In addition, we also consider the presence of receiver position errors and develop the corresponding localization algorithm. Numerical simulations are conducted to demonstrate the localization performance of the proposed algorithms by comparing them with the CRLB.
Index Term
-Fixed point iteration (FPI), semidefinite programming (SDP), single moving receiver, target localization, time-of-arrival (TOA).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336384/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336384/</guid>
</item>
<item>
<title>The Neural-SRP Method for Universal Robust Multi-Source Tracking</title>
<description><p><span><big>The Neural-SRP Method for Universal Robust Multi-Source Tracking</big></span><br></p><p><span><small><i>Eric Grinstein; Christopher M. Hicks; Toon van Waterschoot; Mike Brookes; Patrick A. Naylor</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340057</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Neural networks have achieved state-of-the-art performance on the task of acoustic Direction-of-Arrival (DOA) estimation using microphone arrays. Neural models can be classified as end-to-end or hybrid, each class showing advantages and disadvantages. This work introduces Neural-SRP, an end-to-end neural network architecture for DOA estimation inspired by the classical Steered Response Power (SRP) method, which overcomes limitations of current neural models. We evaluate the architecture on multiple scenarios, namely, multi-source DOA tracking and single-source DOA tracking under the presence of directional and diffuse noise. The experiments demonstrate that our proposed method compares favourably in terms of computational and localization performance with established neural methods on various recorded and simulated benchmark datasets.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345765/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345765/</guid>
</item>
<item>
<title>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</title>
<description><p><span><big>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</big></span><br></p><p><span><small><i>Alexander Venus; Erik Leitinger; Stefan Tertinek; Klaus Witrisal</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper presents a neural-enhanced probabilistic model and corresponding factor graph-based sum-product algorithm for robust localization and tracking in multipath-prone environments. The introduced hybrid probabilistic model consists of physics-based and data-driven measurement models capturing the information contained in both, the line-of-sight (LOS) component as well as in multipath components (NLOS components). The physics-based and data-driven models are embedded in a joint Bayesian framework allowing to derive from first principles a factor graph-based algorithm that fuses the information of these models. The proposed algorithm uses radio signal measurements from multiple base stations to robustly estimate the mobile agent's position together with all model parameters. It provides high localization accuracy by exploiting the position-related information of the LOS component via the physics-based model and robustness by exploiting the geometric imprint of multipath components independent of the propagation channel via the data-driven model. In a challenging numerical experiment involving obstructed LOS situations to all anchors, we show that the proposed sequential algorithm significantly outperforms state-of-the-art methods and attains the posterior Cramér-Rao lower bound even with training data limited to local regions.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336409/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336409/</guid>
</item>
<item>
<title>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</title>
<description><p><span><big>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</big></span><br></p><p><span><small><i>AMIT EFRAIM; Joseph M. Francos</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In solving a model fitting problem, the existence of outliers in the set of measurements can have a devastating effect on the solution accuracy. Traditionally, in order to overcome this problem, robust point cloud registration algorithms are composed of transformation hypothesis generation, followed by hypothesis evaluation aimed at selecting the best hypothesized estimate. Hypotheses evaluation is commonly performed using the sample consensus criterion. However, since this criterion accounts only for the consensus size, it fails when the maximal sample consensus is incorrect. We propose a new hypothesis evaluation approach, generalizing the sample consensus approach, where instead of the sample consensus, the transformation that maximizes the point clouds feature correlation is selected as the best hypothesis. The feature vector at each point contains information such as on local geometry and semantic context. Utilizing this information in the hypotheses evaluation and selection procedure allows for a correct decision even when the hypothesis yielding the maximal sample consensus is false. Consequently, the probability of selecting the correct model increases. We show both mathematically and empirically that substituting the sample consensus criterion with the proposed point cloud feature correlation hypothesis test (PC-FCHT) lowers the probability of large registration errors, compared to using the special case of sample consensus. The proposed PC-FCHT is applicable to any algorithm that follows the hypothesis generation and evaluation scheme, potentially improving the success rates of a wide family of point cloud registration algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345750/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345750/</guid>
</item>
<item>
<title>Joint PAPR and OBP Reduction for NC-OFDM Systems</title>
<description><p><span><big>Joint PAPR and OBP Reduction for NC-OFDM Systems</big></span><br></p><p><span><small><i>Hsuan-Fu Wang; Fang-Biau Ueng; Bo-Heng Yeh</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3329757</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The spectrum resource is always a critical issue for wireless communications since it directly impacts the data rate and capacity. However, the problem of spectrum resource scarcity always exists. Moreover, spectrum resource scarcity becomes more severe as new communication technologies and wireless applications sprout. Noncontiguous orthogonal frequency division multiplexing (NC-OFDM) is a multicarrier method for bandwidth utilization. Unfortunately, this system has two fatal defects: high peak-to-average power ratio (PAPR) and considerable out-of-band power (OBP), which are detrimental to the system's performance. To solve these two problems, we propose a convex optimization-based method for joint PAPR and OBP reduction in NC-OFDM Systems. The strategy is to permit the secondary user to utilize the unoccupied spectrum of the primary user with dynamic spectrum sharing (DSS) based on a cognitive radio network (CRN). To this end, a flexible system operating over noncontiguous bands and DSS scenarios is necessary. The simulation results have shown that our method could effectively improve the overall performance and outperform other schemes, i.e., projections onto convex sets (POCS) and alternating projections onto convex and non-convex sets (APOCNCS), without harming the transmission of the primary system. The collaboration between secondary and primary systems is viable with the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10305259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10305259/</guid>
</item>
<item>
<title>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</title>
<description><p><span><big>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</big></span><br></p><p><span><small><i>Gal Itzhak; Israel Cohen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339433</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article presents a Kronecker-product (KP) beamforming approach incorporating sparse concentric circular arrays (SCCAs). The locations of the microphones on the SCCA are optimized concerning the broadband array directivity over a wide range of direction-of-arrival (DOA) deviations of a desired signal. A maximum directivity factor (MDF) sub-beamformer is derived accordingly with the optimal locations. Then, we propose two global beamformers obtained as a Kronecker product of a uniform linear array (ULA) and the SCCA sub-beamformer. The global beamformers differ by the type of the ULA, which is designed either as an MDF sub-beamformer along the
$\mathsf {x}$
-axis or as a maximum white noise gain sub-beamformer along the
$\mathsf {y}$
-axis. We analyze the performance of the proposed beamformers in terms of the directivity factor, the white noise gain, and their spatial beampatterns. Compared to traditional beamformers, the proposed beamformers exhibit considerably larger tolerance to DOA deviations concerning both the azimuth and elevation angles. Experimental results with speech signals in noisy and reverberant environments demonstrate that the proposed approach outperforms traditional beamformers regarding the perceptual evaluation of speech quality (PESQ) and short-time objective intelligibility (STOI) scores when the desired speech signals deviate from the nominal DOA.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342869/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342869/</guid>
</item>
<item>
<title>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</title>
<description><p><span><big>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</big></span><br></p><p><span><small><i>Karn N. Watcharasupat; Chih-Wei Wu; Yiwei Ding; Iroro Orife; Aaron J. Hipple; Phillip A. Williams; Scott Kramer; Alexander Lerch; William Wolcott</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339428</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Cinematic audio source separation is a relatively new subtask of audio source separation, with the aim of extracting the dialogue, music, and effects stems from their mixture. In this work, we developed a model generalizing the Bandsplit RNN for any complete or overcomplete partitions of the frequency axis. Psychoacoustically motivated frequency scales were used to inform the band definitions which are now defined with redundancy for more reliable feature extraction. A loss function motivated by the signal-to-noise ratio and the sparsity-promoting property of the 1-norm was proposed. We additionally exploit the information-sharing property of a common-encoder setup to reduce computational complexity during both training and inference, improve separation performance for hard-to-generalize classes of sounds, and allow flexibility during inference time with detachable decoders. Our best model sets the state of the art on the Divide and Remaster dataset with performance above the ideal ratio mask for the dialogue stem.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342812/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342812/</guid>
</item>
<item>
<title>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</title>
<description><p><span><big>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</big></span><br></p><p><span><small><i>Miguel Ferrer; María de Diego; Alberto Gonzalez</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340106</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The LMS algorithm is widely employed in adaptive systems due to its robustness, simplicity, and reasonable performance. However, it is well known that this algorithm suffers from a slow convergence speed when dealing with colored reference signals. Numerous variants and alternative algorithms have been proposed to address this issue, though all of them entail an increase in computational cost. Among the proposed alternatives, the affine projection algorithm stands out. This algorithm has the peculiarity of starting from
$N$
data vectors of the reference signal. It transforms these vectors into as many data vectors suitably normalized in energy and mutually orthogonal. In this work, we propose a version of the LMS algorithm that, similar to the affine projection algorithm, starts from
$N$
data vectors of the reference signal but corrects them by using only a scalar factor that functions as a convergence step. Our goal is to align the behavior of this algorithm with the behavior of the affine projection algorithm without significantly increasing the computational cost of the LMS.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345730/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345730/</guid>
</item>
<item>
<title>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</title>
<description><p><span><big>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</big></span><br></p><p><span><small><i>Zongyao Li; Ren Togo; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340616</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article proposes a method for transferring knowledge of semantic segmentation from a labeled source domain to an unlabeled target domain without using the source-domain data. Such a problem is called source-data-free domain adaptation, in which a pre-trained source-domain model and the unlabeled target-domain data are used to transfer the label knowledge across the domains. Like most previous methods, our method uses pseudo labels for distilling and transferring the source-domain knowledge. On the basis of the pseudo-label learning, our method improves the domain adaptation performance in two innovative ways: 1) reducing the domain differences by source-data-free style transfer and 2) exploring the style diversity within the target domain by style modification. To this end, we introduce two additional modules: 1) an inter-domain style transfer module which aligns the feature statistics of the source and target domains before producing the pseudo labels thereby improving the pseudo labels' accuracy, and 2) an intra-domain style modification module which modifies the image styles within the target domain for learning intra-domain style-invariant features. Our method with the two modules outperforms previous source-data-free domain adaptation methods in two commonly used benchmarks. Moreover, our method is well compatible with the previous methods for further improvement.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10356748/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10356748/</guid>
</item>
<item>
<title>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</title>
<description><p><span><big>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</big></span><br></p><p><span><small><i>Tobias Kabzinski; Peter Jax</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337721</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Quasi-continuous system identification of time-variant linear acoustic systems can be applied in various audio signal processing applications when numerous acoustic transfer functions must be measured. A prominent application is measuring head-related transfer functions. We treat the underlying multiple-input-multiple-output (MIMO) system identification problem in a state-space model as a joint estimation problem for states, representing impulse responses, and state-space model parameters using the expectation maximization (EM) algorithm. We address limitations of prior work by imposing different model structures, especially for dependencies within a (transformed) state vector. This results in block diagonal matrix structures, for which we derive M-step update rules. Making assumptions about this model structure and choosing a block size for a given application define the computational complexity. In examples, we found that applying this framework yields improvements of up to 10 dB in relative system distance in comparison to a conventional method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334061/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334061/</guid>
</item>
<item>
<title>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</title>
<description><p><span><big>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</big></span><br></p><p><span><small><i>Damir Rakhimov; Martin Haardt</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337729</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this paper, we present an analytical performance assessment of 2-D Tensor ESPRIT in terms of physical parameters. We show that the error in the
$r$
-mode depends only on two components, irrespective of the dimensionality of the problem. We obtain analytical expressions in closed form for the mean squared error (MSE) in each dimension as a function of the signal-to-noise (SNR) ratio, the array steering matrices, the number of antennas, the number of snapshots, the selection matrices, and the signal correlation. The derived expressions allow a better understanding of the difference in performance between the tensor and the matrix versions of the ESPRIT algorithm. The simulation results confirm the coincidence between the presented analytical expression and the curves obtained via Monte Carlo trials. We analyze the behavior of each of the two error components in different scenarios.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334446/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334446/</guid>
</item>
<item>
<title>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</title>
<description><p><span><big>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</big></span><br></p><p><span><small><i>Yaman Kındap; Simon Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343341</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this work we study linear vector stochastic differential equation (SDE) models driven by the generalised hyperbolic (GH) Lévy process for inference in continuous-time non-Gaussian filtering problems. The GH family of stochastic processes offers a flexible framework for modelling of non-Gaussian, heavy-tailed characteristics and includes the normal inverse-Gaussian, variance-gamma and Student-t processes as special cases. We present continuous-time simulation methods for the solution of vector SDE models driven by GH processes and novel inference methodologies using a variant of sequential Markov chain Monte Carlo (MCMC). As an example a particular formulation of Langevin dynamics is studied within this framework. The model is applied to both a synthetically generated data set and a real-world financial series to demonstrate its capabilities.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360268/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360268/</guid>
</item>
<item>
<title>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</title>
<description><p><span><big>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</big></span><br></p><p><span><small><i>James M. Cozens; Simon J. Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344048</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper proposes a probabilistic approach for extracting time-varying and irregular time signature information from polyphonic audio extracts, subsequently providing beat and bar line positions given inferred time signature divisions. This is achieved via dynamically evaluating the beat tempo as a function of time through finding an optimal compromise in beat and bar alignment in the time and tempo domains. Time signature divisions are determined based on a new representation, termed the Metrogram, that presents time-varying information regarding rhythmic and metric periodicities in the Tempogram. Our methodology is characterised by its ability to provide a distribution over metric interpretations, offering insights into the diverse ways music can be rhythmically perceived. Results indicate high-level accuracy for a variety of polyphonic extracts containing irregular, complex, irrational, and time-varying time signatures. Accuracy rivalling state-of-the-art methodologies is also reported in a beat tracking task performed on the standard Ballroom Dataset. The paper offers insights into the field of dynamic time signature recognition and beat tracking, offering a valuable and versatile resource for the analysis, composition, and performance of music.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363392/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363392/</guid>
</item>
<item>
<title>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</title>
<description><p><span><big>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</big></span><br></p><p><span><small><i>Yuto Watanabe; Ren Togo; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343335</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Although text-guided image manipulation approaches have demonstrated highly accurate performance for editing the appearance of images in a virtual or simple scenario, their real-world applications face significant challenges. The primary cause of these challenges is the misalignment in the distribution of training and real-world data, which leads to unstable text-guided image manipulation. In this work, we propose a novel framework called TolerantGAN and tackle the new task of real-world text-guided image manipulation independent of the training data. To achieve this, we introduce two key concepts of a border smoothly connection module (BSCM) and a manipulation direction-based attention module (MDAM). BSCM smoothens the misalignment in the distribution of training and real-world data. MDAM extracts only regions highly relevant for image manipulation and assists in reconstructing unobserved objects in the training data. For in-the-wild input images of various classes, TolerantGAN robustly outperforms the state-of-the-art methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360283/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360283/</guid>
</item>
<item>
<title>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</title>
<description><p><span><big>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</big></span><br></p><p><span><small><i>Anastasia Avdeeva; Aleksei Gusev; Tseren Andzhukaev; Artem Ivanov</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343342</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Whispered speech is a quiet voice without vocalization. One of the common cases of using whispered speech is a technique that can help overcome stuttering. But whispered speech can be uncomfortable and difficult to understand in everyday communication. To address these problems, we propose a method of low-delayed whisper-to-speech voice conversion, which can be useful in real life communication of people with disordered speech. As part of our research, we study the impact of streaming Automatic Speech Recognition models on the quality of voice conversion, comparing different streaming models and methods for model adaptation to streaming settings, and showing the importance of using such models in cases of low-delayed voice conversion.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360259/</guid>
</item>
<item>
<title>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</title>
<description><p><span><big>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</big></span><br></p><p><span><small><i>Denis C. Ilie-Ablachim; Andra Băltoiu; Bogdan Dumitrescu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344313</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We propose sparse representation and dictionary learning algorithms for dictionaries whose atoms are characterized by Gaussian probability distributions around some central atoms. This extension of the space covered by the atoms permits a better characterization of localized features of the signals. The representations are computed by trading-off representation error and the probability of the actual atoms used in the representation. We propose two types of algorithms: a greedy one, similar in spirit with Orthogonal Matching Pursuit, and one based on L1-regularization. We apply our new algorithms to unsupervised anomaly detection, where the representation error is used as anomaly score. The flexibility of our approach appears to improve more the representations of the many normal signals than those of the few outliers, at least for an anomaly type called dependency, thus improving the detection quality. Comparison with both standard dictionary learning algorithms and established anomaly detection methods is favorable.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365333/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365333/</guid>
</item>
<item>
<title>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</title>
<description><p><span><big>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</big></span><br></p><p><span><small><i>Yuya Moroto; Yingrui Ye; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344079</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>There are various sentiment theories for categorizing human sentiments into several discrete sentiment categories, which means that the theory used for training sentiment prediction methods does not always match that used in the test phase. As a solution to this problem, zero-shot visual sentiment prediction methods have been proposed to predict unseen sentiments for which no images are available in the training phase. However, the training of these previous zero-shot methods relies on a single sentiment theory, which limits their ability to handle sentiments from other theories. Thus, this article proposes a more robust zero-shot visual sentiment prediction method that can handle cross-domain sentiments defined in different sentiment theories. Specifically, by focusing on the fact that sentiments are abstract concepts common to humans regardless of whether their theories are different, we incorporate knowledge distillation into our method to construct a teacher–student model that can train the implicit relationships between sentiments defined in different sentiment theories. Furthermore, to enhance sentiment discrimination capability and strengthen the implicit relationships between sentiments, we introduce a novel sentiment loss between the teacher and student models. In this way, our model becomes robust to unseen sentiments by exploiting the implicit relationships between sentiments. The contributions of this article are the introduction of knowledge distillation and a novel sentiment loss between the teacher and student models for zero-shot visual sentiment prediction, and improved performance of zero-shot visual sentiment prediction. Experiments on several open datasets demonstrate the effectiveness of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363382/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363382/</guid>
</item>
<item>
<title>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</title>
<description><p><span><big>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</big></span><br></p><p><span><small><i>Chengen Liu; Geert Leus; Elvin Isufi</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339376</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The edge flow reconstruction task consists of retreiving edge flow signals from corrupted or incomplete measurements. This is typically solved by a regularized optimization problem on higher-order networks such as simplicial complexes and the corresponding regularizers are chosen based on prior knowledge. Tailoring this prior to the setting of interest can be challenging or it may not even be possible. Thus, we consider to learn this prior knowledge via a model-based deep learning approach. We propose a new regularized optimization problem for the simplicial edge flow reconstruction task, the simplicial ElasticNet, which combines the advantages of the
$\ell _{1}$
and
$\ell _{2}$
norms. We solve the simplicial ElasticNet problem via the multi-block alternating direction method of multipliers (ADMM) algorithm and provide conditions on its convergence. By unrolling the ADMM iterative steps, we develop a model-based neural network with a low requirement on the number of training data. This unrolling network replaces the fixed parameters in the iterative algorithm by learnable weights, thus exploiting the neural network's learning capability while preserving the iterative algorithm's interpretability. We enhance this unrolling network via simplicial convolutional filters to aggregate information from the edge flow neighbors, ultimately, improving the network learning expressivity. Extensive experiments on real-world and synthetic datasets validate the proposed approaches and show considerable improvements over both baselines and traditional non-model-based neural networks.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342735/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342735/</guid>
</item>
<item>
<title>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</title>
<description><p><span><big>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</big></span><br></p><p><span><small><i>Jim Beckers; Bart Van Erp; Ziyue Zhao; Kirill Kondrashov; Bert De Vries</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337718</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Bayesian model reduction provides an efficient approach for comparing the performance of all nested sub-models of a model, without re-evaluating any of these sub-models. Until now, Bayesian model reduction has been applied mainly in the computational neuroscience community on simple models. In this paper, we formulate and apply Bayesian model reduction to perform principled pruning of Bayesian neural networks, based on variational free energy minimization. Direct application of Bayesian model reduction, however, gives rise to approximation errors. Therefore, a novel iterative pruning algorithm is presented to alleviate the problems arising with naive Bayesian model reduction, as supported experimentally on the publicly available UCI datasets for different inference algorithms. This novel parameter pruning scheme solves the shortcomings of current state-of-the-art pruning methods that are used by the signal processing community. The proposed approach has a clear stopping criterion and minimizes the same objective that is used during training. Next to these benefits, our experiments indicate better model performance in comparison to state-of-the-art pruning schemes.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334001/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334001/</guid>
</item>
<item>
<title>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</title>
<description><p><span><big>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</big></span><br></p><p><span><small><i>Anh Minh Truong; Wilfried Philips; Peter Veelaert</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340064</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Many image-based rendering (IBR) methods rely on depth estimates obtained from structured light or time-of-flight depth sensors to synthesize novel views from sparse camera networks. However, these estimates often contain missing or noisy regions, resulting in an incorrect mapping between source and target views. This situation makes the fusion process more challenging, as the visual information is misaligned, inconsistent, or missing. In this work, we first implement a lightweight network based on the transformer, which is well-known for its capability to model long-range relationships within input data, to extract spatial features from color images. These features are then used to enhance the quality of completed depth maps. Furthermore, we combine a sequential deep neural network with a spatial attention mechanism to effectively fuse the projected features from multiple source viewpoints. This approach enables us to integrate information from an arbitrary number of source viewpoints as well as improve accuracy in synthesized views. Experimental results on challenging datasets demonstrate that our method achieves superior synthesized image quality compared to state-of-the-art (SOTA) methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345792/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345792/</guid>
</item>
<item>
<title>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</title>
<description><p><span><big>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</big></span><br></p><p><span><small><i>Reza Mirzaeifard; Naveen K. D. Venkategowda; Vinay Chakravarthi Gogineni; Stefan Werner</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344395</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper investigates quantile regression in the presence of non-convex and non-smooth sparse penalties, such as the minimax concave penalty (MCP) and smoothly clipped absolute deviation (SCAD). The non-smooth and non-convex nature of these problems often leads to convergence difficulties for many algorithms. While iterative techniques such as coordinate descent and local linear approximation can facilitate convergence, the process is often slow. This sluggish pace is primarily due to the need to run these approximation techniques until full convergence at each step, a requirement we term as a
secondary convergence iteration
. To accelerate the convergence speed, we employ the alternating direction method of multipliers (ADMM) and introduce a novel single-loop smoothing ADMM algorithm with an increasing penalty parameter, named SIAD, specifically tailored for sparse-penalized quantile regression. We first delve into the convergence properties of the proposed SIAD algorithm and establish the necessary conditions for convergence. Theoretically, we confirm a convergence rate of
$o({k^{-\frac{1}{4}}})$
for the sub-gradient bound of the augmented Lagrangian, where
$k$
denotes the number of iterations. Subsequently, we provide numerical results to showcase the effectiveness of the SIAD algorithm. Our findings highlight that the SIAD method outperforms existing approaches, providing a faster and more stable solution for sparse-penalized quantile regression.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365338/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365338/</guid>
</item>
<item>
<title>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</title>
<description><p><span><big>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</big></span><br></p><p><span><small><i>Oliver Lang; Christian Hofbauer; Reinhard Feger; Mario Huemer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343308</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>A promising waveform candidate for future joint sensing and communication systems is orthogonal frequency-division multiplexing (OFDM). For such systems, supporting multiple transmit antennas requires multiplexing methods for the generation of orthogonal transmit signals, where equidistant subcarrier interleaving (ESI) is the most popular multiplexing method. In this work, we analyze a multiplexing method called Doppler-division multiplexing (DDM). This method applies a phase shift from OFDM symbol to OFDM symbol to separate signals transmitted by different Tx antennas along the velocity axis of the range-Doppler map. The main focus of this work lies on the implications of DDM on the communication task. It will be shown that for DDM, the channels observed in the communication receiver are heavily time-varying, preventing any meaningful transmission of data when not taken into account. In this work, a communication system designed to combat these time-varying channels is proposed, which includes methods for data estimation, synchronization, and channel estimation. Bit error ratio (BER) simulations demonstrate the superiority of this communications system compared to ESI-based systems.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360223/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360223/</guid>
</item>
<item>
<title>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</title>
<description><p><span><big>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</big></span><br></p><p><span><small><i>Nils L. Westhausen; Bernd T. Meyer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343320</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we introduce a causal low-latency low-complexity approach for binaural multichannel blind speaker separation in noisy reverberant conditions. The model, referred to as Group Communication Binaural Filter and Sum Network (GCBFSnet) predicts complex filters for filter-and-sum beamforming in the time-frequency domain. We apply Group Communication (GC), i.e., latent model variables are split into groups and processed with a shared sequence model with the aim of reducing the complexity of a simple model only containing one convolutional and one recurrent module. With GC we are able to reduce the size of the model by up to 83% and the complexity up to 73% compared to the model without GC, while mostly retaining performance. Even for the smallest model configuration, GCBFSnet matches the performance of a low-complexity TasNet baseline in most metrics despite the larger size and higher number of required operations of the baseline.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360275/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360275/</guid>
</item>
<item>
<title>Reverse Ordering Techniques for Attention-Based Channel Prediction</title>
<description><p><span><big>Reverse Ordering Techniques for Attention-Based Channel Prediction</big></span><br></p><p><span><small><i>Valentina Rizzello; Benedikt Böck; Michael Joham; Wolfgang Utschick</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344024</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Channel state information (CSI) is crucial for enhancing the performance of wireless systems by allowing to adjust the transmission strategies based on the current channel conditions. However, obtaining precise CSI is difficult because of the fast-changing channel conditions caused by multi-path fading. An inaccurate CSI hinders the performance of various adaptive wireless systems, highlighting the need for channel prediction techniques to effectively mitigate the drawbacks of outdated CSI. Conventional methods typically depend on assumptions regarding user velocity or require knowledge of the Doppler frequency. In contrast to existing approaches, we aim for a more robust and practical solution by training neural networks without making any assumptions about user velocity, relying solely on noisy channel observations during training. Specifically, we adapt both the sequence-to-sequence with attention (Seq2Seq-attn) and transformer models for channel prediction. Additionally, a new technique called reverse positional encoding is introduced in the transformer model to improve the robustness of the model against varying sequence lengths. Similarly, the encoder outputs of the Seq2Seq-attn model are reversed prior to the application of attention mechanisms. By means of simulations, we show that these proposed techniques enable the models to effectively capture relationships within sequences of channel snapshots without increasing the complexity. Importantly, this capability remains robust across varying sequence lengths, representing a substantial improvement over existing methodologies.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363354/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363354/</guid>
</item>
<item>
<title>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</title>
<description><p><span><big>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</big></span><br></p><p><span><small><i>Sarina Meyer; Xiaoxiao Miao; Ngoc Thang Vu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344375</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Speaker anonymization is the task of modifying a speech recording such that the original speaker cannot be identified anymore. Since the first Voice Privacy Challenge in 2020, along with the release of a framework, the popularity of this research topic is continually increasing. However, the comparison and combination of different anonymization approaches remains challenging due to the complexity of evaluation and the absence of user-friendly research frameworks. We therefore propose an efficient speaker anonymization and evaluation framework based on a modular and easily extendable structure, almost fully in Python. The framework facilitates the orchestration of several anonymization approaches in parallel and allows for interfacing between different techniques. Furthermore, we propose modifications to common evaluation methods which improves the quality of the evaluation and reduces their computation time by 65 to 95%, depending on the metric. Our code is fully open source.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365329/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365329/</guid>
</item>
<item>
<title>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</title>
<description><p><span><big>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</big></span><br></p><p><span><small><i>Alessandro Ilic Mezza; Matteo Amerena; Alberto Bernardini; Augusto Sarti</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343318</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Real-time audio communications over IP have become essential to our daily lives. Packet-switched networks, however, are inherently prone to jitter and data losses, thus creating a strong need for effective packet loss concealment (PLC) techniques. Though solutions based on deep learning have made significant progress in that direction as far as speech is concerned, extending the use of such methods to applications of Networked Music Performance (NMP) presents significant challenges, including high fidelity requirements, higher sampling rates, and stringent temporal constraints associated to the simultaneous interaction between remote musicians. In this article, we present PARCnet, a hybrid PLC method that utilizes a feed-forward neural network to estimate the time-domain residual signal of a parallel linear autoregressive model. Objective metrics and a listening test show that PARCnet provides state-of-the-art results while enabling real-time operation on CPU.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360264/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360264/</guid>
</item>
<item>
<title>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</title>
<description><p><span><big>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</big></span><br></p><p><span><small><i>Katerina Zmolikova; Michael Syskind Pedersen; Jesper Jensen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Supervised learning-based speech enhancement methods often work remarkably well in acoustic situations represented in the training corpus but generalize poorly to out-of-domain situations, i.e. situations not seen during training. This stands in the way of further improvement of these methods in realistic scenarios, as collecting paired noisy-clean recordings in the target application domain is typically not feasible. Recording noisy-only in-domain data is, though, much more practical. In this article, we tackle the problem of unsupervised domain adaptation in speech enhancement. Specifically, we propose a way to use in-domain noisy-only data in the training of a neural network to improve upon a model trained solely on out-of-domain paired data. For this, we make use of masked spectrogram prediction, a technique from self-supervised learning that aims to interpolate masked regions of a spectrogram. We hypothesize that masked spectrogram prediction encourages learning of features that represent well both speech and noise components of the noisy signals. These features can then be used to train a more robust speech enhancement system. We evaluate the proposed method on the VoiceBank-DEMAND and LibriFSD50k databases, with WSJ0-CHiME3 serving as the out-of-domain database. We show that the proposed method outperforms both the out-of-domain system and the baseline approaches, i.e. RemixIT and noisy-target training, and also combines well with the previously proposed RemixIT method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360251/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360251/</guid>
</item>
<item>
<title>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</title>
<description><p><span><big>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</big></span><br></p><p><span><small><i>Alexander Karataev; Christian Forsch; Laura Cottatellucci</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3348343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We consider a cell-free massive multiple-input multiple-output (CF-MaMIMO) communication system in the uplink transmission and propose a novel algorithm for blind or semi-blind joint channel estimation and data detection (JCD). We formulate the problem in the framework of bilinear inference and develop a solution based on the expectation propagation (EP) method for both channel estimation and data detection. We propose a new approximation of the joint a posteriori distribution of the channel and data whose representation as a factor graph enables the application of the EP approach using the message-passing technique, local low-complexity computations at the nodes, and an effective modeling of channel-data interplay. The derived algorithm, called bilinear-EP JCD, allows for a distributed implementation among access points (APs) and the central processing unit (CPU) and has polynomial complexity. Our simulation results show that it outperforms other EP-based state-of-the-art polynomial time algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10378663/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10378663/</guid>
</item>
<item>
<title>Adversarial Representation Learning for Robust Privacy Preservation in Audio</title>
<description><p><span><big>Adversarial Representation Learning for Robust Privacy Preservation in Audio</big></span><br></p><p><span><small><i>Shayan Gharib; Minh Tran; Diep Luong; Konstantinos Drossos; Tuomas Virtanen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Sound event detection systems are widely used in various applications such as surveillance and environmental monitoring where data is automatically collected, processed, and sent to a cloud for sound recognition. However, this process may inadvertently reveal sensitive information about users or their surroundings, hence raising privacy concerns. In this study, we propose a novel adversarial training method for learning representations of audio recordings that effectively prevents the detection of speech activity from the latent features of the recordings. The proposed method trains a model to generate invariant latent representations of speech-containing audio recordings that cannot be distinguished from non-speech recordings by a speech classifier. The novelty of our work is in the optimization algorithm, where the speech classifier's weights are regularly replaced with the weights of classifiers trained in a supervised manner. This increases the discrimination power of the speech classifier constantly during the adversarial training, motivating the model to generate latent representations in which speech is not distinguishable, even using new speech classifiers trained outside the adversarial training loop. The proposed method is evaluated against a baseline approach with no privacy measures and a prior adversarial training method, demonstrating a significant reduction in privacy violations compared to the baseline approach. Additionally, we show that the prior adversarial method is practically ineffective for this purpose.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10379095/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10379095/</guid>
</item>
<item>
<title>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</title>
<description><p><span><big>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</big></span><br></p><p><span><small><i>Yasaman Parhizkar; Gene Cheung; Andrew W. Eckford</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>It is a popular hypothesis in neuroscience that ganglion cells in the retina are activated by selectively detecting visual features in an observed scene. While ganglion cell firings can be predicted via data-trained deep neural nets, the n |
|
Successfully generated as following: http://localhost:1200/ieee/journal/8782710 - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Open Journal of Signal Processing</title>
<link>https://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=8782710</link>
<atom:link href="http://localhost:1200/ieee/journal/8782710" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Open Journal of Signal Processing - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Tue, 09 Apr 2024 13:47:40 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Synthbuster: Towards Detection of Diffusion Model Generated Images</title>
<description><p><span><big>Synthbuster: Towards Detection of Diffusion Model Generated Images</big></span><br></p><p><span><small><i>Quentin Bammey</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337714</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Synthetically-generated images are getting increasingly popular. Diffusion models have advanced to the stage where even non-experts can generate photo-realistic images from a simple text prompt. They expand creative horizons but also open a Pandora's box of potential disinformation risks. In this context, the present corpus of synthetic image detection techniques, primarily focusing on older generative models like Generative Adversarial Networks, finds itself ill-equipped to deal with this emerging trend. Recognizing this challenge, we introduce a method specifically designed to detect synthetic images produced by diffusion models. Our approach capitalizes on the inherent frequency artefacts left behind during the diffusion process. Spectral analysis is used to highlight the artefacts in the Fourier transform of a residual image, which are used to distinguish real from fake images. The proposed method can detect diffusion-model-generated images even under mild
jpeg
compression, and generalizes relatively well to unknown models. By pioneering this novel approach, we aim to fortify forensic methodologies and ignite further research into the detection of AI-generated images.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334046/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334046/</guid>
</item>
<item>
<title>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</title>
<description><p><span><big>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</big></span><br></p><p><span><small><i>Yanbin Zou; Jingna Fan; Zekai Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we consider using time-of-arrival (TOA) measurements from a single moving receiver to locate a moving target at constant velocity that emits a periodic signal with unknown signal period. First, we give the TOA measurement model and deduce the Cram
$\acute{\text{e}}$
r-Rao lower bounds (CRLB). Then, we formulate a nonlinear least squares (NLS) problem to estimate the unknown parameters. We use semidefinite programming (SDP) techniques to relax the nonconvex NLS problem. However, it is shown that the SDP localization algorithm cannot provide a high-quality solution. Subsequently, we develop a fixed point iteration (FPI) method to improve the performance of the SDP algorithm. In addition, we also consider the presence of receiver position errors and develop the corresponding localization algorithm. Numerical simulations are conducted to demonstrate the localization performance of the proposed algorithms by comparing them with the CRLB.
Index Term
-Fixed point iteration (FPI), semidefinite programming (SDP), single moving receiver, target localization, time-of-arrival (TOA).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336384/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336384/</guid>
</item>
<item>
<title>The Neural-SRP Method for Universal Robust Multi-Source Tracking</title>
<description><p><span><big>The Neural-SRP Method for Universal Robust Multi-Source Tracking</big></span><br></p><p><span><small><i>Eric Grinstein; Christopher M. Hicks; Toon van Waterschoot; Mike Brookes; Patrick A. Naylor</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340057</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Neural networks have achieved state-of-the-art performance on the task of acoustic Direction-of-Arrival (DOA) estimation using microphone arrays. Neural models can be classified as end-to-end or hybrid, each class showing advantages and disadvantages. This work introduces Neural-SRP, an end-to-end neural network architecture for DOA estimation inspired by the classical Steered Response Power (SRP) method, which overcomes limitations of current neural models. We evaluate the architecture on multiple scenarios, namely, multi-source DOA tracking and single-source DOA tracking under the presence of directional and diffuse noise. The experiments demonstrate that our proposed method compares favourably in terms of computational and localization performance with established neural methods on various recorded and simulated benchmark datasets.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345765/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345765/</guid>
</item>
<item>
<title>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</title>
<description><p><span><big>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</big></span><br></p><p><span><small><i>Alexander Venus; Erik Leitinger; Stefan Tertinek; Klaus Witrisal</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper presents a neural-enhanced probabilistic model and corresponding factor graph-based sum-product algorithm for robust localization and tracking in multipath-prone environments. The introduced hybrid probabilistic model consists of physics-based and data-driven measurement models capturing the information contained in both, the line-of-sight (LOS) component as well as in multipath components (NLOS components). The physics-based and data-driven models are embedded in a joint Bayesian framework allowing to derive from first principles a factor graph-based algorithm that fuses the information of these models. The proposed algorithm uses radio signal measurements from multiple base stations to robustly estimate the mobile agent's position together with all model parameters. It provides high localization accuracy by exploiting the position-related information of the LOS component via the physics-based model and robustness by exploiting the geometric imprint of multipath components independent of the propagation channel via the data-driven model. In a challenging numerical experiment involving obstructed LOS situations to all anchors, we show that the proposed sequential algorithm significantly outperforms state-of-the-art methods and attains the posterior Cramér-Rao lower bound even with training data limited to local regions.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336409/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336409/</guid>
</item>
<item>
<title>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</title>
<description><p><span><big>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</big></span><br></p><p><span><small><i>AMIT EFRAIM; Joseph M. Francos</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In solving a model fitting problem, the existence of outliers in the set of measurements can have a devastating effect on the solution accuracy. Traditionally, in order to overcome this problem, robust point cloud registration algorithms are composed of transformation hypothesis generation, followed by hypothesis evaluation aimed at selecting the best hypothesized estimate. Hypotheses evaluation is commonly performed using the sample consensus criterion. However, since this criterion accounts only for the consensus size, it fails when the maximal sample consensus is incorrect. We propose a new hypothesis evaluation approach, generalizing the sample consensus approach, where instead of the sample consensus, the transformation that maximizes the point clouds feature correlation is selected as the best hypothesis. The feature vector at each point contains information such as on local geometry and semantic context. Utilizing this information in the hypotheses evaluation and selection procedure allows for a correct decision even when the hypothesis yielding the maximal sample consensus is false. Consequently, the probability of selecting the correct model increases. We show both mathematically and empirically that substituting the sample consensus criterion with the proposed point cloud feature correlation hypothesis test (PC-FCHT) lowers the probability of large registration errors, compared to using the special case of sample consensus. The proposed PC-FCHT is applicable to any algorithm that follows the hypothesis generation and evaluation scheme, potentially improving the success rates of a wide family of point cloud registration algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345750/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345750/</guid>
</item>
<item>
<title>Joint PAPR and OBP Reduction for NC-OFDM Systems</title>
<description><p><span><big>Joint PAPR and OBP Reduction for NC-OFDM Systems</big></span><br></p><p><span><small><i>Hsuan-Fu Wang; Fang-Biau Ueng; Bo-Heng Yeh</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3329757</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The spectrum resource is always a critical issue for wireless communications since it directly impacts the data rate and capacity. However, the problem of spectrum resource scarcity always exists. Moreover, spectrum resource scarcity becomes more severe as new communication technologies and wireless applications sprout. Noncontiguous orthogonal frequency division multiplexing (NC-OFDM) is a multicarrier method for bandwidth utilization. Unfortunately, this system has two fatal defects: high peak-to-average power ratio (PAPR) and considerable out-of-band power (OBP), which are detrimental to the system's performance. To solve these two problems, we propose a convex optimization-based method for joint PAPR and OBP reduction in NC-OFDM Systems. The strategy is to permit the secondary user to utilize the unoccupied spectrum of the primary user with dynamic spectrum sharing (DSS) based on a cognitive radio network (CRN). To this end, a flexible system operating over noncontiguous bands and DSS scenarios is necessary. The simulation results have shown that our method could effectively improve the overall performance and outperform other schemes, i.e., projections onto convex sets (POCS) and alternating projections onto convex and non-convex sets (APOCNCS), without harming the transmission of the primary system. The collaboration between secondary and primary systems is viable with the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10305259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10305259/</guid>
</item>
<item>
<title>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</title>
<description><p><span><big>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</big></span><br></p><p><span><small><i>Gal Itzhak; Israel Cohen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339433</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article presents a Kronecker-product (KP) beamforming approach incorporating sparse concentric circular arrays (SCCAs). The locations of the microphones on the SCCA are optimized concerning the broadband array directivity over a wide range of direction-of-arrival (DOA) deviations of a desired signal. A maximum directivity factor (MDF) sub-beamformer is derived accordingly with the optimal locations. Then, we propose two global beamformers obtained as a Kronecker product of a uniform linear array (ULA) and the SCCA sub-beamformer. The global beamformers differ by the type of the ULA, which is designed either as an MDF sub-beamformer along the
$\mathsf {x}$
-axis or as a maximum white noise gain sub-beamformer along the
$\mathsf {y}$
-axis. We analyze the performance of the proposed beamformers in terms of the directivity factor, the white noise gain, and their spatial beampatterns. Compared to traditional beamformers, the proposed beamformers exhibit considerably larger tolerance to DOA deviations concerning both the azimuth and elevation angles. Experimental results with speech signals in noisy and reverberant environments demonstrate that the proposed approach outperforms traditional beamformers regarding the perceptual evaluation of speech quality (PESQ) and short-time objective intelligibility (STOI) scores when the desired speech signals deviate from the nominal DOA.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342869/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342869/</guid>
</item>
<item>
<title>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</title>
<description><p><span><big>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</big></span><br></p><p><span><small><i>Karn N. Watcharasupat; Chih-Wei Wu; Yiwei Ding; Iroro Orife; Aaron J. Hipple; Phillip A. Williams; Scott Kramer; Alexander Lerch; William Wolcott</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339428</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Cinematic audio source separation is a relatively new subtask of audio source separation, with the aim of extracting the dialogue, music, and effects stems from their mixture. In this work, we developed a model generalizing the Bandsplit RNN for any complete or overcomplete partitions of the frequency axis. Psychoacoustically motivated frequency scales were used to inform the band definitions which are now defined with redundancy for more reliable feature extraction. A loss function motivated by the signal-to-noise ratio and the sparsity-promoting property of the 1-norm was proposed. We additionally exploit the information-sharing property of a common-encoder setup to reduce computational complexity during both training and inference, improve separation performance for hard-to-generalize classes of sounds, and allow flexibility during inference time with detachable decoders. Our best model sets the state of the art on the Divide and Remaster dataset with performance above the ideal ratio mask for the dialogue stem.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342812/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342812/</guid>
</item>
<item>
<title>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</title>
<description><p><span><big>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</big></span><br></p><p><span><small><i>Miguel Ferrer; María de Diego; Alberto Gonzalez</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340106</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The LMS algorithm is widely employed in adaptive systems due to its robustness, simplicity, and reasonable performance. However, it is well known that this algorithm suffers from a slow convergence speed when dealing with colored reference signals. Numerous variants and alternative algorithms have been proposed to address this issue, though all of them entail an increase in computational cost. Among the proposed alternatives, the affine projection algorithm stands out. This algorithm has the peculiarity of starting from
$N$
data vectors of the reference signal. It transforms these vectors into as many data vectors suitably normalized in energy and mutually orthogonal. In this work, we propose a version of the LMS algorithm that, similar to the affine projection algorithm, starts from
$N$
data vectors of the reference signal but corrects them by using only a scalar factor that functions as a convergence step. Our goal is to align the behavior of this algorithm with the behavior of the affine projection algorithm without significantly increasing the computational cost of the LMS.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345730/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345730/</guid>
</item>
<item>
<title>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</title>
<description><p><span><big>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</big></span><br></p><p><span><small><i>Zongyao Li; Ren Togo; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340616</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article proposes a method for transferring knowledge of semantic segmentation from a labeled source domain to an unlabeled target domain without using the source-domain data. Such a problem is called source-data-free domain adaptation, in which a pre-trained source-domain model and the unlabeled target-domain data are used to transfer the label knowledge across the domains. Like most previous methods, our method uses pseudo labels for distilling and transferring the source-domain knowledge. On the basis of the pseudo-label learning, our method improves the domain adaptation performance in two innovative ways: 1) reducing the domain differences by source-data-free style transfer and 2) exploring the style diversity within the target domain by style modification. To this end, we introduce two additional modules: 1) an inter-domain style transfer module which aligns the feature statistics of the source and target domains before producing the pseudo labels thereby improving the pseudo labels' accuracy, and 2) an intra-domain style modification module which modifies the image styles within the target domain for learning intra-domain style-invariant features. Our method with the two modules outperforms previous source-data-free domain adaptation methods in two commonly used benchmarks. Moreover, our method is well compatible with the previous methods for further improvement.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10356748/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10356748/</guid>
</item>
<item>
<title>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</title>
<description><p><span><big>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</big></span><br></p><p><span><small><i>Tobias Kabzinski; Peter Jax</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337721</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Quasi-continuous system identification of time-variant linear acoustic systems can be applied in various audio signal processing applications when numerous acoustic transfer functions must be measured. A prominent application is measuring head-related transfer functions. We treat the underlying multiple-input-multiple-output (MIMO) system identification problem in a state-space model as a joint estimation problem for states, representing impulse responses, and state-space model parameters using the expectation maximization (EM) algorithm. We address limitations of prior work by imposing different model structures, especially for dependencies within a (transformed) state vector. This results in block diagonal matrix structures, for which we derive M-step update rules. Making assumptions about this model structure and choosing a block size for a given application define the computational complexity. In examples, we found that applying this framework yields improvements of up to 10 dB in relative system distance in comparison to a conventional method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334061/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334061/</guid>
</item>
<item>
<title>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</title>
<description><p><span><big>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</big></span><br></p><p><span><small><i>Damir Rakhimov; Martin Haardt</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337729</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this paper, we present an analytical performance assessment of 2-D Tensor ESPRIT in terms of physical parameters. We show that the error in the
$r$
-mode depends only on two components, irrespective of the dimensionality of the problem. We obtain analytical expressions in closed form for the mean squared error (MSE) in each dimension as a function of the signal-to-noise (SNR) ratio, the array steering matrices, the number of antennas, the number of snapshots, the selection matrices, and the signal correlation. The derived expressions allow a better understanding of the difference in performance between the tensor and the matrix versions of the ESPRIT algorithm. The simulation results confirm the coincidence between the presented analytical expression and the curves obtained via Monte Carlo trials. We analyze the behavior of each of the two error components in different scenarios.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334446/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334446/</guid>
</item>
<item>
<title>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</title>
<description><p><span><big>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</big></span><br></p><p><span><small><i>Yaman Kındap; Simon Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343341</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this work we study linear vector stochastic differential equation (SDE) models driven by the generalised hyperbolic (GH) Lévy process for inference in continuous-time non-Gaussian filtering problems. The GH family of stochastic processes offers a flexible framework for modelling of non-Gaussian, heavy-tailed characteristics and includes the normal inverse-Gaussian, variance-gamma and Student-t processes as special cases. We present continuous-time simulation methods for the solution of vector SDE models driven by GH processes and novel inference methodologies using a variant of sequential Markov chain Monte Carlo (MCMC). As an example a particular formulation of Langevin dynamics is studied within this framework. The model is applied to both a synthetically generated data set and a real-world financial series to demonstrate its capabilities.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360268/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360268/</guid>
</item>
<item>
<title>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</title>
<description><p><span><big>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</big></span><br></p><p><span><small><i>James M. Cozens; Simon J. Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344048</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper proposes a probabilistic approach for extracting time-varying and irregular time signature information from polyphonic audio extracts, subsequently providing beat and bar line positions given inferred time signature divisions. This is achieved via dynamically evaluating the beat tempo as a function of time through finding an optimal compromise in beat and bar alignment in the time and tempo domains. Time signature divisions are determined based on a new representation, termed the Metrogram, that presents time-varying information regarding rhythmic and metric periodicities in the Tempogram. Our methodology is characterised by its ability to provide a distribution over metric interpretations, offering insights into the diverse ways music can be rhythmically perceived. Results indicate high-level accuracy for a variety of polyphonic extracts containing irregular, complex, irrational, and time-varying time signatures. Accuracy rivalling state-of-the-art methodologies is also reported in a beat tracking task performed on the standard Ballroom Dataset. The paper offers insights into the field of dynamic time signature recognition and beat tracking, offering a valuable and versatile resource for the analysis, composition, and performance of music.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363392/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363392/</guid>
</item>
<item>
<title>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</title>
<description><p><span><big>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</big></span><br></p><p><span><small><i>Yuto Watanabe; Ren Togo; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343335</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Although text-guided image manipulation approaches have demonstrated highly accurate performance for editing the appearance of images in a virtual or simple scenario, their real-world applications face significant challenges. The primary cause of these challenges is the misalignment in the distribution of training and real-world data, which leads to unstable text-guided image manipulation. In this work, we propose a novel framework called TolerantGAN and tackle the new task of real-world text-guided image manipulation independent of the training data. To achieve this, we introduce two key concepts of a border smoothly connection module (BSCM) and a manipulation direction-based attention module (MDAM). BSCM smoothens the misalignment in the distribution of training and real-world data. MDAM extracts only regions highly relevant for image manipulation and assists in reconstructing unobserved objects in the training data. For in-the-wild input images of various classes, TolerantGAN robustly outperforms the state-of-the-art methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360283/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360283/</guid>
</item>
<item>
<title>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</title>
<description><p><span><big>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</big></span><br></p><p><span><small><i>Anastasia Avdeeva; Aleksei Gusev; Tseren Andzhukaev; Artem Ivanov</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343342</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Whispered speech is a quiet voice without vocalization. One of the common cases of using whispered speech is a technique that can help overcome stuttering. But whispered speech can be uncomfortable and difficult to understand in everyday communication. To address these problems, we propose a method of low-delayed whisper-to-speech voice conversion, which can be useful in real life communication of people with disordered speech. As part of our research, we study the impact of streaming Automatic Speech Recognition models on the quality of voice conversion, comparing different streaming models and methods for model adaptation to streaming settings, and showing the importance of using such models in cases of low-delayed voice conversion.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360259/</guid>
</item>
<item>
<title>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</title>
<description><p><span><big>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</big></span><br></p><p><span><small><i>Denis C. Ilie-Ablachim; Andra Băltoiu; Bogdan Dumitrescu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344313</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We propose sparse representation and dictionary learning algorithms for dictionaries whose atoms are characterized by Gaussian probability distributions around some central atoms. This extension of the space covered by the atoms permits a better characterization of localized features of the signals. The representations are computed by trading-off representation error and the probability of the actual atoms used in the representation. We propose two types of algorithms: a greedy one, similar in spirit with Orthogonal Matching Pursuit, and one based on L1-regularization. We apply our new algorithms to unsupervised anomaly detection, where the representation error is used as anomaly score. The flexibility of our approach appears to improve more the representations of the many normal signals than those of the few outliers, at least for an anomaly type called dependency, thus improving the detection quality. Comparison with both standard dictionary learning algorithms and established anomaly detection methods is favorable.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365333/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365333/</guid>
</item>
<item>
<title>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</title>
<description><p><span><big>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</big></span><br></p><p><span><small><i>Yuya Moroto; Yingrui Ye; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344079</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>There are various sentiment theories for categorizing human sentiments into several discrete sentiment categories, which means that the theory used for training sentiment prediction methods does not always match that used in the test phase. As a solution to this problem, zero-shot visual sentiment prediction methods have been proposed to predict unseen sentiments for which no images are available in the training phase. However, the training of these previous zero-shot methods relies on a single sentiment theory, which limits their ability to handle sentiments from other theories. Thus, this article proposes a more robust zero-shot visual sentiment prediction method that can handle cross-domain sentiments defined in different sentiment theories. Specifically, by focusing on the fact that sentiments are abstract concepts common to humans regardless of whether their theories are different, we incorporate knowledge distillation into our method to construct a teacher–student model that can train the implicit relationships between sentiments defined in different sentiment theories. Furthermore, to enhance sentiment discrimination capability and strengthen the implicit relationships between sentiments, we introduce a novel sentiment loss between the teacher and student models. In this way, our model becomes robust to unseen sentiments by exploiting the implicit relationships between sentiments. The contributions of this article are the introduction of knowledge distillation and a novel sentiment loss between the teacher and student models for zero-shot visual sentiment prediction, and improved performance of zero-shot visual sentiment prediction. Experiments on several open datasets demonstrate the effectiveness of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363382/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363382/</guid>
</item>
<item>
<title>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</title>
<description><p><span><big>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</big></span><br></p><p><span><small><i>Chengen Liu; Geert Leus; Elvin Isufi</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339376</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The edge flow reconstruction task consists of retreiving edge flow signals from corrupted or incomplete measurements. This is typically solved by a regularized optimization problem on higher-order networks such as simplicial complexes and the corresponding regularizers are chosen based on prior knowledge. Tailoring this prior to the setting of interest can be challenging or it may not even be possible. Thus, we consider to learn this prior knowledge via a model-based deep learning approach. We propose a new regularized optimization problem for the simplicial edge flow reconstruction task, the simplicial ElasticNet, which combines the advantages of the
$\ell _{1}$
and
$\ell _{2}$
norms. We solve the simplicial ElasticNet problem via the multi-block alternating direction method of multipliers (ADMM) algorithm and provide conditions on its convergence. By unrolling the ADMM iterative steps, we develop a model-based neural network with a low requirement on the number of training data. This unrolling network replaces the fixed parameters in the iterative algorithm by learnable weights, thus exploiting the neural network's learning capability while preserving the iterative algorithm's interpretability. We enhance this unrolling network via simplicial convolutional filters to aggregate information from the edge flow neighbors, ultimately, improving the network learning expressivity. Extensive experiments on real-world and synthetic datasets validate the proposed approaches and show considerable improvements over both baselines and traditional non-model-based neural networks.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342735/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342735/</guid>
</item>
<item>
<title>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</title>
<description><p><span><big>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</big></span><br></p><p><span><small><i>Jim Beckers; Bart Van Erp; Ziyue Zhao; Kirill Kondrashov; Bert De Vries</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337718</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Bayesian model reduction provides an efficient approach for comparing the performance of all nested sub-models of a model, without re-evaluating any of these sub-models. Until now, Bayesian model reduction has been applied mainly in the computational neuroscience community on simple models. In this paper, we formulate and apply Bayesian model reduction to perform principled pruning of Bayesian neural networks, based on variational free energy minimization. Direct application of Bayesian model reduction, however, gives rise to approximation errors. Therefore, a novel iterative pruning algorithm is presented to alleviate the problems arising with naive Bayesian model reduction, as supported experimentally on the publicly available UCI datasets for different inference algorithms. This novel parameter pruning scheme solves the shortcomings of current state-of-the-art pruning methods that are used by the signal processing community. The proposed approach has a clear stopping criterion and minimizes the same objective that is used during training. Next to these benefits, our experiments indicate better model performance in comparison to state-of-the-art pruning schemes.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334001/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334001/</guid>
</item>
<item>
<title>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</title>
<description><p><span><big>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</big></span><br></p><p><span><small><i>Anh Minh Truong; Wilfried Philips; Peter Veelaert</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340064</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Many image-based rendering (IBR) methods rely on depth estimates obtained from structured light or time-of-flight depth sensors to synthesize novel views from sparse camera networks. However, these estimates often contain missing or noisy regions, resulting in an incorrect mapping between source and target views. This situation makes the fusion process more challenging, as the visual information is misaligned, inconsistent, or missing. In this work, we first implement a lightweight network based on the transformer, which is well-known for its capability to model long-range relationships within input data, to extract spatial features from color images. These features are then used to enhance the quality of completed depth maps. Furthermore, we combine a sequential deep neural network with a spatial attention mechanism to effectively fuse the projected features from multiple source viewpoints. This approach enables us to integrate information from an arbitrary number of source viewpoints as well as improve accuracy in synthesized views. Experimental results on challenging datasets demonstrate that our method achieves superior synthesized image quality compared to state-of-the-art (SOTA) methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345792/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345792/</guid>
</item>
<item>
<title>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</title>
<description><p><span><big>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</big></span><br></p><p><span><small><i>Reza Mirzaeifard; Naveen K. D. Venkategowda; Vinay Chakravarthi Gogineni; Stefan Werner</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344395</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper investigates quantile regression in the presence of non-convex and non-smooth sparse penalties, such as the minimax concave penalty (MCP) and smoothly clipped absolute deviation (SCAD). The non-smooth and non-convex nature of these problems often leads to convergence difficulties for many algorithms. While iterative techniques such as coordinate descent and local linear approximation can facilitate convergence, the process is often slow. This sluggish pace is primarily due to the need to run these approximation techniques until full convergence at each step, a requirement we term as a
secondary convergence iteration
. To accelerate the convergence speed, we employ the alternating direction method of multipliers (ADMM) and introduce a novel single-loop smoothing ADMM algorithm with an increasing penalty parameter, named SIAD, specifically tailored for sparse-penalized quantile regression. We first delve into the convergence properties of the proposed SIAD algorithm and establish the necessary conditions for convergence. Theoretically, we confirm a convergence rate of
$o({k^{-\frac{1}{4}}})$
for the sub-gradient bound of the augmented Lagrangian, where
$k$
denotes the number of iterations. Subsequently, we provide numerical results to showcase the effectiveness of the SIAD algorithm. Our findings highlight that the SIAD method outperforms existing approaches, providing a faster and more stable solution for sparse-penalized quantile regression.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365338/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365338/</guid>
</item>
<item>
<title>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</title>
<description><p><span><big>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</big></span><br></p><p><span><small><i>Oliver Lang; Christian Hofbauer; Reinhard Feger; Mario Huemer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343308</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>A promising waveform candidate for future joint sensing and communication systems is orthogonal frequency-division multiplexing (OFDM). For such systems, supporting multiple transmit antennas requires multiplexing methods for the generation of orthogonal transmit signals, where equidistant subcarrier interleaving (ESI) is the most popular multiplexing method. In this work, we analyze a multiplexing method called Doppler-division multiplexing (DDM). This method applies a phase shift from OFDM symbol to OFDM symbol to separate signals transmitted by different Tx antennas along the velocity axis of the range-Doppler map. The main focus of this work lies on the implications of DDM on the communication task. It will be shown that for DDM, the channels observed in the communication receiver are heavily time-varying, preventing any meaningful transmission of data when not taken into account. In this work, a communication system designed to combat these time-varying channels is proposed, which includes methods for data estimation, synchronization, and channel estimation. Bit error ratio (BER) simulations demonstrate the superiority of this communications system compared to ESI-based systems.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360223/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360223/</guid>
</item>
<item>
<title>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</title>
<description><p><span><big>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</big></span><br></p><p><span><small><i>Nils L. Westhausen; Bernd T. Meyer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343320</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we introduce a causal low-latency low-complexity approach for binaural multichannel blind speaker separation in noisy reverberant conditions. The model, referred to as Group Communication Binaural Filter and Sum Network (GCBFSnet) predicts complex filters for filter-and-sum beamforming in the time-frequency domain. We apply Group Communication (GC), i.e., latent model variables are split into groups and processed with a shared sequence model with the aim of reducing the complexity of a simple model only containing one convolutional and one recurrent module. With GC we are able to reduce the size of the model by up to 83% and the complexity up to 73% compared to the model without GC, while mostly retaining performance. Even for the smallest model configuration, GCBFSnet matches the performance of a low-complexity TasNet baseline in most metrics despite the larger size and higher number of required operations of the baseline.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360275/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360275/</guid>
</item>
<item>
<title>Reverse Ordering Techniques for Attention-Based Channel Prediction</title>
<description><p><span><big>Reverse Ordering Techniques for Attention-Based Channel Prediction</big></span><br></p><p><span><small><i>Valentina Rizzello; Benedikt Böck; Michael Joham; Wolfgang Utschick</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344024</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Channel state information (CSI) is crucial for enhancing the performance of wireless systems by allowing to adjust the transmission strategies based on the current channel conditions. However, obtaining precise CSI is difficult because of the fast-changing channel conditions caused by multi-path fading. An inaccurate CSI hinders the performance of various adaptive wireless systems, highlighting the need for channel prediction techniques to effectively mitigate the drawbacks of outdated CSI. Conventional methods typically depend on assumptions regarding user velocity or require knowledge of the Doppler frequency. In contrast to existing approaches, we aim for a more robust and practical solution by training neural networks without making any assumptions about user velocity, relying solely on noisy channel observations during training. Specifically, we adapt both the sequence-to-sequence with attention (Seq2Seq-attn) and transformer models for channel prediction. Additionally, a new technique called reverse positional encoding is introduced in the transformer model to improve the robustness of the model against varying sequence lengths. Similarly, the encoder outputs of the Seq2Seq-attn model are reversed prior to the application of attention mechanisms. By means of simulations, we show that these proposed techniques enable the models to effectively capture relationships within sequences of channel snapshots without increasing the complexity. Importantly, this capability remains robust across varying sequence lengths, representing a substantial improvement over existing methodologies.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363354/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363354/</guid>
</item>
<item>
<title>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</title>
<description><p><span><big>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</big></span><br></p><p><span><small><i>Sarina Meyer; Xiaoxiao Miao; Ngoc Thang Vu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344375</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Speaker anonymization is the task of modifying a speech recording such that the original speaker cannot be identified anymore. Since the first Voice Privacy Challenge in 2020, along with the release of a framework, the popularity of this research topic is continually increasing. However, the comparison and combination of different anonymization approaches remains challenging due to the complexity of evaluation and the absence of user-friendly research frameworks. We therefore propose an efficient speaker anonymization and evaluation framework based on a modular and easily extendable structure, almost fully in Python. The framework facilitates the orchestration of several anonymization approaches in parallel and allows for interfacing between different techniques. Furthermore, we propose modifications to common evaluation methods which improves the quality of the evaluation and reduces their computation time by 65 to 95%, depending on the metric. Our code is fully open source.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365329/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365329/</guid>
</item>
<item>
<title>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</title>
<description><p><span><big>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</big></span><br></p><p><span><small><i>Alessandro Ilic Mezza; Matteo Amerena; Alberto Bernardini; Augusto Sarti</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343318</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Real-time audio communications over IP have become essential to our daily lives. Packet-switched networks, however, are inherently prone to jitter and data losses, thus creating a strong need for effective packet loss concealment (PLC) techniques. Though solutions based on deep learning have made significant progress in that direction as far as speech is concerned, extending the use of such methods to applications of Networked Music Performance (NMP) presents significant challenges, including high fidelity requirements, higher sampling rates, and stringent temporal constraints associated to the simultaneous interaction between remote musicians. In this article, we present PARCnet, a hybrid PLC method that utilizes a feed-forward neural network to estimate the time-domain residual signal of a parallel linear autoregressive model. Objective metrics and a listening test show that PARCnet provides state-of-the-art results while enabling real-time operation on CPU.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360264/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360264/</guid>
</item>
<item>
<title>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</title>
<description><p><span><big>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</big></span><br></p><p><span><small><i>Katerina Zmolikova; Michael Syskind Pedersen; Jesper Jensen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Supervised learning-based speech enhancement methods often work remarkably well in acoustic situations represented in the training corpus but generalize poorly to out-of-domain situations, i.e. situations not seen during training. This stands in the way of further improvement of these methods in realistic scenarios, as collecting paired noisy-clean recordings in the target application domain is typically not feasible. Recording noisy-only in-domain data is, though, much more practical. In this article, we tackle the problem of unsupervised domain adaptation in speech enhancement. Specifically, we propose a way to use in-domain noisy-only data in the training of a neural network to improve upon a model trained solely on out-of-domain paired data. For this, we make use of masked spectrogram prediction, a technique from self-supervised learning that aims to interpolate masked regions of a spectrogram. We hypothesize that masked spectrogram prediction encourages learning of features that represent well both speech and noise components of the noisy signals. These features can then be used to train a more robust speech enhancement system. We evaluate the proposed method on the VoiceBank-DEMAND and LibriFSD50k databases, with WSJ0-CHiME3 serving as the out-of-domain database. We show that the proposed method outperforms both the out-of-domain system and the baseline approaches, i.e. RemixIT and noisy-target training, and also combines well with the previously proposed RemixIT method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360251/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360251/</guid>
</item>
<item>
<title>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</title>
<description><p><span><big>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</big></span><br></p><p><span><small><i>Alexander Karataev; Christian Forsch; Laura Cottatellucci</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3348343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We consider a cell-free massive multiple-input multiple-output (CF-MaMIMO) communication system in the uplink transmission and propose a novel algorithm for blind or semi-blind joint channel estimation and data detection (JCD). We formulate the problem in the framework of bilinear inference and develop a solution based on the expectation propagation (EP) method for both channel estimation and data detection. We propose a new approximation of the joint a posteriori distribution of the channel and data whose representation as a factor graph enables the application of the EP approach using the message-passing technique, local low-complexity computations at the nodes, and an effective modeling of channel-data interplay. The derived algorithm, called bilinear-EP JCD, allows for a distributed implementation among access points (APs) and the central processing unit (CPU) and has polynomial complexity. Our simulation results show that it outperforms other EP-based state-of-the-art polynomial time algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10378663/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10378663/</guid>
</item>
<item>
<title>Adversarial Representation Learning for Robust Privacy Preservation in Audio</title>
<description><p><span><big>Adversarial Representation Learning for Robust Privacy Preservation in Audio</big></span><br></p><p><span><small><i>Shayan Gharib; Minh Tran; Diep Luong; Konstantinos Drossos; Tuomas Virtanen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Sound event detection systems are widely used in various applications such as surveillance and environmental monitoring where data is automatically collected, processed, and sent to a cloud for sound recognition. However, this process may inadvertently reveal sensitive information about users or their surroundings, hence raising privacy concerns. In this study, we propose a novel adversarial training method for learning representations of audio recordings that effectively prevents the detection of speech activity from the latent features of the recordings. The proposed method trains a model to generate invariant latent representations of speech-containing audio recordings that cannot be distinguished from non-speech recordings by a speech classifier. The novelty of our work is in the optimization algorithm, where the speech classifier's weights are regularly replaced with the weights of classifiers trained in a supervised manner. This increases the discrimination power of the speech classifier constantly during the adversarial training, motivating the model to generate latent representations in which speech is not distinguishable, even using new speech classifiers trained outside the adversarial training loop. The proposed method is evaluated against a baseline approach with no privacy measures and a prior adversarial training method, demonstrating a significant reduction in privacy violations compared to the baseline approach. Additionally, we show that the prior adversarial method is practically ineffective for this purpose.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10379095/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10379095/</guid>
</item>
<item>
<title>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</title>
<description><p><span><big>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</big></span><br></p><p><span><small><i>Yasaman Parhizkar; Gene Cheung; Andrew W. Eckford</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>It is a popular hypothesis in neuroscience that ganglion cells in the retina are activated by selectively detecting visual features in an observed scene. While ganglion cell firings can be predicted via data-trained deep neural nets, the n |
Derekmini
changed the title
got to ofetch in /ieee/journal.tsgot to ofetch in /ieee/journal.ts and /ieee/earlyaccess.ts
|
Successfully generated as following: http://localhost:1200/ieee/journal/8782710 - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Open Journal of Signal Processing</title>
<link>https://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=8782710</link>
<atom:link href="http://localhost:1200/ieee/journal/8782710" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Open Journal of Signal Processing - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Wed, 10 Apr 2024 00:45:00 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Synthbuster: Towards Detection of Diffusion Model Generated Images</title>
<description><p><span><big>Synthbuster: Towards Detection of Diffusion Model Generated Images</big></span><br></p><p><span><small><i>Quentin Bammey</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337714</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Synthetically-generated images are getting increasingly popular. Diffusion models have advanced to the stage where even non-experts can generate photo-realistic images from a simple text prompt. They expand creative horizons but also open a Pandora's box of potential disinformation risks. In this context, the present corpus of synthetic image detection techniques, primarily focusing on older generative models like Generative Adversarial Networks, finds itself ill-equipped to deal with this emerging trend. Recognizing this challenge, we introduce a method specifically designed to detect synthetic images produced by diffusion models. Our approach capitalizes on the inherent frequency artefacts left behind during the diffusion process. Spectral analysis is used to highlight the artefacts in the Fourier transform of a residual image, which are used to distinguish real from fake images. The proposed method can detect diffusion-model-generated images even under mild
jpeg
compression, and generalizes relatively well to unknown models. By pioneering this novel approach, we aim to fortify forensic methodologies and ignite further research into the detection of AI-generated images.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334046/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334046/</guid>
</item>
<item>
<title>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</title>
<description><p><span><big>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</big></span><br></p><p><span><small><i>Yanbin Zou; Jingna Fan; Zekai Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we consider using time-of-arrival (TOA) measurements from a single moving receiver to locate a moving target at constant velocity that emits a periodic signal with unknown signal period. First, we give the TOA measurement model and deduce the Cram
$\acute{\text{e}}$
r-Rao lower bounds (CRLB). Then, we formulate a nonlinear least squares (NLS) problem to estimate the unknown parameters. We use semidefinite programming (SDP) techniques to relax the nonconvex NLS problem. However, it is shown that the SDP localization algorithm cannot provide a high-quality solution. Subsequently, we develop a fixed point iteration (FPI) method to improve the performance of the SDP algorithm. In addition, we also consider the presence of receiver position errors and develop the corresponding localization algorithm. Numerical simulations are conducted to demonstrate the localization performance of the proposed algorithms by comparing them with the CRLB.
Index Term
-Fixed point iteration (FPI), semidefinite programming (SDP), single moving receiver, target localization, time-of-arrival (TOA).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336384/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336384/</guid>
</item>
<item>
<title>The Neural-SRP Method for Universal Robust Multi-Source Tracking</title>
<description><p><span><big>The Neural-SRP Method for Universal Robust Multi-Source Tracking</big></span><br></p><p><span><small><i>Eric Grinstein; Christopher M. Hicks; Toon van Waterschoot; Mike Brookes; Patrick A. Naylor</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340057</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Neural networks have achieved state-of-the-art performance on the task of acoustic Direction-of-Arrival (DOA) estimation using microphone arrays. Neural models can be classified as end-to-end or hybrid, each class showing advantages and disadvantages. This work introduces Neural-SRP, an end-to-end neural network architecture for DOA estimation inspired by the classical Steered Response Power (SRP) method, which overcomes limitations of current neural models. We evaluate the architecture on multiple scenarios, namely, multi-source DOA tracking and single-source DOA tracking under the presence of directional and diffuse noise. The experiments demonstrate that our proposed method compares favourably in terms of computational and localization performance with established neural methods on various recorded and simulated benchmark datasets.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345765/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345765/</guid>
</item>
<item>
<title>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</title>
<description><p><span><big>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</big></span><br></p><p><span><small><i>Alexander Venus; Erik Leitinger; Stefan Tertinek; Klaus Witrisal</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper presents a neural-enhanced probabilistic model and corresponding factor graph-based sum-product algorithm for robust localization and tracking in multipath-prone environments. The introduced hybrid probabilistic model consists of physics-based and data-driven measurement models capturing the information contained in both, the line-of-sight (LOS) component as well as in multipath components (NLOS components). The physics-based and data-driven models are embedded in a joint Bayesian framework allowing to derive from first principles a factor graph-based algorithm that fuses the information of these models. The proposed algorithm uses radio signal measurements from multiple base stations to robustly estimate the mobile agent's position together with all model parameters. It provides high localization accuracy by exploiting the position-related information of the LOS component via the physics-based model and robustness by exploiting the geometric imprint of multipath components independent of the propagation channel via the data-driven model. In a challenging numerical experiment involving obstructed LOS situations to all anchors, we show that the proposed sequential algorithm significantly outperforms state-of-the-art methods and attains the posterior Cramér-Rao lower bound even with training data limited to local regions.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336409/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336409/</guid>
</item>
<item>
<title>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</title>
<description><p><span><big>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</big></span><br></p><p><span><small><i>AMIT EFRAIM; Joseph M. Francos</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In solving a model fitting problem, the existence of outliers in the set of measurements can have a devastating effect on the solution accuracy. Traditionally, in order to overcome this problem, robust point cloud registration algorithms are composed of transformation hypothesis generation, followed by hypothesis evaluation aimed at selecting the best hypothesized estimate. Hypotheses evaluation is commonly performed using the sample consensus criterion. However, since this criterion accounts only for the consensus size, it fails when the maximal sample consensus is incorrect. We propose a new hypothesis evaluation approach, generalizing the sample consensus approach, where instead of the sample consensus, the transformation that maximizes the point clouds feature correlation is selected as the best hypothesis. The feature vector at each point contains information such as on local geometry and semantic context. Utilizing this information in the hypotheses evaluation and selection procedure allows for a correct decision even when the hypothesis yielding the maximal sample consensus is false. Consequently, the probability of selecting the correct model increases. We show both mathematically and empirically that substituting the sample consensus criterion with the proposed point cloud feature correlation hypothesis test (PC-FCHT) lowers the probability of large registration errors, compared to using the special case of sample consensus. The proposed PC-FCHT is applicable to any algorithm that follows the hypothesis generation and evaluation scheme, potentially improving the success rates of a wide family of point cloud registration algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345750/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345750/</guid>
</item>
<item>
<title>Joint PAPR and OBP Reduction for NC-OFDM Systems</title>
<description><p><span><big>Joint PAPR and OBP Reduction for NC-OFDM Systems</big></span><br></p><p><span><small><i>Hsuan-Fu Wang; Fang-Biau Ueng; Bo-Heng Yeh</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3329757</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The spectrum resource is always a critical issue for wireless communications since it directly impacts the data rate and capacity. However, the problem of spectrum resource scarcity always exists. Moreover, spectrum resource scarcity becomes more severe as new communication technologies and wireless applications sprout. Noncontiguous orthogonal frequency division multiplexing (NC-OFDM) is a multicarrier method for bandwidth utilization. Unfortunately, this system has two fatal defects: high peak-to-average power ratio (PAPR) and considerable out-of-band power (OBP), which are detrimental to the system's performance. To solve these two problems, we propose a convex optimization-based method for joint PAPR and OBP reduction in NC-OFDM Systems. The strategy is to permit the secondary user to utilize the unoccupied spectrum of the primary user with dynamic spectrum sharing (DSS) based on a cognitive radio network (CRN). To this end, a flexible system operating over noncontiguous bands and DSS scenarios is necessary. The simulation results have shown that our method could effectively improve the overall performance and outperform other schemes, i.e., projections onto convex sets (POCS) and alternating projections onto convex and non-convex sets (APOCNCS), without harming the transmission of the primary system. The collaboration between secondary and primary systems is viable with the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10305259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10305259/</guid>
</item>
<item>
<title>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</title>
<description><p><span><big>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</big></span><br></p><p><span><small><i>Gal Itzhak; Israel Cohen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339433</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article presents a Kronecker-product (KP) beamforming approach incorporating sparse concentric circular arrays (SCCAs). The locations of the microphones on the SCCA are optimized concerning the broadband array directivity over a wide range of direction-of-arrival (DOA) deviations of a desired signal. A maximum directivity factor (MDF) sub-beamformer is derived accordingly with the optimal locations. Then, we propose two global beamformers obtained as a Kronecker product of a uniform linear array (ULA) and the SCCA sub-beamformer. The global beamformers differ by the type of the ULA, which is designed either as an MDF sub-beamformer along the
$\mathsf {x}$
-axis or as a maximum white noise gain sub-beamformer along the
$\mathsf {y}$
-axis. We analyze the performance of the proposed beamformers in terms of the directivity factor, the white noise gain, and their spatial beampatterns. Compared to traditional beamformers, the proposed beamformers exhibit considerably larger tolerance to DOA deviations concerning both the azimuth and elevation angles. Experimental results with speech signals in noisy and reverberant environments demonstrate that the proposed approach outperforms traditional beamformers regarding the perceptual evaluation of speech quality (PESQ) and short-time objective intelligibility (STOI) scores when the desired speech signals deviate from the nominal DOA.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342869/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342869/</guid>
</item>
<item>
<title>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</title>
<description><p><span><big>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</big></span><br></p><p><span><small><i>Karn N. Watcharasupat; Chih-Wei Wu; Yiwei Ding; Iroro Orife; Aaron J. Hipple; Phillip A. Williams; Scott Kramer; Alexander Lerch; William Wolcott</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339428</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Cinematic audio source separation is a relatively new subtask of audio source separation, with the aim of extracting the dialogue, music, and effects stems from their mixture. In this work, we developed a model generalizing the Bandsplit RNN for any complete or overcomplete partitions of the frequency axis. Psychoacoustically motivated frequency scales were used to inform the band definitions which are now defined with redundancy for more reliable feature extraction. A loss function motivated by the signal-to-noise ratio and the sparsity-promoting property of the 1-norm was proposed. We additionally exploit the information-sharing property of a common-encoder setup to reduce computational complexity during both training and inference, improve separation performance for hard-to-generalize classes of sounds, and allow flexibility during inference time with detachable decoders. Our best model sets the state of the art on the Divide and Remaster dataset with performance above the ideal ratio mask for the dialogue stem.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342812/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342812/</guid>
</item>
<item>
<title>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</title>
<description><p><span><big>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</big></span><br></p><p><span><small><i>Miguel Ferrer; María de Diego; Alberto Gonzalez</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340106</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The LMS algorithm is widely employed in adaptive systems due to its robustness, simplicity, and reasonable performance. However, it is well known that this algorithm suffers from a slow convergence speed when dealing with colored reference signals. Numerous variants and alternative algorithms have been proposed to address this issue, though all of them entail an increase in computational cost. Among the proposed alternatives, the affine projection algorithm stands out. This algorithm has the peculiarity of starting from
$N$
data vectors of the reference signal. It transforms these vectors into as many data vectors suitably normalized in energy and mutually orthogonal. In this work, we propose a version of the LMS algorithm that, similar to the affine projection algorithm, starts from
$N$
data vectors of the reference signal but corrects them by using only a scalar factor that functions as a convergence step. Our goal is to align the behavior of this algorithm with the behavior of the affine projection algorithm without significantly increasing the computational cost of the LMS.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345730/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345730/</guid>
</item>
<item>
<title>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</title>
<description><p><span><big>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</big></span><br></p><p><span><small><i>Zongyao Li; Ren Togo; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340616</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article proposes a method for transferring knowledge of semantic segmentation from a labeled source domain to an unlabeled target domain without using the source-domain data. Such a problem is called source-data-free domain adaptation, in which a pre-trained source-domain model and the unlabeled target-domain data are used to transfer the label knowledge across the domains. Like most previous methods, our method uses pseudo labels for distilling and transferring the source-domain knowledge. On the basis of the pseudo-label learning, our method improves the domain adaptation performance in two innovative ways: 1) reducing the domain differences by source-data-free style transfer and 2) exploring the style diversity within the target domain by style modification. To this end, we introduce two additional modules: 1) an inter-domain style transfer module which aligns the feature statistics of the source and target domains before producing the pseudo labels thereby improving the pseudo labels' accuracy, and 2) an intra-domain style modification module which modifies the image styles within the target domain for learning intra-domain style-invariant features. Our method with the two modules outperforms previous source-data-free domain adaptation methods in two commonly used benchmarks. Moreover, our method is well compatible with the previous methods for further improvement.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10356748/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10356748/</guid>
</item>
<item>
<title>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</title>
<description><p><span><big>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</big></span><br></p><p><span><small><i>Tobias Kabzinski; Peter Jax</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337721</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Quasi-continuous system identification of time-variant linear acoustic systems can be applied in various audio signal processing applications when numerous acoustic transfer functions must be measured. A prominent application is measuring head-related transfer functions. We treat the underlying multiple-input-multiple-output (MIMO) system identification problem in a state-space model as a joint estimation problem for states, representing impulse responses, and state-space model parameters using the expectation maximization (EM) algorithm. We address limitations of prior work by imposing different model structures, especially for dependencies within a (transformed) state vector. This results in block diagonal matrix structures, for which we derive M-step update rules. Making assumptions about this model structure and choosing a block size for a given application define the computational complexity. In examples, we found that applying this framework yields improvements of up to 10 dB in relative system distance in comparison to a conventional method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334061/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334061/</guid>
</item>
<item>
<title>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</title>
<description><p><span><big>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</big></span><br></p><p><span><small><i>Damir Rakhimov; Martin Haardt</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337729</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this paper, we present an analytical performance assessment of 2-D Tensor ESPRIT in terms of physical parameters. We show that the error in the
$r$
-mode depends only on two components, irrespective of the dimensionality of the problem. We obtain analytical expressions in closed form for the mean squared error (MSE) in each dimension as a function of the signal-to-noise (SNR) ratio, the array steering matrices, the number of antennas, the number of snapshots, the selection matrices, and the signal correlation. The derived expressions allow a better understanding of the difference in performance between the tensor and the matrix versions of the ESPRIT algorithm. The simulation results confirm the coincidence between the presented analytical expression and the curves obtained via Monte Carlo trials. We analyze the behavior of each of the two error components in different scenarios.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334446/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334446/</guid>
</item>
<item>
<title>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</title>
<description><p><span><big>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</big></span><br></p><p><span><small><i>Yaman Kındap; Simon Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343341</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this work we study linear vector stochastic differential equation (SDE) models driven by the generalised hyperbolic (GH) Lévy process for inference in continuous-time non-Gaussian filtering problems. The GH family of stochastic processes offers a flexible framework for modelling of non-Gaussian, heavy-tailed characteristics and includes the normal inverse-Gaussian, variance-gamma and Student-t processes as special cases. We present continuous-time simulation methods for the solution of vector SDE models driven by GH processes and novel inference methodologies using a variant of sequential Markov chain Monte Carlo (MCMC). As an example a particular formulation of Langevin dynamics is studied within this framework. The model is applied to both a synthetically generated data set and a real-world financial series to demonstrate its capabilities.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360268/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360268/</guid>
</item>
<item>
<title>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</title>
<description><p><span><big>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</big></span><br></p><p><span><small><i>James M. Cozens; Simon J. Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344048</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper proposes a probabilistic approach for extracting time-varying and irregular time signature information from polyphonic audio extracts, subsequently providing beat and bar line positions given inferred time signature divisions. This is achieved via dynamically evaluating the beat tempo as a function of time through finding an optimal compromise in beat and bar alignment in the time and tempo domains. Time signature divisions are determined based on a new representation, termed the Metrogram, that presents time-varying information regarding rhythmic and metric periodicities in the Tempogram. Our methodology is characterised by its ability to provide a distribution over metric interpretations, offering insights into the diverse ways music can be rhythmically perceived. Results indicate high-level accuracy for a variety of polyphonic extracts containing irregular, complex, irrational, and time-varying time signatures. Accuracy rivalling state-of-the-art methodologies is also reported in a beat tracking task performed on the standard Ballroom Dataset. The paper offers insights into the field of dynamic time signature recognition and beat tracking, offering a valuable and versatile resource for the analysis, composition, and performance of music.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363392/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363392/</guid>
</item>
<item>
<title>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</title>
<description><p><span><big>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</big></span><br></p><p><span><small><i>Yuto Watanabe; Ren Togo; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343335</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Although text-guided image manipulation approaches have demonstrated highly accurate performance for editing the appearance of images in a virtual or simple scenario, their real-world applications face significant challenges. The primary cause of these challenges is the misalignment in the distribution of training and real-world data, which leads to unstable text-guided image manipulation. In this work, we propose a novel framework called TolerantGAN and tackle the new task of real-world text-guided image manipulation independent of the training data. To achieve this, we introduce two key concepts of a border smoothly connection module (BSCM) and a manipulation direction-based attention module (MDAM). BSCM smoothens the misalignment in the distribution of training and real-world data. MDAM extracts only regions highly relevant for image manipulation and assists in reconstructing unobserved objects in the training data. For in-the-wild input images of various classes, TolerantGAN robustly outperforms the state-of-the-art methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360283/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360283/</guid>
</item>
<item>
<title>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</title>
<description><p><span><big>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</big></span><br></p><p><span><small><i>Anastasia Avdeeva; Aleksei Gusev; Tseren Andzhukaev; Artem Ivanov</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343342</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Whispered speech is a quiet voice without vocalization. One of the common cases of using whispered speech is a technique that can help overcome stuttering. But whispered speech can be uncomfortable and difficult to understand in everyday communication. To address these problems, we propose a method of low-delayed whisper-to-speech voice conversion, which can be useful in real life communication of people with disordered speech. As part of our research, we study the impact of streaming Automatic Speech Recognition models on the quality of voice conversion, comparing different streaming models and methods for model adaptation to streaming settings, and showing the importance of using such models in cases of low-delayed voice conversion.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360259/</guid>
</item>
<item>
<title>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</title>
<description><p><span><big>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</big></span><br></p><p><span><small><i>Denis C. Ilie-Ablachim; Andra Băltoiu; Bogdan Dumitrescu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344313</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We propose sparse representation and dictionary learning algorithms for dictionaries whose atoms are characterized by Gaussian probability distributions around some central atoms. This extension of the space covered by the atoms permits a better characterization of localized features of the signals. The representations are computed by trading-off representation error and the probability of the actual atoms used in the representation. We propose two types of algorithms: a greedy one, similar in spirit with Orthogonal Matching Pursuit, and one based on L1-regularization. We apply our new algorithms to unsupervised anomaly detection, where the representation error is used as anomaly score. The flexibility of our approach appears to improve more the representations of the many normal signals than those of the few outliers, at least for an anomaly type called dependency, thus improving the detection quality. Comparison with both standard dictionary learning algorithms and established anomaly detection methods is favorable.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365333/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365333/</guid>
</item>
<item>
<title>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</title>
<description><p><span><big>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</big></span><br></p><p><span><small><i>Yuya Moroto; Yingrui Ye; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344079</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>There are various sentiment theories for categorizing human sentiments into several discrete sentiment categories, which means that the theory used for training sentiment prediction methods does not always match that used in the test phase. As a solution to this problem, zero-shot visual sentiment prediction methods have been proposed to predict unseen sentiments for which no images are available in the training phase. However, the training of these previous zero-shot methods relies on a single sentiment theory, which limits their ability to handle sentiments from other theories. Thus, this article proposes a more robust zero-shot visual sentiment prediction method that can handle cross-domain sentiments defined in different sentiment theories. Specifically, by focusing on the fact that sentiments are abstract concepts common to humans regardless of whether their theories are different, we incorporate knowledge distillation into our method to construct a teacher–student model that can train the implicit relationships between sentiments defined in different sentiment theories. Furthermore, to enhance sentiment discrimination capability and strengthen the implicit relationships between sentiments, we introduce a novel sentiment loss between the teacher and student models. In this way, our model becomes robust to unseen sentiments by exploiting the implicit relationships between sentiments. The contributions of this article are the introduction of knowledge distillation and a novel sentiment loss between the teacher and student models for zero-shot visual sentiment prediction, and improved performance of zero-shot visual sentiment prediction. Experiments on several open datasets demonstrate the effectiveness of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363382/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363382/</guid>
</item>
<item>
<title>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</title>
<description><p><span><big>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</big></span><br></p><p><span><small><i>Chengen Liu; Geert Leus; Elvin Isufi</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339376</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The edge flow reconstruction task consists of retreiving edge flow signals from corrupted or incomplete measurements. This is typically solved by a regularized optimization problem on higher-order networks such as simplicial complexes and the corresponding regularizers are chosen based on prior knowledge. Tailoring this prior to the setting of interest can be challenging or it may not even be possible. Thus, we consider to learn this prior knowledge via a model-based deep learning approach. We propose a new regularized optimization problem for the simplicial edge flow reconstruction task, the simplicial ElasticNet, which combines the advantages of the
$\ell _{1}$
and
$\ell _{2}$
norms. We solve the simplicial ElasticNet problem via the multi-block alternating direction method of multipliers (ADMM) algorithm and provide conditions on its convergence. By unrolling the ADMM iterative steps, we develop a model-based neural network with a low requirement on the number of training data. This unrolling network replaces the fixed parameters in the iterative algorithm by learnable weights, thus exploiting the neural network's learning capability while preserving the iterative algorithm's interpretability. We enhance this unrolling network via simplicial convolutional filters to aggregate information from the edge flow neighbors, ultimately, improving the network learning expressivity. Extensive experiments on real-world and synthetic datasets validate the proposed approaches and show considerable improvements over both baselines and traditional non-model-based neural networks.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342735/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342735/</guid>
</item>
<item>
<title>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</title>
<description><p><span><big>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</big></span><br></p><p><span><small><i>Jim Beckers; Bart Van Erp; Ziyue Zhao; Kirill Kondrashov; Bert De Vries</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337718</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Bayesian model reduction provides an efficient approach for comparing the performance of all nested sub-models of a model, without re-evaluating any of these sub-models. Until now, Bayesian model reduction has been applied mainly in the computational neuroscience community on simple models. In this paper, we formulate and apply Bayesian model reduction to perform principled pruning of Bayesian neural networks, based on variational free energy minimization. Direct application of Bayesian model reduction, however, gives rise to approximation errors. Therefore, a novel iterative pruning algorithm is presented to alleviate the problems arising with naive Bayesian model reduction, as supported experimentally on the publicly available UCI datasets for different inference algorithms. This novel parameter pruning scheme solves the shortcomings of current state-of-the-art pruning methods that are used by the signal processing community. The proposed approach has a clear stopping criterion and minimizes the same objective that is used during training. Next to these benefits, our experiments indicate better model performance in comparison to state-of-the-art pruning schemes.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334001/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334001/</guid>
</item>
<item>
<title>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</title>
<description><p><span><big>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</big></span><br></p><p><span><small><i>Anh Minh Truong; Wilfried Philips; Peter Veelaert</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340064</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Many image-based rendering (IBR) methods rely on depth estimates obtained from structured light or time-of-flight depth sensors to synthesize novel views from sparse camera networks. However, these estimates often contain missing or noisy regions, resulting in an incorrect mapping between source and target views. This situation makes the fusion process more challenging, as the visual information is misaligned, inconsistent, or missing. In this work, we first implement a lightweight network based on the transformer, which is well-known for its capability to model long-range relationships within input data, to extract spatial features from color images. These features are then used to enhance the quality of completed depth maps. Furthermore, we combine a sequential deep neural network with a spatial attention mechanism to effectively fuse the projected features from multiple source viewpoints. This approach enables us to integrate information from an arbitrary number of source viewpoints as well as improve accuracy in synthesized views. Experimental results on challenging datasets demonstrate that our method achieves superior synthesized image quality compared to state-of-the-art (SOTA) methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345792/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345792/</guid>
</item>
<item>
<title>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</title>
<description><p><span><big>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</big></span><br></p><p><span><small><i>Reza Mirzaeifard; Naveen K. D. Venkategowda; Vinay Chakravarthi Gogineni; Stefan Werner</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344395</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper investigates quantile regression in the presence of non-convex and non-smooth sparse penalties, such as the minimax concave penalty (MCP) and smoothly clipped absolute deviation (SCAD). The non-smooth and non-convex nature of these problems often leads to convergence difficulties for many algorithms. While iterative techniques such as coordinate descent and local linear approximation can facilitate convergence, the process is often slow. This sluggish pace is primarily due to the need to run these approximation techniques until full convergence at each step, a requirement we term as a
secondary convergence iteration
. To accelerate the convergence speed, we employ the alternating direction method of multipliers (ADMM) and introduce a novel single-loop smoothing ADMM algorithm with an increasing penalty parameter, named SIAD, specifically tailored for sparse-penalized quantile regression. We first delve into the convergence properties of the proposed SIAD algorithm and establish the necessary conditions for convergence. Theoretically, we confirm a convergence rate of
$o({k^{-\frac{1}{4}}})$
for the sub-gradient bound of the augmented Lagrangian, where
$k$
denotes the number of iterations. Subsequently, we provide numerical results to showcase the effectiveness of the SIAD algorithm. Our findings highlight that the SIAD method outperforms existing approaches, providing a faster and more stable solution for sparse-penalized quantile regression.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365338/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365338/</guid>
</item>
<item>
<title>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</title>
<description><p><span><big>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</big></span><br></p><p><span><small><i>Oliver Lang; Christian Hofbauer; Reinhard Feger; Mario Huemer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343308</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>A promising waveform candidate for future joint sensing and communication systems is orthogonal frequency-division multiplexing (OFDM). For such systems, supporting multiple transmit antennas requires multiplexing methods for the generation of orthogonal transmit signals, where equidistant subcarrier interleaving (ESI) is the most popular multiplexing method. In this work, we analyze a multiplexing method called Doppler-division multiplexing (DDM). This method applies a phase shift from OFDM symbol to OFDM symbol to separate signals transmitted by different Tx antennas along the velocity axis of the range-Doppler map. The main focus of this work lies on the implications of DDM on the communication task. It will be shown that for DDM, the channels observed in the communication receiver are heavily time-varying, preventing any meaningful transmission of data when not taken into account. In this work, a communication system designed to combat these time-varying channels is proposed, which includes methods for data estimation, synchronization, and channel estimation. Bit error ratio (BER) simulations demonstrate the superiority of this communications system compared to ESI-based systems.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360223/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360223/</guid>
</item>
<item>
<title>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</title>
<description><p><span><big>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</big></span><br></p><p><span><small><i>Nils L. Westhausen; Bernd T. Meyer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343320</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we introduce a causal low-latency low-complexity approach for binaural multichannel blind speaker separation in noisy reverberant conditions. The model, referred to as Group Communication Binaural Filter and Sum Network (GCBFSnet) predicts complex filters for filter-and-sum beamforming in the time-frequency domain. We apply Group Communication (GC), i.e., latent model variables are split into groups and processed with a shared sequence model with the aim of reducing the complexity of a simple model only containing one convolutional and one recurrent module. With GC we are able to reduce the size of the model by up to 83% and the complexity up to 73% compared to the model without GC, while mostly retaining performance. Even for the smallest model configuration, GCBFSnet matches the performance of a low-complexity TasNet baseline in most metrics despite the larger size and higher number of required operations of the baseline.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360275/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360275/</guid>
</item>
<item>
<title>Reverse Ordering Techniques for Attention-Based Channel Prediction</title>
<description><p><span><big>Reverse Ordering Techniques for Attention-Based Channel Prediction</big></span><br></p><p><span><small><i>Valentina Rizzello; Benedikt Böck; Michael Joham; Wolfgang Utschick</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344024</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Channel state information (CSI) is crucial for enhancing the performance of wireless systems by allowing to adjust the transmission strategies based on the current channel conditions. However, obtaining precise CSI is difficult because of the fast-changing channel conditions caused by multi-path fading. An inaccurate CSI hinders the performance of various adaptive wireless systems, highlighting the need for channel prediction techniques to effectively mitigate the drawbacks of outdated CSI. Conventional methods typically depend on assumptions regarding user velocity or require knowledge of the Doppler frequency. In contrast to existing approaches, we aim for a more robust and practical solution by training neural networks without making any assumptions about user velocity, relying solely on noisy channel observations during training. Specifically, we adapt both the sequence-to-sequence with attention (Seq2Seq-attn) and transformer models for channel prediction. Additionally, a new technique called reverse positional encoding is introduced in the transformer model to improve the robustness of the model against varying sequence lengths. Similarly, the encoder outputs of the Seq2Seq-attn model are reversed prior to the application of attention mechanisms. By means of simulations, we show that these proposed techniques enable the models to effectively capture relationships within sequences of channel snapshots without increasing the complexity. Importantly, this capability remains robust across varying sequence lengths, representing a substantial improvement over existing methodologies.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363354/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363354/</guid>
</item>
<item>
<title>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</title>
<description><p><span><big>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</big></span><br></p><p><span><small><i>Sarina Meyer; Xiaoxiao Miao; Ngoc Thang Vu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344375</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Speaker anonymization is the task of modifying a speech recording such that the original speaker cannot be identified anymore. Since the first Voice Privacy Challenge in 2020, along with the release of a framework, the popularity of this research topic is continually increasing. However, the comparison and combination of different anonymization approaches remains challenging due to the complexity of evaluation and the absence of user-friendly research frameworks. We therefore propose an efficient speaker anonymization and evaluation framework based on a modular and easily extendable structure, almost fully in Python. The framework facilitates the orchestration of several anonymization approaches in parallel and allows for interfacing between different techniques. Furthermore, we propose modifications to common evaluation methods which improves the quality of the evaluation and reduces their computation time by 65 to 95%, depending on the metric. Our code is fully open source.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365329/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365329/</guid>
</item>
<item>
<title>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</title>
<description><p><span><big>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</big></span><br></p><p><span><small><i>Alessandro Ilic Mezza; Matteo Amerena; Alberto Bernardini; Augusto Sarti</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343318</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Real-time audio communications over IP have become essential to our daily lives. Packet-switched networks, however, are inherently prone to jitter and data losses, thus creating a strong need for effective packet loss concealment (PLC) techniques. Though solutions based on deep learning have made significant progress in that direction as far as speech is concerned, extending the use of such methods to applications of Networked Music Performance (NMP) presents significant challenges, including high fidelity requirements, higher sampling rates, and stringent temporal constraints associated to the simultaneous interaction between remote musicians. In this article, we present PARCnet, a hybrid PLC method that utilizes a feed-forward neural network to estimate the time-domain residual signal of a parallel linear autoregressive model. Objective metrics and a listening test show that PARCnet provides state-of-the-art results while enabling real-time operation on CPU.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360264/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360264/</guid>
</item>
<item>
<title>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</title>
<description><p><span><big>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</big></span><br></p><p><span><small><i>Katerina Zmolikova; Michael Syskind Pedersen; Jesper Jensen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Supervised learning-based speech enhancement methods often work remarkably well in acoustic situations represented in the training corpus but generalize poorly to out-of-domain situations, i.e. situations not seen during training. This stands in the way of further improvement of these methods in realistic scenarios, as collecting paired noisy-clean recordings in the target application domain is typically not feasible. Recording noisy-only in-domain data is, though, much more practical. In this article, we tackle the problem of unsupervised domain adaptation in speech enhancement. Specifically, we propose a way to use in-domain noisy-only data in the training of a neural network to improve upon a model trained solely on out-of-domain paired data. For this, we make use of masked spectrogram prediction, a technique from self-supervised learning that aims to interpolate masked regions of a spectrogram. We hypothesize that masked spectrogram prediction encourages learning of features that represent well both speech and noise components of the noisy signals. These features can then be used to train a more robust speech enhancement system. We evaluate the proposed method on the VoiceBank-DEMAND and LibriFSD50k databases, with WSJ0-CHiME3 serving as the out-of-domain database. We show that the proposed method outperforms both the out-of-domain system and the baseline approaches, i.e. RemixIT and noisy-target training, and also combines well with the previously proposed RemixIT method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360251/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360251/</guid>
</item>
<item>
<title>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</title>
<description><p><span><big>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</big></span><br></p><p><span><small><i>Alexander Karataev; Christian Forsch; Laura Cottatellucci</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3348343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We consider a cell-free massive multiple-input multiple-output (CF-MaMIMO) communication system in the uplink transmission and propose a novel algorithm for blind or semi-blind joint channel estimation and data detection (JCD). We formulate the problem in the framework of bilinear inference and develop a solution based on the expectation propagation (EP) method for both channel estimation and data detection. We propose a new approximation of the joint a posteriori distribution of the channel and data whose representation as a factor graph enables the application of the EP approach using the message-passing technique, local low-complexity computations at the nodes, and an effective modeling of channel-data interplay. The derived algorithm, called bilinear-EP JCD, allows for a distributed implementation among access points (APs) and the central processing unit (CPU) and has polynomial complexity. Our simulation results show that it outperforms other EP-based state-of-the-art polynomial time algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10378663/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10378663/</guid>
</item>
<item>
<title>Adversarial Representation Learning for Robust Privacy Preservation in Audio</title>
<description><p><span><big>Adversarial Representation Learning for Robust Privacy Preservation in Audio</big></span><br></p><p><span><small><i>Shayan Gharib; Minh Tran; Diep Luong; Konstantinos Drossos; Tuomas Virtanen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Sound event detection systems are widely used in various applications such as surveillance and environmental monitoring where data is automatically collected, processed, and sent to a cloud for sound recognition. However, this process may inadvertently reveal sensitive information about users or their surroundings, hence raising privacy concerns. In this study, we propose a novel adversarial training method for learning representations of audio recordings that effectively prevents the detection of speech activity from the latent features of the recordings. The proposed method trains a model to generate invariant latent representations of speech-containing audio recordings that cannot be distinguished from non-speech recordings by a speech classifier. The novelty of our work is in the optimization algorithm, where the speech classifier's weights are regularly replaced with the weights of classifiers trained in a supervised manner. This increases the discrimination power of the speech classifier constantly during the adversarial training, motivating the model to generate latent representations in which speech is not distinguishable, even using new speech classifiers trained outside the adversarial training loop. The proposed method is evaluated against a baseline approach with no privacy measures and a prior adversarial training method, demonstrating a significant reduction in privacy violations compared to the baseline approach. Additionally, we show that the prior adversarial method is practically ineffective for this purpose.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10379095/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10379095/</guid>
</item>
<item>
<title>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</title>
<description><p><span><big>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</big></span><br></p><p><span><small><i>Yasaman Parhizkar; Gene Cheung; Andrew W. Eckford</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>It is a popular hypothesis in neuroscience that ganglion cells in the retina are activated by selectively detecting visual features in an observed scene. While ganglion cell firings can be predicted via data-trained deep neural nets, the n |
|
Successfully generated as following: http://localhost:1200/ieee/journal/8782710 - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Open Journal of Signal Processing</title>
<link>https://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=8782710</link>
<atom:link href="http://localhost:1200/ieee/journal/8782710" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Open Journal of Signal Processing - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Wed, 10 Apr 2024 00:53:27 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Synthbuster: Towards Detection of Diffusion Model Generated Images</title>
<description><p><span><big>Synthbuster: Towards Detection of Diffusion Model Generated Images</big></span><br></p><p><span><small><i>Quentin Bammey</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337714</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Synthetically-generated images are getting increasingly popular. Diffusion models have advanced to the stage where even non-experts can generate photo-realistic images from a simple text prompt. They expand creative horizons but also open a Pandora's box of potential disinformation risks. In this context, the present corpus of synthetic image detection techniques, primarily focusing on older generative models like Generative Adversarial Networks, finds itself ill-equipped to deal with this emerging trend. Recognizing this challenge, we introduce a method specifically designed to detect synthetic images produced by diffusion models. Our approach capitalizes on the inherent frequency artefacts left behind during the diffusion process. Spectral analysis is used to highlight the artefacts in the Fourier transform of a residual image, which are used to distinguish real from fake images. The proposed method can detect diffusion-model-generated images even under mild
jpeg
compression, and generalizes relatively well to unknown models. By pioneering this novel approach, we aim to fortify forensic methodologies and ignite further research into the detection of AI-generated images.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334046/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334046/</guid>
</item>
<item>
<title>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</title>
<description><p><span><big>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</big></span><br></p><p><span><small><i>Yanbin Zou; Jingna Fan; Zekai Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we consider using time-of-arrival (TOA) measurements from a single moving receiver to locate a moving target at constant velocity that emits a periodic signal with unknown signal period. First, we give the TOA measurement model and deduce the Cram
$\acute{\text{e}}$
r-Rao lower bounds (CRLB). Then, we formulate a nonlinear least squares (NLS) problem to estimate the unknown parameters. We use semidefinite programming (SDP) techniques to relax the nonconvex NLS problem. However, it is shown that the SDP localization algorithm cannot provide a high-quality solution. Subsequently, we develop a fixed point iteration (FPI) method to improve the performance of the SDP algorithm. In addition, we also consider the presence of receiver position errors and develop the corresponding localization algorithm. Numerical simulations are conducted to demonstrate the localization performance of the proposed algorithms by comparing them with the CRLB.
Index Term
-Fixed point iteration (FPI), semidefinite programming (SDP), single moving receiver, target localization, time-of-arrival (TOA).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336384/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336384/</guid>
</item>
<item>
<title>The Neural-SRP Method for Universal Robust Multi-Source Tracking</title>
<description><p><span><big>The Neural-SRP Method for Universal Robust Multi-Source Tracking</big></span><br></p><p><span><small><i>Eric Grinstein; Christopher M. Hicks; Toon van Waterschoot; Mike Brookes; Patrick A. Naylor</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340057</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Neural networks have achieved state-of-the-art performance on the task of acoustic Direction-of-Arrival (DOA) estimation using microphone arrays. Neural models can be classified as end-to-end or hybrid, each class showing advantages and disadvantages. This work introduces Neural-SRP, an end-to-end neural network architecture for DOA estimation inspired by the classical Steered Response Power (SRP) method, which overcomes limitations of current neural models. We evaluate the architecture on multiple scenarios, namely, multi-source DOA tracking and single-source DOA tracking under the presence of directional and diffuse noise. The experiments demonstrate that our proposed method compares favourably in terms of computational and localization performance with established neural methods on various recorded and simulated benchmark datasets.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345765/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345765/</guid>
</item>
<item>
<title>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</title>
<description><p><span><big>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</big></span><br></p><p><span><small><i>Alexander Venus; Erik Leitinger; Stefan Tertinek; Klaus Witrisal</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper presents a neural-enhanced probabilistic model and corresponding factor graph-based sum-product algorithm for robust localization and tracking in multipath-prone environments. The introduced hybrid probabilistic model consists of physics-based and data-driven measurement models capturing the information contained in both, the line-of-sight (LOS) component as well as in multipath components (NLOS components). The physics-based and data-driven models are embedded in a joint Bayesian framework allowing to derive from first principles a factor graph-based algorithm that fuses the information of these models. The proposed algorithm uses radio signal measurements from multiple base stations to robustly estimate the mobile agent's position together with all model parameters. It provides high localization accuracy by exploiting the position-related information of the LOS component via the physics-based model and robustness by exploiting the geometric imprint of multipath components independent of the propagation channel via the data-driven model. In a challenging numerical experiment involving obstructed LOS situations to all anchors, we show that the proposed sequential algorithm significantly outperforms state-of-the-art methods and attains the posterior Cramér-Rao lower bound even with training data limited to local regions.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336409/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336409/</guid>
</item>
<item>
<title>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</title>
<description><p><span><big>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</big></span><br></p><p><span><small><i>AMIT EFRAIM; Joseph M. Francos</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In solving a model fitting problem, the existence of outliers in the set of measurements can have a devastating effect on the solution accuracy. Traditionally, in order to overcome this problem, robust point cloud registration algorithms are composed of transformation hypothesis generation, followed by hypothesis evaluation aimed at selecting the best hypothesized estimate. Hypotheses evaluation is commonly performed using the sample consensus criterion. However, since this criterion accounts only for the consensus size, it fails when the maximal sample consensus is incorrect. We propose a new hypothesis evaluation approach, generalizing the sample consensus approach, where instead of the sample consensus, the transformation that maximizes the point clouds feature correlation is selected as the best hypothesis. The feature vector at each point contains information such as on local geometry and semantic context. Utilizing this information in the hypotheses evaluation and selection procedure allows for a correct decision even when the hypothesis yielding the maximal sample consensus is false. Consequently, the probability of selecting the correct model increases. We show both mathematically and empirically that substituting the sample consensus criterion with the proposed point cloud feature correlation hypothesis test (PC-FCHT) lowers the probability of large registration errors, compared to using the special case of sample consensus. The proposed PC-FCHT is applicable to any algorithm that follows the hypothesis generation and evaluation scheme, potentially improving the success rates of a wide family of point cloud registration algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345750/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345750/</guid>
</item>
<item>
<title>Joint PAPR and OBP Reduction for NC-OFDM Systems</title>
<description><p><span><big>Joint PAPR and OBP Reduction for NC-OFDM Systems</big></span><br></p><p><span><small><i>Hsuan-Fu Wang; Fang-Biau Ueng; Bo-Heng Yeh</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3329757</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The spectrum resource is always a critical issue for wireless communications since it directly impacts the data rate and capacity. However, the problem of spectrum resource scarcity always exists. Moreover, spectrum resource scarcity becomes more severe as new communication technologies and wireless applications sprout. Noncontiguous orthogonal frequency division multiplexing (NC-OFDM) is a multicarrier method for bandwidth utilization. Unfortunately, this system has two fatal defects: high peak-to-average power ratio (PAPR) and considerable out-of-band power (OBP), which are detrimental to the system's performance. To solve these two problems, we propose a convex optimization-based method for joint PAPR and OBP reduction in NC-OFDM Systems. The strategy is to permit the secondary user to utilize the unoccupied spectrum of the primary user with dynamic spectrum sharing (DSS) based on a cognitive radio network (CRN). To this end, a flexible system operating over noncontiguous bands and DSS scenarios is necessary. The simulation results have shown that our method could effectively improve the overall performance and outperform other schemes, i.e., projections onto convex sets (POCS) and alternating projections onto convex and non-convex sets (APOCNCS), without harming the transmission of the primary system. The collaboration between secondary and primary systems is viable with the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10305259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10305259/</guid>
</item>
<item>
<title>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</title>
<description><p><span><big>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</big></span><br></p><p><span><small><i>Gal Itzhak; Israel Cohen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339433</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article presents a Kronecker-product (KP) beamforming approach incorporating sparse concentric circular arrays (SCCAs). The locations of the microphones on the SCCA are optimized concerning the broadband array directivity over a wide range of direction-of-arrival (DOA) deviations of a desired signal. A maximum directivity factor (MDF) sub-beamformer is derived accordingly with the optimal locations. Then, we propose two global beamformers obtained as a Kronecker product of a uniform linear array (ULA) and the SCCA sub-beamformer. The global beamformers differ by the type of the ULA, which is designed either as an MDF sub-beamformer along the
$\mathsf {x}$
-axis or as a maximum white noise gain sub-beamformer along the
$\mathsf {y}$
-axis. We analyze the performance of the proposed beamformers in terms of the directivity factor, the white noise gain, and their spatial beampatterns. Compared to traditional beamformers, the proposed beamformers exhibit considerably larger tolerance to DOA deviations concerning both the azimuth and elevation angles. Experimental results with speech signals in noisy and reverberant environments demonstrate that the proposed approach outperforms traditional beamformers regarding the perceptual evaluation of speech quality (PESQ) and short-time objective intelligibility (STOI) scores when the desired speech signals deviate from the nominal DOA.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342869/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342869/</guid>
</item>
<item>
<title>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</title>
<description><p><span><big>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</big></span><br></p><p><span><small><i>Karn N. Watcharasupat; Chih-Wei Wu; Yiwei Ding; Iroro Orife; Aaron J. Hipple; Phillip A. Williams; Scott Kramer; Alexander Lerch; William Wolcott</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339428</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Cinematic audio source separation is a relatively new subtask of audio source separation, with the aim of extracting the dialogue, music, and effects stems from their mixture. In this work, we developed a model generalizing the Bandsplit RNN for any complete or overcomplete partitions of the frequency axis. Psychoacoustically motivated frequency scales were used to inform the band definitions which are now defined with redundancy for more reliable feature extraction. A loss function motivated by the signal-to-noise ratio and the sparsity-promoting property of the 1-norm was proposed. We additionally exploit the information-sharing property of a common-encoder setup to reduce computational complexity during both training and inference, improve separation performance for hard-to-generalize classes of sounds, and allow flexibility during inference time with detachable decoders. Our best model sets the state of the art on the Divide and Remaster dataset with performance above the ideal ratio mask for the dialogue stem.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342812/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342812/</guid>
</item>
<item>
<title>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</title>
<description><p><span><big>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</big></span><br></p><p><span><small><i>Miguel Ferrer; María de Diego; Alberto Gonzalez</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340106</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The LMS algorithm is widely employed in adaptive systems due to its robustness, simplicity, and reasonable performance. However, it is well known that this algorithm suffers from a slow convergence speed when dealing with colored reference signals. Numerous variants and alternative algorithms have been proposed to address this issue, though all of them entail an increase in computational cost. Among the proposed alternatives, the affine projection algorithm stands out. This algorithm has the peculiarity of starting from
$N$
data vectors of the reference signal. It transforms these vectors into as many data vectors suitably normalized in energy and mutually orthogonal. In this work, we propose a version of the LMS algorithm that, similar to the affine projection algorithm, starts from
$N$
data vectors of the reference signal but corrects them by using only a scalar factor that functions as a convergence step. Our goal is to align the behavior of this algorithm with the behavior of the affine projection algorithm without significantly increasing the computational cost of the LMS.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345730/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345730/</guid>
</item>
<item>
<title>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</title>
<description><p><span><big>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</big></span><br></p><p><span><small><i>Zongyao Li; Ren Togo; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340616</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article proposes a method for transferring knowledge of semantic segmentation from a labeled source domain to an unlabeled target domain without using the source-domain data. Such a problem is called source-data-free domain adaptation, in which a pre-trained source-domain model and the unlabeled target-domain data are used to transfer the label knowledge across the domains. Like most previous methods, our method uses pseudo labels for distilling and transferring the source-domain knowledge. On the basis of the pseudo-label learning, our method improves the domain adaptation performance in two innovative ways: 1) reducing the domain differences by source-data-free style transfer and 2) exploring the style diversity within the target domain by style modification. To this end, we introduce two additional modules: 1) an inter-domain style transfer module which aligns the feature statistics of the source and target domains before producing the pseudo labels thereby improving the pseudo labels' accuracy, and 2) an intra-domain style modification module which modifies the image styles within the target domain for learning intra-domain style-invariant features. Our method with the two modules outperforms previous source-data-free domain adaptation methods in two commonly used benchmarks. Moreover, our method is well compatible with the previous methods for further improvement.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10356748/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10356748/</guid>
</item>
<item>
<title>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</title>
<description><p><span><big>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</big></span><br></p><p><span><small><i>Tobias Kabzinski; Peter Jax</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337721</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Quasi-continuous system identification of time-variant linear acoustic systems can be applied in various audio signal processing applications when numerous acoustic transfer functions must be measured. A prominent application is measuring head-related transfer functions. We treat the underlying multiple-input-multiple-output (MIMO) system identification problem in a state-space model as a joint estimation problem for states, representing impulse responses, and state-space model parameters using the expectation maximization (EM) algorithm. We address limitations of prior work by imposing different model structures, especially for dependencies within a (transformed) state vector. This results in block diagonal matrix structures, for which we derive M-step update rules. Making assumptions about this model structure and choosing a block size for a given application define the computational complexity. In examples, we found that applying this framework yields improvements of up to 10 dB in relative system distance in comparison to a conventional method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334061/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334061/</guid>
</item>
<item>
<title>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</title>
<description><p><span><big>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</big></span><br></p><p><span><small><i>Damir Rakhimov; Martin Haardt</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337729</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this paper, we present an analytical performance assessment of 2-D Tensor ESPRIT in terms of physical parameters. We show that the error in the
$r$
-mode depends only on two components, irrespective of the dimensionality of the problem. We obtain analytical expressions in closed form for the mean squared error (MSE) in each dimension as a function of the signal-to-noise (SNR) ratio, the array steering matrices, the number of antennas, the number of snapshots, the selection matrices, and the signal correlation. The derived expressions allow a better understanding of the difference in performance between the tensor and the matrix versions of the ESPRIT algorithm. The simulation results confirm the coincidence between the presented analytical expression and the curves obtained via Monte Carlo trials. We analyze the behavior of each of the two error components in different scenarios.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334446/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334446/</guid>
</item>
<item>
<title>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</title>
<description><p><span><big>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</big></span><br></p><p><span><small><i>Yaman Kındap; Simon Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343341</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this work we study linear vector stochastic differential equation (SDE) models driven by the generalised hyperbolic (GH) Lévy process for inference in continuous-time non-Gaussian filtering problems. The GH family of stochastic processes offers a flexible framework for modelling of non-Gaussian, heavy-tailed characteristics and includes the normal inverse-Gaussian, variance-gamma and Student-t processes as special cases. We present continuous-time simulation methods for the solution of vector SDE models driven by GH processes and novel inference methodologies using a variant of sequential Markov chain Monte Carlo (MCMC). As an example a particular formulation of Langevin dynamics is studied within this framework. The model is applied to both a synthetically generated data set and a real-world financial series to demonstrate its capabilities.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360268/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360268/</guid>
</item>
<item>
<title>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</title>
<description><p><span><big>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</big></span><br></p><p><span><small><i>James M. Cozens; Simon J. Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344048</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper proposes a probabilistic approach for extracting time-varying and irregular time signature information from polyphonic audio extracts, subsequently providing beat and bar line positions given inferred time signature divisions. This is achieved via dynamically evaluating the beat tempo as a function of time through finding an optimal compromise in beat and bar alignment in the time and tempo domains. Time signature divisions are determined based on a new representation, termed the Metrogram, that presents time-varying information regarding rhythmic and metric periodicities in the Tempogram. Our methodology is characterised by its ability to provide a distribution over metric interpretations, offering insights into the diverse ways music can be rhythmically perceived. Results indicate high-level accuracy for a variety of polyphonic extracts containing irregular, complex, irrational, and time-varying time signatures. Accuracy rivalling state-of-the-art methodologies is also reported in a beat tracking task performed on the standard Ballroom Dataset. The paper offers insights into the field of dynamic time signature recognition and beat tracking, offering a valuable and versatile resource for the analysis, composition, and performance of music.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363392/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363392/</guid>
</item>
<item>
<title>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</title>
<description><p><span><big>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</big></span><br></p><p><span><small><i>Yuto Watanabe; Ren Togo; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343335</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Although text-guided image manipulation approaches have demonstrated highly accurate performance for editing the appearance of images in a virtual or simple scenario, their real-world applications face significant challenges. The primary cause of these challenges is the misalignment in the distribution of training and real-world data, which leads to unstable text-guided image manipulation. In this work, we propose a novel framework called TolerantGAN and tackle the new task of real-world text-guided image manipulation independent of the training data. To achieve this, we introduce two key concepts of a border smoothly connection module (BSCM) and a manipulation direction-based attention module (MDAM). BSCM smoothens the misalignment in the distribution of training and real-world data. MDAM extracts only regions highly relevant for image manipulation and assists in reconstructing unobserved objects in the training data. For in-the-wild input images of various classes, TolerantGAN robustly outperforms the state-of-the-art methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360283/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360283/</guid>
</item>
<item>
<title>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</title>
<description><p><span><big>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</big></span><br></p><p><span><small><i>Anastasia Avdeeva; Aleksei Gusev; Tseren Andzhukaev; Artem Ivanov</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343342</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Whispered speech is a quiet voice without vocalization. One of the common cases of using whispered speech is a technique that can help overcome stuttering. But whispered speech can be uncomfortable and difficult to understand in everyday communication. To address these problems, we propose a method of low-delayed whisper-to-speech voice conversion, which can be useful in real life communication of people with disordered speech. As part of our research, we study the impact of streaming Automatic Speech Recognition models on the quality of voice conversion, comparing different streaming models and methods for model adaptation to streaming settings, and showing the importance of using such models in cases of low-delayed voice conversion.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360259/</guid>
</item>
<item>
<title>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</title>
<description><p><span><big>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</big></span><br></p><p><span><small><i>Denis C. Ilie-Ablachim; Andra Băltoiu; Bogdan Dumitrescu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344313</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We propose sparse representation and dictionary learning algorithms for dictionaries whose atoms are characterized by Gaussian probability distributions around some central atoms. This extension of the space covered by the atoms permits a better characterization of localized features of the signals. The representations are computed by trading-off representation error and the probability of the actual atoms used in the representation. We propose two types of algorithms: a greedy one, similar in spirit with Orthogonal Matching Pursuit, and one based on L1-regularization. We apply our new algorithms to unsupervised anomaly detection, where the representation error is used as anomaly score. The flexibility of our approach appears to improve more the representations of the many normal signals than those of the few outliers, at least for an anomaly type called dependency, thus improving the detection quality. Comparison with both standard dictionary learning algorithms and established anomaly detection methods is favorable.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365333/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365333/</guid>
</item>
<item>
<title>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</title>
<description><p><span><big>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</big></span><br></p><p><span><small><i>Yuya Moroto; Yingrui Ye; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344079</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>There are various sentiment theories for categorizing human sentiments into several discrete sentiment categories, which means that the theory used for training sentiment prediction methods does not always match that used in the test phase. As a solution to this problem, zero-shot visual sentiment prediction methods have been proposed to predict unseen sentiments for which no images are available in the training phase. However, the training of these previous zero-shot methods relies on a single sentiment theory, which limits their ability to handle sentiments from other theories. Thus, this article proposes a more robust zero-shot visual sentiment prediction method that can handle cross-domain sentiments defined in different sentiment theories. Specifically, by focusing on the fact that sentiments are abstract concepts common to humans regardless of whether their theories are different, we incorporate knowledge distillation into our method to construct a teacher–student model that can train the implicit relationships between sentiments defined in different sentiment theories. Furthermore, to enhance sentiment discrimination capability and strengthen the implicit relationships between sentiments, we introduce a novel sentiment loss between the teacher and student models. In this way, our model becomes robust to unseen sentiments by exploiting the implicit relationships between sentiments. The contributions of this article are the introduction of knowledge distillation and a novel sentiment loss between the teacher and student models for zero-shot visual sentiment prediction, and improved performance of zero-shot visual sentiment prediction. Experiments on several open datasets demonstrate the effectiveness of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363382/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363382/</guid>
</item>
<item>
<title>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</title>
<description><p><span><big>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</big></span><br></p><p><span><small><i>Chengen Liu; Geert Leus; Elvin Isufi</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339376</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The edge flow reconstruction task consists of retreiving edge flow signals from corrupted or incomplete measurements. This is typically solved by a regularized optimization problem on higher-order networks such as simplicial complexes and the corresponding regularizers are chosen based on prior knowledge. Tailoring this prior to the setting of interest can be challenging or it may not even be possible. Thus, we consider to learn this prior knowledge via a model-based deep learning approach. We propose a new regularized optimization problem for the simplicial edge flow reconstruction task, the simplicial ElasticNet, which combines the advantages of the
$\ell _{1}$
and
$\ell _{2}$
norms. We solve the simplicial ElasticNet problem via the multi-block alternating direction method of multipliers (ADMM) algorithm and provide conditions on its convergence. By unrolling the ADMM iterative steps, we develop a model-based neural network with a low requirement on the number of training data. This unrolling network replaces the fixed parameters in the iterative algorithm by learnable weights, thus exploiting the neural network's learning capability while preserving the iterative algorithm's interpretability. We enhance this unrolling network via simplicial convolutional filters to aggregate information from the edge flow neighbors, ultimately, improving the network learning expressivity. Extensive experiments on real-world and synthetic datasets validate the proposed approaches and show considerable improvements over both baselines and traditional non-model-based neural networks.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342735/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342735/</guid>
</item>
<item>
<title>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</title>
<description><p><span><big>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</big></span><br></p><p><span><small><i>Jim Beckers; Bart Van Erp; Ziyue Zhao; Kirill Kondrashov; Bert De Vries</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337718</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Bayesian model reduction provides an efficient approach for comparing the performance of all nested sub-models of a model, without re-evaluating any of these sub-models. Until now, Bayesian model reduction has been applied mainly in the computational neuroscience community on simple models. In this paper, we formulate and apply Bayesian model reduction to perform principled pruning of Bayesian neural networks, based on variational free energy minimization. Direct application of Bayesian model reduction, however, gives rise to approximation errors. Therefore, a novel iterative pruning algorithm is presented to alleviate the problems arising with naive Bayesian model reduction, as supported experimentally on the publicly available UCI datasets for different inference algorithms. This novel parameter pruning scheme solves the shortcomings of current state-of-the-art pruning methods that are used by the signal processing community. The proposed approach has a clear stopping criterion and minimizes the same objective that is used during training. Next to these benefits, our experiments indicate better model performance in comparison to state-of-the-art pruning schemes.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334001/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334001/</guid>
</item>
<item>
<title>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</title>
<description><p><span><big>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</big></span><br></p><p><span><small><i>Anh Minh Truong; Wilfried Philips; Peter Veelaert</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340064</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Many image-based rendering (IBR) methods rely on depth estimates obtained from structured light or time-of-flight depth sensors to synthesize novel views from sparse camera networks. However, these estimates often contain missing or noisy regions, resulting in an incorrect mapping between source and target views. This situation makes the fusion process more challenging, as the visual information is misaligned, inconsistent, or missing. In this work, we first implement a lightweight network based on the transformer, which is well-known for its capability to model long-range relationships within input data, to extract spatial features from color images. These features are then used to enhance the quality of completed depth maps. Furthermore, we combine a sequential deep neural network with a spatial attention mechanism to effectively fuse the projected features from multiple source viewpoints. This approach enables us to integrate information from an arbitrary number of source viewpoints as well as improve accuracy in synthesized views. Experimental results on challenging datasets demonstrate that our method achieves superior synthesized image quality compared to state-of-the-art (SOTA) methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345792/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345792/</guid>
</item>
<item>
<title>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</title>
<description><p><span><big>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</big></span><br></p><p><span><small><i>Reza Mirzaeifard; Naveen K. D. Venkategowda; Vinay Chakravarthi Gogineni; Stefan Werner</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344395</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper investigates quantile regression in the presence of non-convex and non-smooth sparse penalties, such as the minimax concave penalty (MCP) and smoothly clipped absolute deviation (SCAD). The non-smooth and non-convex nature of these problems often leads to convergence difficulties for many algorithms. While iterative techniques such as coordinate descent and local linear approximation can facilitate convergence, the process is often slow. This sluggish pace is primarily due to the need to run these approximation techniques until full convergence at each step, a requirement we term as a
secondary convergence iteration
. To accelerate the convergence speed, we employ the alternating direction method of multipliers (ADMM) and introduce a novel single-loop smoothing ADMM algorithm with an increasing penalty parameter, named SIAD, specifically tailored for sparse-penalized quantile regression. We first delve into the convergence properties of the proposed SIAD algorithm and establish the necessary conditions for convergence. Theoretically, we confirm a convergence rate of
$o({k^{-\frac{1}{4}}})$
for the sub-gradient bound of the augmented Lagrangian, where
$k$
denotes the number of iterations. Subsequently, we provide numerical results to showcase the effectiveness of the SIAD algorithm. Our findings highlight that the SIAD method outperforms existing approaches, providing a faster and more stable solution for sparse-penalized quantile regression.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365338/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365338/</guid>
</item>
<item>
<title>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</title>
<description><p><span><big>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</big></span><br></p><p><span><small><i>Oliver Lang; Christian Hofbauer; Reinhard Feger; Mario Huemer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343308</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>A promising waveform candidate for future joint sensing and communication systems is orthogonal frequency-division multiplexing (OFDM). For such systems, supporting multiple transmit antennas requires multiplexing methods for the generation of orthogonal transmit signals, where equidistant subcarrier interleaving (ESI) is the most popular multiplexing method. In this work, we analyze a multiplexing method called Doppler-division multiplexing (DDM). This method applies a phase shift from OFDM symbol to OFDM symbol to separate signals transmitted by different Tx antennas along the velocity axis of the range-Doppler map. The main focus of this work lies on the implications of DDM on the communication task. It will be shown that for DDM, the channels observed in the communication receiver are heavily time-varying, preventing any meaningful transmission of data when not taken into account. In this work, a communication system designed to combat these time-varying channels is proposed, which includes methods for data estimation, synchronization, and channel estimation. Bit error ratio (BER) simulations demonstrate the superiority of this communications system compared to ESI-based systems.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360223/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360223/</guid>
</item>
<item>
<title>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</title>
<description><p><span><big>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</big></span><br></p><p><span><small><i>Nils L. Westhausen; Bernd T. Meyer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343320</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we introduce a causal low-latency low-complexity approach for binaural multichannel blind speaker separation in noisy reverberant conditions. The model, referred to as Group Communication Binaural Filter and Sum Network (GCBFSnet) predicts complex filters for filter-and-sum beamforming in the time-frequency domain. We apply Group Communication (GC), i.e., latent model variables are split into groups and processed with a shared sequence model with the aim of reducing the complexity of a simple model only containing one convolutional and one recurrent module. With GC we are able to reduce the size of the model by up to 83% and the complexity up to 73% compared to the model without GC, while mostly retaining performance. Even for the smallest model configuration, GCBFSnet matches the performance of a low-complexity TasNet baseline in most metrics despite the larger size and higher number of required operations of the baseline.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360275/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360275/</guid>
</item>
<item>
<title>Reverse Ordering Techniques for Attention-Based Channel Prediction</title>
<description><p><span><big>Reverse Ordering Techniques for Attention-Based Channel Prediction</big></span><br></p><p><span><small><i>Valentina Rizzello; Benedikt Böck; Michael Joham; Wolfgang Utschick</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344024</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Channel state information (CSI) is crucial for enhancing the performance of wireless systems by allowing to adjust the transmission strategies based on the current channel conditions. However, obtaining precise CSI is difficult because of the fast-changing channel conditions caused by multi-path fading. An inaccurate CSI hinders the performance of various adaptive wireless systems, highlighting the need for channel prediction techniques to effectively mitigate the drawbacks of outdated CSI. Conventional methods typically depend on assumptions regarding user velocity or require knowledge of the Doppler frequency. In contrast to existing approaches, we aim for a more robust and practical solution by training neural networks without making any assumptions about user velocity, relying solely on noisy channel observations during training. Specifically, we adapt both the sequence-to-sequence with attention (Seq2Seq-attn) and transformer models for channel prediction. Additionally, a new technique called reverse positional encoding is introduced in the transformer model to improve the robustness of the model against varying sequence lengths. Similarly, the encoder outputs of the Seq2Seq-attn model are reversed prior to the application of attention mechanisms. By means of simulations, we show that these proposed techniques enable the models to effectively capture relationships within sequences of channel snapshots without increasing the complexity. Importantly, this capability remains robust across varying sequence lengths, representing a substantial improvement over existing methodologies.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363354/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363354/</guid>
</item>
<item>
<title>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</title>
<description><p><span><big>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</big></span><br></p><p><span><small><i>Sarina Meyer; Xiaoxiao Miao; Ngoc Thang Vu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344375</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Speaker anonymization is the task of modifying a speech recording such that the original speaker cannot be identified anymore. Since the first Voice Privacy Challenge in 2020, along with the release of a framework, the popularity of this research topic is continually increasing. However, the comparison and combination of different anonymization approaches remains challenging due to the complexity of evaluation and the absence of user-friendly research frameworks. We therefore propose an efficient speaker anonymization and evaluation framework based on a modular and easily extendable structure, almost fully in Python. The framework facilitates the orchestration of several anonymization approaches in parallel and allows for interfacing between different techniques. Furthermore, we propose modifications to common evaluation methods which improves the quality of the evaluation and reduces their computation time by 65 to 95%, depending on the metric. Our code is fully open source.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365329/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365329/</guid>
</item>
<item>
<title>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</title>
<description><p><span><big>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</big></span><br></p><p><span><small><i>Alessandro Ilic Mezza; Matteo Amerena; Alberto Bernardini; Augusto Sarti</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343318</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Real-time audio communications over IP have become essential to our daily lives. Packet-switched networks, however, are inherently prone to jitter and data losses, thus creating a strong need for effective packet loss concealment (PLC) techniques. Though solutions based on deep learning have made significant progress in that direction as far as speech is concerned, extending the use of such methods to applications of Networked Music Performance (NMP) presents significant challenges, including high fidelity requirements, higher sampling rates, and stringent temporal constraints associated to the simultaneous interaction between remote musicians. In this article, we present PARCnet, a hybrid PLC method that utilizes a feed-forward neural network to estimate the time-domain residual signal of a parallel linear autoregressive model. Objective metrics and a listening test show that PARCnet provides state-of-the-art results while enabling real-time operation on CPU.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360264/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360264/</guid>
</item>
<item>
<title>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</title>
<description><p><span><big>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</big></span><br></p><p><span><small><i>Katerina Zmolikova; Michael Syskind Pedersen; Jesper Jensen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Supervised learning-based speech enhancement methods often work remarkably well in acoustic situations represented in the training corpus but generalize poorly to out-of-domain situations, i.e. situations not seen during training. This stands in the way of further improvement of these methods in realistic scenarios, as collecting paired noisy-clean recordings in the target application domain is typically not feasible. Recording noisy-only in-domain data is, though, much more practical. In this article, we tackle the problem of unsupervised domain adaptation in speech enhancement. Specifically, we propose a way to use in-domain noisy-only data in the training of a neural network to improve upon a model trained solely on out-of-domain paired data. For this, we make use of masked spectrogram prediction, a technique from self-supervised learning that aims to interpolate masked regions of a spectrogram. We hypothesize that masked spectrogram prediction encourages learning of features that represent well both speech and noise components of the noisy signals. These features can then be used to train a more robust speech enhancement system. We evaluate the proposed method on the VoiceBank-DEMAND and LibriFSD50k databases, with WSJ0-CHiME3 serving as the out-of-domain database. We show that the proposed method outperforms both the out-of-domain system and the baseline approaches, i.e. RemixIT and noisy-target training, and also combines well with the previously proposed RemixIT method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360251/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360251/</guid>
</item>
<item>
<title>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</title>
<description><p><span><big>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</big></span><br></p><p><span><small><i>Alexander Karataev; Christian Forsch; Laura Cottatellucci</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3348343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We consider a cell-free massive multiple-input multiple-output (CF-MaMIMO) communication system in the uplink transmission and propose a novel algorithm for blind or semi-blind joint channel estimation and data detection (JCD). We formulate the problem in the framework of bilinear inference and develop a solution based on the expectation propagation (EP) method for both channel estimation and data detection. We propose a new approximation of the joint a posteriori distribution of the channel and data whose representation as a factor graph enables the application of the EP approach using the message-passing technique, local low-complexity computations at the nodes, and an effective modeling of channel-data interplay. The derived algorithm, called bilinear-EP JCD, allows for a distributed implementation among access points (APs) and the central processing unit (CPU) and has polynomial complexity. Our simulation results show that it outperforms other EP-based state-of-the-art polynomial time algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10378663/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10378663/</guid>
</item>
<item>
<title>Adversarial Representation Learning for Robust Privacy Preservation in Audio</title>
<description><p><span><big>Adversarial Representation Learning for Robust Privacy Preservation in Audio</big></span><br></p><p><span><small><i>Shayan Gharib; Minh Tran; Diep Luong; Konstantinos Drossos; Tuomas Virtanen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Sound event detection systems are widely used in various applications such as surveillance and environmental monitoring where data is automatically collected, processed, and sent to a cloud for sound recognition. However, this process may inadvertently reveal sensitive information about users or their surroundings, hence raising privacy concerns. In this study, we propose a novel adversarial training method for learning representations of audio recordings that effectively prevents the detection of speech activity from the latent features of the recordings. The proposed method trains a model to generate invariant latent representations of speech-containing audio recordings that cannot be distinguished from non-speech recordings by a speech classifier. The novelty of our work is in the optimization algorithm, where the speech classifier's weights are regularly replaced with the weights of classifiers trained in a supervised manner. This increases the discrimination power of the speech classifier constantly during the adversarial training, motivating the model to generate latent representations in which speech is not distinguishable, even using new speech classifiers trained outside the adversarial training loop. The proposed method is evaluated against a baseline approach with no privacy measures and a prior adversarial training method, demonstrating a significant reduction in privacy violations compared to the baseline approach. Additionally, we show that the prior adversarial method is practically ineffective for this purpose.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10379095/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10379095/</guid>
</item>
<item>
<title>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</title>
<description><p><span><big>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</big></span><br></p><p><span><small><i>Yasaman Parhizkar; Gene Cheung; Andrew W. Eckford</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>It is a popular hypothesis in neuroscience that ganglion cells in the retina are activated by selectively detecting visual features in an observed scene. While ganglion cell firings can be predicted via data-trained deep neural nets, the n... |
http://localhost:1200/ieee/journal/4427201/earlyaccess - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Sensors Journal</title>
<link>https://ieeexplore.ieee.org/xpl/tocresult.jsp?isnumber=4427201</link>
<atom:link href="http://localhost:1200/ieee/journal/4427201/earlyaccess" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Sensors Journal - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Wed, 10 Apr 2024 12:28:55 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Torsion Sensor Based on Helical Long-Period Grating Inscribed in the Etched Double Cladding Fiber</title>
<description><p><span><big>Torsion Sensor Based on Helical Long-Period Grating Inscribed in the Etched Double Cladding Fiber</big></span><br></p><p><span><small><i>Yanping He; Yuehui Ma; Chen Jiang; Yunqi Liu</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373623</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>We proposed a high-sensitivity torsion sensor based on the helical long-period grating (HLPG) inscribed in the cladding-etched double cladding fiber (DCF). The torsion sensitivity has been greatly improved by reducing the cladding diameter of the DCF, from 0.342 nm/(rad/m) to 1.162 nm/(rad/m), which is one order of magnitude higher than that of the conventional long-period fiber grating. The cladding mode coupling characteristics during the etching process are also investigated experimentally. The DCF-HLPG has a low strain and temperature sensitivity, and the maximum strain sensitivity is only about 0.0022 nm/με. The proposed DCF-HLPG has a potential application as a high-sensitivity torsion sensor with low strain and temperature crosstalk.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466518/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466518/</guid>
</item>
<item>
<title>Deep-Learning-Based Prediction Algorithm for Fuel Cell Electric Vehicle Energy with Shift Mixup</title>
<description><p><span><big>Deep-Learning-Based Prediction Algorithm for Fuel Cell Electric Vehicle Energy with Shift Mixup</big></span><br></p><p><span><small><i>Tae-Ho Kim; Jae-Heung Cho; Young-Kwang Kim; Joon-Hyuk Chang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373078</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The automobile industry is switching from fossil-fuel-based energy sources to green energy sources. In particular, because hydrogen has a high energy density and efficiency compared to other energy sources, it is a major research field for commercial vehicles that require large amounts of energy and travel long distances on average, such as buses and trucks. In fuel cell electric vehicles (FCEVs), maintaining an adequate energy production intensity is necessary to improve energy production efficiency and prevent falling into a state of inoperability owing to a lack of energy. In this case, energy prediction becomes a significant factor. In this study, a deep-learning-based prediction method for FCEV powertrain energy is proposed. The proposed method uses only internal data, which can be obtained from the vehicle, and does not require external information regarding future routes. Additionally, we designed a model considering the vehicle data time-series characteristics, and proposed a shift mixup, which is a data augmentation method that does not compromise the vehicle’s dynamic characteristics, to address the data shortage problem. Furthermore, a pretext task-learning method that can improve model performance without external data during inference is introduced. This method includes pretext tasks designed specifically for the vehicle domain. Finally, a distance-based loss mask and contrastive learning that the representation can learn semantic information are proposed. Experimental results for the actual driving dataset show that our method improved by 28.05% and by 30.04% compared to the exponential moving average in terms of root mean squared error and mean absolute error, respectively. We demonstrate with energy management strategy (EMS) that an effective energy prediction algorithm helps sustain an optimal state of charge (SOC).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466522/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466522/</guid>
</item>
<item>
<title>Effect of substrate temperature on Cu2-xO growth for enhanced photodetector performance in visible region</title>
<description><p><span><big>Effect of substrate temperature on Cu2-xO growth for enhanced photodetector performance in visible region</big></span><br></p><p><span><small><i>Karthickraja Ramakrishnan; Y. Ashok Kumar Reddy; B. Ajitha</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373768</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The present study investigates the effect of substrate temperature (30-300°C) on the sputter-deposited Cu
2-x
O films for improved photodetector performance in the visible region. Cubic structured Cu2O phase is observed for all deposited films through structural analysis. Increasing substrate temperature causes the formation of nanospheres with higher grain size at 200°C confirmed by the morphological images. Moreover, all the deposited thin films show a strong absorption band in the visible region (526-550 nm). Among deposited films, a lower bandgap (2.02 eV) is observed for the film deposited at 200°C due to the higher surface area-to-volume ratio and increased grain size. The better crystalline nature, more copper vacancy, and formation of nanospheres with a larger grain size of the film deposited at 200°C result in higher mobility and conductivity. Further, the photodetector performance of the Au/Cu
2-x
O/Au test-devices is studied under the visible light (λ=530 nm) irradiation. Among samples, the sample grown at 200°C showed higher photocurrent (36.10 μA), I
on
/I
off
ratio (183.50), photoresponsivity (1.127 A/W), specific detectivity (22.40×10
11
Jones), and faster response times (τ
r
=92 ms and τ
f
=84 ms) even at lower incident optical power density (0.1283 mW/cm
2
). These improved experimental results signify the potential of the fabricated test-device at 200°C for the optoelectronics devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10471292/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10471292/</guid>
</item>
<item>
<title>A Cavity-Type Pressure Sensor Array with High Anti-Disturbance Performance Inspired by Fish Lateral Line Canal</title>
<description><p><span><big>A Cavity-Type Pressure Sensor Array with High Anti-Disturbance Performance Inspired by Fish Lateral Line Canal</big></span><br></p><p><span><small><i>Qian Yang; Qiao Hu; Liuhao Shan; Guangyu Jiang; Yuanji Yao; Long Tang; Zicai Zhu; Alvo Aabloo</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3381162</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Effective sensing is challenging in real underwater environment swarming with various disturbances. However, most research on underwater sensors has been conducted in still water, which may not provide sufficient insights for environments with disturbances. In addressing this issue, our paper not only proposes the development of anti-disturbance sensor units but also delves into evaluating the localization capabilities of artificial lateral line arrays in disturbed water environments. Concretely, inspired by the lateral line canal system of fish, we proposed a cavity-type pressure sensing device (denoted as a C-sensor) with notable anti-disturbance performance. On one hand, a 3D-printed resin cavity is filled with silicone oil, with flexible latex film packaging the upper and lower openings to achieve anti-disturbance capabilities akin to those of fish lateral line canals. On the other hand, employing IPMC materials based on ion sensing principles, the sensor generates electrical signals similar to sensing cilia in fish lateral line systems. The results demonstrated that the cavityless sensor (denoted as N-sensor) showed self-oscillation behavior, leading to significant errors in weak target signal detection and even sensor failure in extreme cases. However, the C-sensor exhibits remarkable sensing performance in still water and water disturbances, thereby confirming the effectiveness of its cavity-type structure in mitigating the impact of water disturbances. Furthermore, we developed two distinct artificial lateral line arrays based on the above sensors (C-array and N-array). The experimental results showed a marked 78% reduction in mean localization error for C-array compared to N-array. Additonally, the C-array also exhibited good ability in multiple dipoles detection. Such results indicate the superior applicability of the C-array within real complex underwater environments, promoting the practical application of IPMC sensors.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10486206/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10486206/</guid>
</item>
<item>
<title>Multi-sensor multiple extended objects tracking based on the message passing</title>
<description><p><span><big>Multi-sensor multiple extended objects tracking based on the message passing</big></span><br></p><p><span><small><i>Yuansheng Li; Tao Shen; Lin Gao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3384560</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Multiple extended objects (EOs) tracking has attracted a lot of attention due to the fast development of high-resolution sensors. A remarkable feature of EO, compared to the traditional point target, is that an EO normally produces more than one measurement, resulting in challenges for finding the associations among objects and measurements. In this paper, we are interested in tracking an unknown number of EOs with multiple sensors by resorting to the message passing method. The existence probability and belief of each EO are explicitly estimated, where the belief is modeled by a mixture gamma Gaussian inverse Wishart (GGIW) distribution, so as to jointly estimate the measurement rate (MR), centroid state and extension. The marginal posterior of EOs is approximately by belief, which is obtained by running the loopy sum-product algorithm (LSPA) on a suitably devised factor graph. As a result, the computational load of the proposed algorithm increases linearly with respect to the number of targets, thus admitting the scalability. Simulation experiments are carried out to verify the performance of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10495766/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10495766/</guid>
</item>
<item>
<title>Dual-Resonance Split Ring Resonator Metasurface for Terahertz Biosensing</title>
<description><p><span><big>Dual-Resonance Split Ring Resonator Metasurface for Terahertz Biosensing</big></span><br></p><p><span><small><i>Arslan Asim; Michael Cada; Yuan Ma; Alan Fine</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3376290</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this paper, the ‘terahertz gap’ has been addressed by designing a novel THz metasurface for potential use in biosensing applications. The metasurface sensor employs Surface Plasmon Resonance (SPR). It operates in the 0 - 1 THz band. Two sharp reflection dips are provided by the sensor, which serve as indicators of analyte refractive index variations. Geometrical as well as compositional parameters of the biosensor design have been studied to optimize the performance in the targeted frequency band. The sensor design shows compatibility with different metals. The performance of the metasurface with gold, copper and aluminum has been investigated. The metasurface geometry is decently resilient to fabrication tolerances. The sensor maintains its resonance conditions when the angle of incidence is changed with minor deviations in the spectral response, but the polarization state of the incident terahertz beam clearly disturbs the absorption peak. Therefore, the sensing performance is restricted to a maximum allowable incidence angle of 20 degrees and circularly polarized terahertz beams. The resonance conditions for the metasurface appear around 0.4 and 0.7 THz. Both resonances have been investigated with respect to changes in the analyte refractive index. The chosen refractive index range is 1 to 1.5. The sensor response is calibrated by plotting the resonance frequency versus the refractive index. Least squares regression technique has been used to extract a data model for sensor response. Comparison of the proposed design with contemporary works has been incorporated into the paper. The sensor provides sensitivities of 0.1614 THz/ RIU and 0.23 THz/ RIU. The electromagnetic simulations have been carried out through the Finite Element Method (FEM).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10474307/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10474307/</guid>
</item>
<item>
<title>An Intestine based Bio-Compatible Humidity Sensor for Environmental and Medical Measurements</title>
<description><p><span><big>An Intestine based Bio-Compatible Humidity Sensor for Environmental and Medical Measurements</big></span><br></p><p><span><small><i>Emrehan Yavsan; Muhammet Rojhat Kara; Mehmet Akif Erismis</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3367032</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The bio-compatible devices suitable for recycling and bio-degrading can be achieved with organic materials in nature. In this work, a bio-compatible capacitive humidity sensor is presented for reducing the amount of electronic waste and contributing to the sustainability of natural resources and the future. The sensor consists of 3 layers. The first layer is the processed intestine layer of cattle. Bio-compatibility is achieved with this layer. In addition to being a highly absorbing tissue, the intestine has been used for centuries for long-term preservation of meat based food. Correspondingly, the developed sensor is found to be more durable and long-lasting than other natural-material based humidity sensors in the literature. The other layers of the sensor are interdigitated copper electrodes and a 0.2 mm thick thin film strip. Thin film strip increases mechanical strength as well as flexibility. The developed sensor prototype was subjected to various tests in the humidity range of 20%-90%. In these tests, the hysteresis characteristic of the sensor, its response-recovery time, and its long-term stability and short-term step responses were examined. Moreover, as a possible application in medicine, the sensor can be used to detect breathing cycles. The sensor’s response and recovery times were measured as 8.72’ and 4.47’, respectively, possibly attributed to the stabilization of our test setup, while the sensor successfully detected deep, normal and fast breathing. Despite being kept in an uncontrolled environment, the sensor continued to operate consistently for breath measurements after 56 weeks, which is more than a year.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10445282/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10445282/</guid>
</item>
<item>
<title>Deep neural network assisted terahertz metasurface sensors for the detection of lung cancer biomarkers</title>
<description><p><span><big>Deep neural network assisted terahertz metasurface sensors for the detection of lung cancer biomarkers</big></span><br></p><p><span><small><i>Jie Chen; Fangrong Hu; Xiaoya Ma; Mo Yang; Shangjun Lin; An Su</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3384578</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>MicroRNA (miRNA) is closely related to various cancers, and the change in its expression level is closely related to the development and death of tumor cells. Here, we design and manufacture a terahertz (THz) metasurface sensor to realize the concentration detection and category identification of the miRNAs related to lung cancer. We established two spectral classification datasets, one contains 9 concentrations of miRNAs, and the other contains 3 categories of lung cancer-related miRNAs. And then, we used a deep neural network (DNN) algorithm to classify these spectral datasets and obtained a LOD (Limit-of-Detection) of 100 aM for miRNAs. Moreover, this method can identify the categories of miRNAs. Compared with the other five machine learning (ML) algorithms, the proposed neural network framework achieves the best classification results. This work provides a new way for the detection and identification of trace nucleic acid and biomarkers of many cancers.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10495770/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10495770/</guid>
</item>
<item>
<title>A Linearized Temperature Measurement System Based on the B-mode of SC-cut Quartz Crystal</title>
<description><p><span><big>A Linearized Temperature Measurement System Based on the B-mode of SC-cut Quartz Crystal</big></span><br></p><p><span><small><i>Zhiqi Li; Haipeng Zhai; Miao Miao; Wei Zhou</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3374763</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This article demonstrates a linearized temperature measurement system based on the B-mode of SC-cut quartz crystal, which has high precision and good linearity. The proposed system uses a digital linear phase comparison measurement method by selecting phase detection region, which achieves high resolution, and fast response for temperature measurement. Extensive error analysis and performance evaluation of the system were conducted in this study. The developed linearized temperature measurement system demonstrated excellent performance, including: wide temperature measuring range (-10°C to 80°C); providing sufficient resolution and extremely fast response time with a resolution of 0.01°C at 10 μs, a resolution of 5E-5°C at 10 ms, and a resolution of 1E-7°C at 1 s when the temperature keeps stable. It achieves extremely low temperature measurement relative error (R.E.&lt; 0.24%) and nonlinearity (N.L.&lt; 0.18%). The measurement results show that the maximum estimate of the Allan variance is 2.3E-4°C. The proposed system has wide prospects for intelligent applications and can play a crucial role in scenarios that require high-precision and fast-response temperature measurements.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472879/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472879/</guid>
</item>
<item>
<title>Mesoporous silica modified by poly(ionic liquid)s for low humidity sensing</title>
<description><p><span><big>Mesoporous silica modified by poly(ionic liquid)s for low humidity sensing</big></span><br></p><p><span><small><i>Zhiyan Ma; Yaping Song; Hongran Zhao; Sen Liu; Xi Yang; Teng Fei; Tong Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373030</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this study, imidazole-based poly(ionic liquid)s are introduced into acid-treated mesoporous silica by solid grinding and subsequent thermal polymerization methods. The purpose of treating the mesoporous silica with hydrochloric acid is to increase the number of hydroxyl groups on the surface to fix more poly(ionic liquid)s. Due to the synergism of enhanced hydrophilicity and the mesoporous structure, the synthesized composite materials exhibited excellent capacities to adsorb and transport water vapor molecules and are suitable for low humidity sensing applications. In the 7-33% relative humidity (RH) range, the response of the as-prepared sensor reached 806% with a short response/recovery time of 8/13 s.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466490/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466490/</guid>
</item>
<item>
<title>PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards</title>
<description><p><span><big>PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards</big></span><br></p><p><span><small><i>Elias Arbash; Margret Fuchs; Behnood Rasti; Sandra Lorenz; Pedram Ghamisi; Richard Gloaguen</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3380826</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Addressing the critical theme of recycling electronic waste (E-waste), this contribution is dedicated to developing advanced automated data processing pipelines as a basis for decision-making and process control. Aligning with the broader goals of the circular economy and the United Nations (UN) Sustainable Development Goals (SDG), our work leverages non-invasive analysis methods utilizing RGB and hyperspectral imaging data to provide both quantitative and qualitative insights into the E-waste stream composition for optimizing recycling efficiency. In this paper, we introduce ’PCB-Vision’; a pioneering RGB-hyperspectral printed circuit board (PCB) benchmark dataset, comprising 53 RGB images of high spatial resolution paired with their corresponding high spectral resolution hyperspectral data cubes in the visible and near-infrared (VNIR) range. Grounded in open science principles, our dataset provides a comprehensive resource for researchers through high-quality ground truths, focusing on three primary PCB components: integrated circuits (IC), capacitors, and connectors. We provide extensive statistical investigations on the proposed dataset together with the performance of several state-of-the-art (SOTA) models, including U-Net, Attention UNet, Residual U-Net, LinkNet, and DeepLabv3+. By openly sharing this multiscene benchmark dataset along with the baseline codes, we hope to foster transparent, traceable, and comparable developments of advanced data processing across various scientific communities, including, but not limited to, computer vision and remote sensing. Emphasizing our commitment to supporting a collaborative and inclusive scientific community, all materials, including code, data, ground truth, and masks, will be accessible at https://github.com/hifexplo/PCBVision.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10485259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10485259/</guid>
</item>
<item>
<title>DEEP-CARDIO: Recommendation System for Cardiovascular Disease Prediction using IOT Network</title>
<description><p><span><big>DEEP-CARDIO: Recommendation System for Cardiovascular Disease Prediction using IOT Network</big></span><br></p><p><span><small><i>A. Yashudas; Dinesh Gupta; G. C. Prashant; Amit Dua; Dokhyl AlQahtani; A. Siva Krishna Reddy</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373429</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Internet of Things (IoT) based remote healthcare applications provide fast and preventative medical services to the patients at risk. However, predicting heart disease is a complex task and diagnosis results are rarely accurate. To address this issue, a novel Recommendation System for Cardiovascular Disease Prediction Using IoT Network (DEEP-CARDIO) has been proposed for providing prior diagnosis, treatment, and dietary recommendations for cardiac diseases. Initially, the physiological data are collected from the patient’s remotely by using the four bio sensors such as ECG sensor, Pressure sensor, Pulse sensor and Glucose sensor. An Arduino controller receives the collected data from the IoT sensors to predict and diagnose the disease. A cardiovascular disease prediction model is implemented by using BiGRU (Bidirectional-Gated Recurrent Unit) attention model which diagnose the cardiovascular disease and classify into five available cardiovascular classes. The recommendation system provides physical and dietary recommendations to cardiac patients based on the classified data, via user mobile application. The performance of the DEEP-CARDIO is validated by Cloud Simulator (CloudSim) using the real-time Framingham’s and Statlog heart disease dataset. The proposed DEEP CARDIO method achieves an overall accuracy of 99.90% whereas, the MABC-SVM, HCBDA and MLbPM method achieves 86.91%, 88.65% and 93.63% respectively.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472883/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472883/</guid>
</item>
<item>
<title>AuNP-decorated textile as chemo resistive sensor for acetone detection</title>
<description><p><span><big>AuNP-decorated textile as chemo resistive sensor for acetone detection</big></span><br></p><p><span><small><i>S. Casalinuovo; D. Caschera; S. Quaranta; D. Caputo</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2023.3348693</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This paper presents the development of chemo resistive sensors for the detection of volatile organic compounds (VOCs). The proposed sensor is based on citrate-functionalized gold nanoparticles (AuNPs) serving as a sensitive layer deposited on cotton fabric. Impedance variations due to VOC/substrate interaction are used as a detection principle. Specifically, this work focuses on acetone detection after exposing the AuNP-decorated cotton to a CH3COCH3 aqueous solution. Such an interaction resulted in a reduction of the total impedance (i.e., magnitude) of the system. This behavior can be ascribed to Van der Waals forces existing between the C=O group and the citrate moieties adsorbed on the gold nanoparticles, which favor charge injection to the substrate. Response to water was also tested for comparison, assuring that the solvent interacts with the sensitive layer by a different adsorption mechanism, not influencing the overall results. Sensor selectivity was also verified by considering ethanol (representative of alcohol group). Indeed, impedance curves reflect the different type of chemical interaction between the analyte and the substrate. In addition, sensor limit of detection for acetone was found to be 1% v/v, in the considered frequency range. Furthermore, sensor performance in terms of reusability was evaluated, showing that the Au-cotton ability in VOCs detection could be restored after about 90 min with a percentage up to 97 % in the frequency of 1Hz. These results can be considered the starting point for the development of portable, sensitive and user-friendly devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10384362/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10384362/</guid>
</item>
<item>
<title>L-cysteine modified gold nanoparticles for copper (II) ion detection using etched Fiber Bragg Grating sensor</title>
<description><p><span><big>L-cysteine modified gold nanoparticles for copper (II) ion detection using etched Fiber Bragg Grating sensor</big></span><br></p><p><span><small><i>S Srivatzen; B S Kavitha; Asha Prasad; S Asokan</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3374514</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>A highly sensitive and selective optical detection of copper (II) (Cu
2+
) concentration in water has been proposed using L-cysteine functionalized gold nanoparticles (AuNPs) using an etched Fiber Bragg grating sensor (eFBG). The eFBG sensor produces a Bragg wavelength shift (ΔλB) in correspondence to the added Cu
2+
concentration. The amino acid L-Cysteine facilitates the formation of core-satellite nanoassemblies of gold in the presence of Cu
2+
. A linear wavelength shift is observed in accordance with the mediating Cu
2+
concentration. The sensor’s detection limit is found to be 1picomolar (pM) which is well within the world health organization (WHO) acceptable limit for safe drinking water and the dynamic range of the sensor lies between 1pM to 100 μM. The sensor is highly sensitive and selective to Cu
2+
with sensitivity of 230 pm 10
-1
M. The sensor has proven its capability as an ideal probe for onsite, real-time measurement of Cu
2+
in drinking water by giving a recovery percentage of &gt;90% in real water sample additional to good reproducibility (standard error&lt; 20 pm), high sensitivity, portability and specificity.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472453/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472453/</guid>
</item>
<item>
<title>AcHand: Detecting Respiratory Rate of operator via Smartphone Microphone</title>
<description><p><span><big>AcHand: Detecting Respiratory Rate of operator via Smartphone Microphone</big></span><br></p><p><span><small><i>KounKou Vincent; Lin Wang; Nan Jing; Wenyuan Liu</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3372255</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this paper, we dedicate a handgrip-free technique, called ”AcHand”, to continuously detect respiratory rates under acoustic signals sensing via a smartphone. Although the particularity of the AcHand is to follow the subtle movements of the heart during its displacement, we acknowledge specific challenges that can make the signal vulnerable to handgripfree scenarios such as 1) stochastic hand activities: fluctuations in movement can alter the trajectory of the acoustic signal to the microphones and cause random changes in the breathing curve, resulting in complex and unpredictable dynamics, and 2) clicking on the phone screen can disrupt the breathing by generating a large signal variation to create undesirable peaks-to-noisy the signal. However, by analyzing the recorded signal reflection evoked by a chirp sound stimulus, we propose a linear transformation of the Cardioreflector-Manoreflector to analyze the respiratory patterns variation and distinguish weak breathing signals under the handgrip-free. We compare the present signal with the ground truth recorded by ECG to approve the efficiency of the proposed method. Our experiments show AcHand achieves a MAE of 0.341 bpm, which is a significant improvement over the state-of-the-art devices, and enhancing detection capability across handgrip-free scenarios.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10462909/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10462909/</guid>
</item>
<item>
<title>Optimization Method for Node Deployment of Closed-Barrier Coverage in Hybrid Directional Sensor Networks</title>
<description><p><span><big>Optimization Method for Node Deployment of Closed-Barrier Coverage in Hybrid Directional Sensor Networks</big></span><br></p><p><span><small><i>Peng Wang; Yonghua Xiong; Jinhua She; Anjun Yu</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3378998</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Closed-barrier coverage is a new coverage problem that requires sensors on the barrier to form a closed loop. Due to the features of end-to-end connection, constructing closed barriers is a challenging task. In a Hybrid Directional Sensor Network, sensors can both rotate and move, making it more difficult to schedule sensors to construct closed barriers. To address this challenge, we propose an optimization method for node deployment of closed-barrier coverage (CCOM). First, selecting multiple initiators transforms the closed-barrier coverage problem into a simple linear-barrier construction problem. Then, a weighted barrier graph is used to search the optimal barrier path. Finally, design a rotation strategy and moving plan for the sensors, fully utilizing their rotation and moving capabilities to achieve closed-barrier coverage with the minimum number of sensors and energy consumption. The simulation experiment results show that compared with other advanced methods, our proposed method has better performance in terms of barrier construction success rate, minimum required number of mobile sensors, moving distance, and network lifetime.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10478814/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10478814/</guid>
</item>
<item>
<title>Enhanced Multicast Protocol for Low-power and Lossy IoT Networks</title>
<description><p><span><big>Enhanced Multicast Protocol for Low-power and Lossy IoT Networks</big></span><br></p><p><span><small><i>Issam Eddine Lakhlef; Badis Djamaa; Mustapha Reda Senouci; Abbas Bradai</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3375797</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Communication protocols in the Internet of Things (IoT) should take into account the resource-constrained nature of Low-power and Lossy Networks (LLNs). IP multicast protocols allow a packet to be routed from one source to multiple destinations in a single transmission. Hence, resources such as bandwidth, energy, and time are saved for a multitude of LLN applications, ranging from over-the-air programming and information sharing to device configuration and resource discovery. In this context, several multicast routing protocols have recently been proposed for LLNs, including the Multicast Protocol for LLNs (MPL). MPL has proven to be very reliable in many scenarios. However, the great resource consumption, especially in terms of energy and bandwidth, remains the main drawback of this protocol. In this paper, after a detailed overview, we provide an in-depth analysis of the MPL protocol to highlight its functional weaknesses. Then, we propose several improvements that touch on different areas to address such limitations. Extensive realistic simulations and experiments were performed to study the performance of the proposed improvements to MPL. The results obtained show that our proposals outperform the MPL protocol in terms of resource consumption (memory, bandwidth, and energy) while improving its performance in terms of end-to-end delay and maintaining the same reliability of data packet delivery.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10473686/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10473686/</guid>
</item>
<item>
<title>A 7.25μK Ultra-High Resolution MEMS Resonant Thermometer</title>
<description><p><span><big>A 7.25μK Ultra-High Resolution MEMS Resonant Thermometer</big></span><br></p><p><span><small><i>Zheng Wang; Liangbo Ma; Xiaorui Bie; Xingyin Xiong; Zhaoyang Zhai; Wuhao Yang; Yongjian Lu; Xudong Zou</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3370956</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This paper proposes an ultra-high-resolution MEMS resonant thermometer that exploits differences in Young’s modulus and coefficient of thermal expansion (CTE) between structural layers to increase thermal stress due to temperature variation. The increased thermal stress eventually acts on the resonator to produce a large frequency shift, and the temperature coefficient of frequency (TCF) can be increased to 17.4 times the original, which is consistent with theoretical analysis and finite element method simulations. The MEMS resonator chip is manufactured by the standard Silicon-on-insulator process and stacked on a ceramic chip carrier through multiple layers of materials. A self-sustaining oscillator, mainly composed of a packaged MEMS resonator chip and a low-noise application-specific integrated circuit (ASIC), is built to track the resonant frequency shift of the resonator. This MEMS resonant thermometer prototype demonstrates a high temperature coefficient of frequency (TCF) of -866.84ppm/K from -50°C to 40°C and exhibits good linearity. An ultra-high resolution of 7.25μK is achieved in the closed-loop experimental test. This is the best result achieved for a MEMS thermometer employing the resonant sensing paradigm to date.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10460450/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10460450/</guid>
</item>
<item>
<title>Design of MTM-based Multi-band Micro-Biosensor in Terahertz region as perfect absorber for Early-Stage Leukemia Diagnosis with sensitivity 18626373 TH...</title>
<description><p><span><big>Design of MTM-based Multi-band Micro-Biosensor in Terahertz region as perfect absorber for Early-Stage Leukemia Diagnosis with sensitivity 18626373 THz/RIU</big></span><br></p><p><span><small><i>Musa N. Hamza; Mohammad Tariqul Islam</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383522</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this article, a novel highly sensitive biosensor based on perfect metamaterial absorbers is presented. A detailed study is presented for several different models, different types of substrate materials, different types of resonant materials and substrate thicknesses showing very good sensitivity to any changes. The proposed biosensor exhibits high sensitivity to different polarization and incidence angles, ensuring its performance in terms of signal-to-noise ratio. The proposed biosensor was carefully compared with previously designed sensors and biosensors. The proposed biosensor exhibits amazing sensitivity, such as a quality factor of 41.1, a figure of merit (FOM) of 172466416 RIU-1, and a sensitivity (S) of 18626373 THz/RIU. The high sensitivity of the biosensor allows early detection of leukemia, demonstrating significant differences between leukemia and normal blood. Another interesting result of this article is the use of terahertz waves for imaging. Microwave imaging is performed for electric fields, magnetic fields, and power.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10494209/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10494209/</guid>
</item>
<item>
<title>LEPS: A lightweight and effective single-stage detector for pothole segmentation</title>
<description><p><span><big>LEPS: A lightweight and effective single-stage detector for pothole segmentation</big></span><br></p><p><span><small><i>Xiaoning Huang; Jintao Cheng; Qiuchi Xiang; Jun Dong; Jin Wu; Rui Fan; Xiaoyu Tang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2023.3330335</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Currently, the problem of potholes on urban roads is becoming increasingly severe. Identifying and locating road potholes promptly has become a major challenge for urban construction. Therefore, we proposed a lightweight and effective instance segmentation method called LEPS (Lightweight and Effective Pothole Segmentation Detector) for road pothole detection. To extract the image edge information and gradient information of potholes from the feature map more efficiently, we proposed a module that performs the convolutional super-position supplemented with a convolutional kernel to enhance spatial details for the backbone. We have designed a novel module applied to the Neck layer, improving the detection performance while reducing the parameters. To enable accurate segmentation of fine-grained features, we optimized the ProtoNet, which enables the segmentation head to generate high-quality masks for more accurate prediction. We have fully demonstrated the effectiveness of the method through a large number of comparative experiments. Our detector has excellent performance on two authoritative example datasets, URTIRI and COCO, and can successfully be applied to visual sensors to accurately detect and segment road potholes in real environments, accurately LEPS reached 0.892 and 0.648 in terms of Mask for AP50 and AP50:95, which is improved by 4.6% and 20.6% compared to the original model. These results demonstrate its strong competitiveness when compared to other models. Comprehensively, LEPS improves detection accuracy while maintaining lightweight, which allows the model to meet the practical application requirements of edge computing devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10315064/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10315064/</guid>
</item>
<item>
<title>UMHE: Unsupervised Multispectral Homography Estimation</title>
<description><p><span><big>UMHE: Unsupervised Multispectral Homography Estimation</big></span><br></p><p><span><small><i>Jeongmin Shin; Jiwon Kim; Seokjun Kwon; Namil Kim; Soonmin Hwang; Yukyung Choi</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383453</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Multispectral image alignment plays a crucial role in exploiting complementary information between different spectral images. Homography-based image alignment can be a practical solution considering a tradeoff between runtime and accuracy. Existing methods, however, have difficulty with multispectral images due to the additional spectral gap or require expensive human labels to train models. To solve these problems, this paper presents a comprehensive study on multispectral homography estimation in an unsupervised learning manner. We propose a curriculum data augmentation, an effective solution for models learning spectrum-agnostic representation by providing diverse input pairs. We also propose to use the phase congruency loss that explicitly calculates the reconstruction between images based on low-level structural information in the frequency domain. To encourage multispectral alignment research, we introduce a novel FLIR corresponding dataset that has manually labeled local correspondences between multispectral images. Our model achieves state-of-the-art alignment performance on the proposed FLIR correspondence dataset among supervised and unsupervised methods while running at
151 FPS
. Furthermore, our model shows good generalization ability on the M3FD dataset without finetuning.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10494213/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10494213/</guid>
</item>
<item>
<title>Calibration and Evaluation of a Low-Cost Optical Particulate Matter Sensor for Measurement of Lofted Lunar Dust</title>
<description><p><span><big>Calibration and Evaluation of a Low-Cost Optical Particulate Matter Sensor for Measurement of Lofted Lunar Dust</big></span><br></p><p><span><small><i>Abhay Vidwans; Jeffrey Gillis-Davis; Pratim Biswas</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3366436</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Several recent Earth-based investigations employed low-cost particulate matter sensors to address the lack of spatiotemporal resolution in air quality data. The lunar environment also has a particulate matter problem in the form of fine lofted and levitated dust particles. Natural and anthropogenic mobilized dust can cause a slew of difficulties for surface operations (deposition onto radiators, optical components, and mechanical devices). Despite the urgency of mitigating dust on the Moon and other airless bodies, the performance of low-cost sensors has not been critically evaluated for space applications. Upcoming long-term robotic and human exploration missions to the Moon necessitate a robust sensor that can monitor particulate matter levels and establish a spatially and temporally resolved global network. In this work, we calibrate two optical light-scattering particulate matter sensors against research-grade aerosol instruments for measuring the concentration of aerosolized lunar simulants. Sensors showed a stronger dependence on aerosol particle size distribution than particle composition. Vacuum testing showed a significant deviation in performance compared to atmospheric pressure, with a stronger dependency on lunar simulant. The predicted mass deposition, based on sensor output coupled with dust trajectory, was within an order of magnitude of the reference deposition. Our results demonstrate for the first time that low-cost particulate matter sensors can monitor dust concentrations with reasonable accuracy in a vacuum environment, with two caveats. First, precise calibrations must be performed with a dust simulant closely matching the particle size distribution of the target dust, and second, atmospheric pressure calibrations alone are insufficient.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10462021/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10462021/</guid>
</item>
<item>
<title>Reweighting Interacting Multiple-model Algorithm to Overcome Model Competition for Target Tracking in the Hybrid System</title>
<description><p><span><big>Reweighting Interacting Multiple-model Algorithm to Overcome Model Competition for Target Tracking in the Hybrid System</big></span><br></p><p><span><small><i>Guowei Li; Shurui Zhang; Yubing Han; Weixing Sheng; Thia Kirubarajan</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3369854</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The vicious competition of interacting multiple-model algorithm (IMM) is an inherent problem and would produce irreversible effects on IMM estimation results, especially combining with the radar system. In this paper, a novel reweighting IMM (RIMM) is proposed to overcome this issue. Firstly, the theoretical lower bound of model numbers in different situations is respectively provided through the analysis of IMM limitations. Furthermore, certificate the influence of model inaccuracy on the Kalman filter, which illustrates an effective method for reducing errors is increasing model numbers. Thirdly, the definition of model set density and the analysis of the true model space are given, and their connection establishes the standard of how to design the model set or add the model number. Finally, an effective method called RIMM is provided to overcome the competition caused by model increasing. The proposed RIMM holds strong adaptability for different model sets. The simulations of RIMM highlight the correctness and effectiveness of the proposed methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10462948/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10462948/</guid>
</item>
<item>
<title>A Stretchable and Wearable Ultrasonic Transducer Array for Bladder Volume Monitoring Application</title>
<description><p><span><big>A Stretchable and Wearable Ultrasonic Transducer Array for Bladder Volume Monitoring Application</big></span><br></p><p><span><small><i>Cong Pu; Ben Fu; Lehang Guo; Huixiong Xu; Chang Peng</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3382244</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Monitoring bladder volume is an essential application for patients with voiding dysfunction. Current wearable ultrasonic bladder monitors are rigid, bulky and cannot conform to human skin. This study presents a stretchable and wearable ultrasonic transducer array that can both conform to non-planar skin surfaces and continuously monitor bladder volume. The wearable transducer array mainly consists of 4 × 4 piezoelectric transducer elements, stretchable serpentine-based electrodes for electrical interconnection, electrically conductive adhesive as both matching and backing layers, and silicone elastomer for encapsulation. The transducer array has a center frequency of 3.9 MHz, -6 dB acoustic bandwidth of 57.6%, and up to 40% elastic stretchability. The volume of balloon-bladder phantom was estimated by a least square ellipsoid fitting method. With the true volume of balloon-bladder phantoms ranging from 10 mL to 405 mL, the proposed transducer array has a mean absolute percentage error of 9.4%, a mean absolute error of 11.9 mL and a coefficient of determination of 0.975, which demonstrates the capability of the proposed stretchable and wearable ultrasonic transducer array for bladder volume monitoring application.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10489841/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10489841/</guid>
</item>
<item>
<title>Intelligent fault diagnosis of bearing using multiwavelet perception kernel convolutional neural network</title>
<description><p><span><big>Intelligent fault diagnosis of bearing using multiwavelet perception kernel convolutional neural network</big></span><br></p><p><span><small><i>Yuanyuan Zhou; Hang Wang; Yongbin Liu; Xianzeng Liu; Zheng Cao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3370564</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Strong background noise characteristics of vibration signals cause issues with poor identification capability of features by fault diagnostic models. To address this issue, a method is proposed for intelligent fault diagnosis of bearing using multiwavelet perception kernel (MPK) and feature attention convolutional neural network (FA-CNN). First, four multiwavelet perception kernels are constructed to decompose the vibration signals in full-band multilevel. Second, improved multiwavelet information entropy (IMIE) of the frequency band components is calculated. The calculated component entropies of the corresponding frequency bands are integrated to construct frequency band clusters (FBC) from low to high frequencies. Third, joint approximate diagonalization eigen (JADE) is introduced to perform feature fusion for every FBC to eliminate redundant information, and fused features from low to high frequencies are obtained as original inputs. The FA-CNN bearing fault diagnosis framework is constructed for intelligent fault diagnosis of bearings. Finally, the effectiveness of the proposed method is verified by two cases. The results show that the proposed method has high fault feature recognition capability.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10458914/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10458914/</guid>
</item>
<item>
<title>Improved multipath mitigation using multiple trend-surface hemispherical map in GNSS precise point positioning</title>
<description><p><span><big>Improved multipath mitigation using multiple trend-surface hemispherical map in GNSS precise point positioning</big></span><br></p><p><span><small><i>Haijun Yuan; Zhetao Zhang; Xiufeng He; Jinwen Zeng; Hao Wang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3374458</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Global Navigation Satellite System (GNSS) Precise Point Positioning (PPP) is highly appreciated as a positioning technology; however, multipath restricts its accuracy and reliability. In this study, we proposed a Multiple Trend-surface Multipath Hemispherical Map (MT-MHM), where discrepant multipath fading frequency or orbit orientation of each satellite in each azimuth and elevation grid is considered. In a typical static scenario, consecutive 8-day observations are collected to extract multipath and conduct PPP experiments. Compared with traditional multipath hemispherical map based on trend-surface fitting, MT-MHM has better modelling performance of residuals and improves standard deviations of residuals for all satellites. Besides, the positioning errors and required convergence epochs of PPP are reduced by using MT-MHM. For the positioning accuracy of all solutions, MT-MHM exhibits 50.8, 15.2, and 27.9% improvements compared with that without multipath correction in the east, north, and up directions, respectively. In conclusion, our proposed MT-MHM exhibits better performance in terms of residual reduction, convergence time, and positioning accuracy in PPP.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472420/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472420/</guid>
</item>
<item>
<title>Design optimization and characterization of a 3D-printed tactile sensor for tissue palpation</title>
<description><p><span><big>Design optimization and characterization of a 3D-printed tactile sensor for tissue palpation</big></span><br></p><p><span><small><i>D. Lo Presti; L. Zoboli; A. Addabbo; D. Bianchi; A. Dimo; C. Massaroni; V. Altomare; A. Grasso; A. Gizzi; E. Schena</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3369337</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Manual palpation is a crucial medical procedure that relies on surface examination to detect internal tissue abnormalities, heavily reliant on healthcare professionals’ expertise and tactile sensitivity. To tackle these issues, smart palpation systems based on electrical or optical sensors have been developed to perform quantitative tactile measurements, crucial for identifying various solid tumors, including breast and prostate cancer by assessing tissue mechanical properties. In this context, fiber Bragg gratings (FBGs) are emerging as a promising technology due to their advantages (e.g., high metrological properties, multiplexing capacity, and easy packaging) making them ideal for tactile sensing. This study explores the benefits of FBG and 3D printing to develop a tactile sensor for tissue palpation. First, an optimization of the design of the sensing core of a previously developed probe was conducted through finite element analysis. The novel structure addresses the primary limitation of the previous solution, where non-uniform strain distribution on the encapsulated FBG causes compression on the grating with high risk of bending and breakage. In contrast, the modeled geometry ensures FBG elongation during tissue palpation. A 3D printing and characterization of the proposed solution was carried out to investigate the response of the enclosed FBG when pushed against different materials showing promising results in discriminating tissues according to their mechanical properties: the more rigid the indented substrate the higher the sensor output. This property will be fundamental for enhancing early tumor detection through superficial tissue palpation, advancing the efficacy of prevention measures.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10455966/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10455966/</guid>
</item>
<item>
<title>Deep Reinforcement Learning-based Joint Sequence Scheduling and Trajectory Planning in Wireless Rechargeable Sensor Networks</title>
<description><p><span><big>Deep Reinforcement Learning-based Joint Sequence Scheduling and Trajectory Planning in Wireless Rechargeable Sensor Networks</big></span><br></p><p><span><small><i>Chengpeng Jiang; Wencong Chen; Ziyang Wang; Wendong Xiao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373664</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Mobile charging has become a popular and efficient method for replenishing energy. This is done with a mobile charger (MC) and wireless energy transfer technology (WET), which helps to alleviate the issue of energy constraints in wireless rechargeable sensor networks (WRSNs). Notably, designing mobile charging scheduling schemes is essential for improving charging performance. Most current studies assume that the networks are obstacle-free. Unlike the existing studies, this paper focuses on joint sequence scheduling and trajectory planning problems (JSSTP), which assumes that the network has multiple static obstacles. To address this issue, we propose a novel deep reinforcement learning-based JSSTP (DRL-JSSTP) that enables the MC to avoid obstacles and reach the charging target to charge the sensors fully. This approach maximizes energy usage efficiency and sensor survival rate while satisfying MC energy capacity constraints. DRL-JSSTP includes a charging target selector and a trajectory planner, which determine the index of the next charging target and plan the movement trajectory to avoid obstacles, respectively. We adopt 1-D convolutional neural networks to extract feature information about the environment state and gated recurrent units to predict the charging decisions. Simulation results demonstrate that DRL-JSSTP outperforms existing approaches, achieving higher energy usage efficiency and sensor survival rate.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466524/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466524/</guid>
</item>
<item>
<title>A Simultaneous Wirel |
This comment was marked as outdated.
This comment was marked as outdated.
|
Successfully generated as following: http://localhost:1200/ieee/journal/8782710 - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Open Journal of Signal Processing</title>
<link>https://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=8782710</link>
<atom:link href="http://localhost:1200/ieee/journal/8782710" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Open Journal of Signal Processing - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Wed, 10 Apr 2024 12:36:37 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Synthbuster: Towards Detection of Diffusion Model Generated Images</title>
<description><p><span><big>Synthbuster: Towards Detection of Diffusion Model Generated Images</big></span><br></p><p><span><small><i>Quentin Bammey</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337714</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Synthetically-generated images are getting increasingly popular. Diffusion models have advanced to the stage where even non-experts can generate photo-realistic images from a simple text prompt. They expand creative horizons but also open a Pandora's box of potential disinformation risks. In this context, the present corpus of synthetic image detection techniques, primarily focusing on older generative models like Generative Adversarial Networks, finds itself ill-equipped to deal with this emerging trend. Recognizing this challenge, we introduce a method specifically designed to detect synthetic images produced by diffusion models. Our approach capitalizes on the inherent frequency artefacts left behind during the diffusion process. Spectral analysis is used to highlight the artefacts in the Fourier transform of a residual image, which are used to distinguish real from fake images. The proposed method can detect diffusion-model-generated images even under mild
jpeg
compression, and generalizes relatively well to unknown models. By pioneering this novel approach, we aim to fortify forensic methodologies and ignite further research into the detection of AI-generated images.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334046/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334046/</guid>
</item>
<item>
<title>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</title>
<description><p><span><big>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</big></span><br></p><p><span><small><i>Yanbin Zou; Jingna Fan; Zekai Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we consider using time-of-arrival (TOA) measurements from a single moving receiver to locate a moving target at constant velocity that emits a periodic signal with unknown signal period. First, we give the TOA measurement model and deduce the Cram
$\acute{\text{e}}$
r-Rao lower bounds (CRLB). Then, we formulate a nonlinear least squares (NLS) problem to estimate the unknown parameters. We use semidefinite programming (SDP) techniques to relax the nonconvex NLS problem. However, it is shown that the SDP localization algorithm cannot provide a high-quality solution. Subsequently, we develop a fixed point iteration (FPI) method to improve the performance of the SDP algorithm. In addition, we also consider the presence of receiver position errors and develop the corresponding localization algorithm. Numerical simulations are conducted to demonstrate the localization performance of the proposed algorithms by comparing them with the CRLB.
Index Term
-Fixed point iteration (FPI), semidefinite programming (SDP), single moving receiver, target localization, time-of-arrival (TOA).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336384/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336384/</guid>
</item>
<item>
<title>The Neural-SRP Method for Universal Robust Multi-Source Tracking</title>
<description><p><span><big>The Neural-SRP Method for Universal Robust Multi-Source Tracking</big></span><br></p><p><span><small><i>Eric Grinstein; Christopher M. Hicks; Toon van Waterschoot; Mike Brookes; Patrick A. Naylor</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340057</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Neural networks have achieved state-of-the-art performance on the task of acoustic Direction-of-Arrival (DOA) estimation using microphone arrays. Neural models can be classified as end-to-end or hybrid, each class showing advantages and disadvantages. This work introduces Neural-SRP, an end-to-end neural network architecture for DOA estimation inspired by the classical Steered Response Power (SRP) method, which overcomes limitations of current neural models. We evaluate the architecture on multiple scenarios, namely, multi-source DOA tracking and single-source DOA tracking under the presence of directional and diffuse noise. The experiments demonstrate that our proposed method compares favourably in terms of computational and localization performance with established neural methods on various recorded and simulated benchmark datasets.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345765/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345765/</guid>
</item>
<item>
<title>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</title>
<description><p><span><big>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</big></span><br></p><p><span><small><i>Alexander Venus; Erik Leitinger; Stefan Tertinek; Klaus Witrisal</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper presents a neural-enhanced probabilistic model and corresponding factor graph-based sum-product algorithm for robust localization and tracking in multipath-prone environments. The introduced hybrid probabilistic model consists of physics-based and data-driven measurement models capturing the information contained in both, the line-of-sight (LOS) component as well as in multipath components (NLOS components). The physics-based and data-driven models are embedded in a joint Bayesian framework allowing to derive from first principles a factor graph-based algorithm that fuses the information of these models. The proposed algorithm uses radio signal measurements from multiple base stations to robustly estimate the mobile agent's position together with all model parameters. It provides high localization accuracy by exploiting the position-related information of the LOS component via the physics-based model and robustness by exploiting the geometric imprint of multipath components independent of the propagation channel via the data-driven model. In a challenging numerical experiment involving obstructed LOS situations to all anchors, we show that the proposed sequential algorithm significantly outperforms state-of-the-art methods and attains the posterior Cramér-Rao lower bound even with training data limited to local regions.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336409/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336409/</guid>
</item>
<item>
<title>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</title>
<description><p><span><big>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</big></span><br></p><p><span><small><i>AMIT EFRAIM; Joseph M. Francos</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In solving a model fitting problem, the existence of outliers in the set of measurements can have a devastating effect on the solution accuracy. Traditionally, in order to overcome this problem, robust point cloud registration algorithms are composed of transformation hypothesis generation, followed by hypothesis evaluation aimed at selecting the best hypothesized estimate. Hypotheses evaluation is commonly performed using the sample consensus criterion. However, since this criterion accounts only for the consensus size, it fails when the maximal sample consensus is incorrect. We propose a new hypothesis evaluation approach, generalizing the sample consensus approach, where instead of the sample consensus, the transformation that maximizes the point clouds feature correlation is selected as the best hypothesis. The feature vector at each point contains information such as on local geometry and semantic context. Utilizing this information in the hypotheses evaluation and selection procedure allows for a correct decision even when the hypothesis yielding the maximal sample consensus is false. Consequently, the probability of selecting the correct model increases. We show both mathematically and empirically that substituting the sample consensus criterion with the proposed point cloud feature correlation hypothesis test (PC-FCHT) lowers the probability of large registration errors, compared to using the special case of sample consensus. The proposed PC-FCHT is applicable to any algorithm that follows the hypothesis generation and evaluation scheme, potentially improving the success rates of a wide family of point cloud registration algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345750/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345750/</guid>
</item>
<item>
<title>Joint PAPR and OBP Reduction for NC-OFDM Systems</title>
<description><p><span><big>Joint PAPR and OBP Reduction for NC-OFDM Systems</big></span><br></p><p><span><small><i>Hsuan-Fu Wang; Fang-Biau Ueng; Bo-Heng Yeh</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3329757</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The spectrum resource is always a critical issue for wireless communications since it directly impacts the data rate and capacity. However, the problem of spectrum resource scarcity always exists. Moreover, spectrum resource scarcity becomes more severe as new communication technologies and wireless applications sprout. Noncontiguous orthogonal frequency division multiplexing (NC-OFDM) is a multicarrier method for bandwidth utilization. Unfortunately, this system has two fatal defects: high peak-to-average power ratio (PAPR) and considerable out-of-band power (OBP), which are detrimental to the system's performance. To solve these two problems, we propose a convex optimization-based method for joint PAPR and OBP reduction in NC-OFDM Systems. The strategy is to permit the secondary user to utilize the unoccupied spectrum of the primary user with dynamic spectrum sharing (DSS) based on a cognitive radio network (CRN). To this end, a flexible system operating over noncontiguous bands and DSS scenarios is necessary. The simulation results have shown that our method could effectively improve the overall performance and outperform other schemes, i.e., projections onto convex sets (POCS) and alternating projections onto convex and non-convex sets (APOCNCS), without harming the transmission of the primary system. The collaboration between secondary and primary systems is viable with the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10305259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10305259/</guid>
</item>
<item>
<title>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</title>
<description><p><span><big>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</big></span><br></p><p><span><small><i>Gal Itzhak; Israel Cohen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339433</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article presents a Kronecker-product (KP) beamforming approach incorporating sparse concentric circular arrays (SCCAs). The locations of the microphones on the SCCA are optimized concerning the broadband array directivity over a wide range of direction-of-arrival (DOA) deviations of a desired signal. A maximum directivity factor (MDF) sub-beamformer is derived accordingly with the optimal locations. Then, we propose two global beamformers obtained as a Kronecker product of a uniform linear array (ULA) and the SCCA sub-beamformer. The global beamformers differ by the type of the ULA, which is designed either as an MDF sub-beamformer along the
$\mathsf {x}$
-axis or as a maximum white noise gain sub-beamformer along the
$\mathsf {y}$
-axis. We analyze the performance of the proposed beamformers in terms of the directivity factor, the white noise gain, and their spatial beampatterns. Compared to traditional beamformers, the proposed beamformers exhibit considerably larger tolerance to DOA deviations concerning both the azimuth and elevation angles. Experimental results with speech signals in noisy and reverberant environments demonstrate that the proposed approach outperforms traditional beamformers regarding the perceptual evaluation of speech quality (PESQ) and short-time objective intelligibility (STOI) scores when the desired speech signals deviate from the nominal DOA.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342869/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342869/</guid>
</item>
<item>
<title>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</title>
<description><p><span><big>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</big></span><br></p><p><span><small><i>Karn N. Watcharasupat; Chih-Wei Wu; Yiwei Ding; Iroro Orife; Aaron J. Hipple; Phillip A. Williams; Scott Kramer; Alexander Lerch; William Wolcott</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339428</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Cinematic audio source separation is a relatively new subtask of audio source separation, with the aim of extracting the dialogue, music, and effects stems from their mixture. In this work, we developed a model generalizing the Bandsplit RNN for any complete or overcomplete partitions of the frequency axis. Psychoacoustically motivated frequency scales were used to inform the band definitions which are now defined with redundancy for more reliable feature extraction. A loss function motivated by the signal-to-noise ratio and the sparsity-promoting property of the 1-norm was proposed. We additionally exploit the information-sharing property of a common-encoder setup to reduce computational complexity during both training and inference, improve separation performance for hard-to-generalize classes of sounds, and allow flexibility during inference time with detachable decoders. Our best model sets the state of the art on the Divide and Remaster dataset with performance above the ideal ratio mask for the dialogue stem.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342812/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342812/</guid>
</item>
<item>
<title>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</title>
<description><p><span><big>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</big></span><br></p><p><span><small><i>Miguel Ferrer; María de Diego; Alberto Gonzalez</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340106</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The LMS algorithm is widely employed in adaptive systems due to its robustness, simplicity, and reasonable performance. However, it is well known that this algorithm suffers from a slow convergence speed when dealing with colored reference signals. Numerous variants and alternative algorithms have been proposed to address this issue, though all of them entail an increase in computational cost. Among the proposed alternatives, the affine projection algorithm stands out. This algorithm has the peculiarity of starting from
$N$
data vectors of the reference signal. It transforms these vectors into as many data vectors suitably normalized in energy and mutually orthogonal. In this work, we propose a version of the LMS algorithm that, similar to the affine projection algorithm, starts from
$N$
data vectors of the reference signal but corrects them by using only a scalar factor that functions as a convergence step. Our goal is to align the behavior of this algorithm with the behavior of the affine projection algorithm without significantly increasing the computational cost of the LMS.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345730/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345730/</guid>
</item>
<item>
<title>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</title>
<description><p><span><big>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</big></span><br></p><p><span><small><i>Zongyao Li; Ren Togo; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340616</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article proposes a method for transferring knowledge of semantic segmentation from a labeled source domain to an unlabeled target domain without using the source-domain data. Such a problem is called source-data-free domain adaptation, in which a pre-trained source-domain model and the unlabeled target-domain data are used to transfer the label knowledge across the domains. Like most previous methods, our method uses pseudo labels for distilling and transferring the source-domain knowledge. On the basis of the pseudo-label learning, our method improves the domain adaptation performance in two innovative ways: 1) reducing the domain differences by source-data-free style transfer and 2) exploring the style diversity within the target domain by style modification. To this end, we introduce two additional modules: 1) an inter-domain style transfer module which aligns the feature statistics of the source and target domains before producing the pseudo labels thereby improving the pseudo labels' accuracy, and 2) an intra-domain style modification module which modifies the image styles within the target domain for learning intra-domain style-invariant features. Our method with the two modules outperforms previous source-data-free domain adaptation methods in two commonly used benchmarks. Moreover, our method is well compatible with the previous methods for further improvement.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10356748/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10356748/</guid>
</item>
<item>
<title>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</title>
<description><p><span><big>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</big></span><br></p><p><span><small><i>Tobias Kabzinski; Peter Jax</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337721</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Quasi-continuous system identification of time-variant linear acoustic systems can be applied in various audio signal processing applications when numerous acoustic transfer functions must be measured. A prominent application is measuring head-related transfer functions. We treat the underlying multiple-input-multiple-output (MIMO) system identification problem in a state-space model as a joint estimation problem for states, representing impulse responses, and state-space model parameters using the expectation maximization (EM) algorithm. We address limitations of prior work by imposing different model structures, especially for dependencies within a (transformed) state vector. This results in block diagonal matrix structures, for which we derive M-step update rules. Making assumptions about this model structure and choosing a block size for a given application define the computational complexity. In examples, we found that applying this framework yields improvements of up to 10 dB in relative system distance in comparison to a conventional method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334061/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334061/</guid>
</item>
<item>
<title>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</title>
<description><p><span><big>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</big></span><br></p><p><span><small><i>Damir Rakhimov; Martin Haardt</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337729</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this paper, we present an analytical performance assessment of 2-D Tensor ESPRIT in terms of physical parameters. We show that the error in the
$r$
-mode depends only on two components, irrespective of the dimensionality of the problem. We obtain analytical expressions in closed form for the mean squared error (MSE) in each dimension as a function of the signal-to-noise (SNR) ratio, the array steering matrices, the number of antennas, the number of snapshots, the selection matrices, and the signal correlation. The derived expressions allow a better understanding of the difference in performance between the tensor and the matrix versions of the ESPRIT algorithm. The simulation results confirm the coincidence between the presented analytical expression and the curves obtained via Monte Carlo trials. We analyze the behavior of each of the two error components in different scenarios.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334446/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334446/</guid>
</item>
<item>
<title>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</title>
<description><p><span><big>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</big></span><br></p><p><span><small><i>Yaman Kındap; Simon Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343341</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this work we study linear vector stochastic differential equation (SDE) models driven by the generalised hyperbolic (GH) Lévy process for inference in continuous-time non-Gaussian filtering problems. The GH family of stochastic processes offers a flexible framework for modelling of non-Gaussian, heavy-tailed characteristics and includes the normal inverse-Gaussian, variance-gamma and Student-t processes as special cases. We present continuous-time simulation methods for the solution of vector SDE models driven by GH processes and novel inference methodologies using a variant of sequential Markov chain Monte Carlo (MCMC). As an example a particular formulation of Langevin dynamics is studied within this framework. The model is applied to both a synthetically generated data set and a real-world financial series to demonstrate its capabilities.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360268/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360268/</guid>
</item>
<item>
<title>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</title>
<description><p><span><big>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</big></span><br></p><p><span><small><i>James M. Cozens; Simon J. Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344048</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper proposes a probabilistic approach for extracting time-varying and irregular time signature information from polyphonic audio extracts, subsequently providing beat and bar line positions given inferred time signature divisions. This is achieved via dynamically evaluating the beat tempo as a function of time through finding an optimal compromise in beat and bar alignment in the time and tempo domains. Time signature divisions are determined based on a new representation, termed the Metrogram, that presents time-varying information regarding rhythmic and metric periodicities in the Tempogram. Our methodology is characterised by its ability to provide a distribution over metric interpretations, offering insights into the diverse ways music can be rhythmically perceived. Results indicate high-level accuracy for a variety of polyphonic extracts containing irregular, complex, irrational, and time-varying time signatures. Accuracy rivalling state-of-the-art methodologies is also reported in a beat tracking task performed on the standard Ballroom Dataset. The paper offers insights into the field of dynamic time signature recognition and beat tracking, offering a valuable and versatile resource for the analysis, composition, and performance of music.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363392/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363392/</guid>
</item>
<item>
<title>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</title>
<description><p><span><big>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</big></span><br></p><p><span><small><i>Yuto Watanabe; Ren Togo; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343335</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Although text-guided image manipulation approaches have demonstrated highly accurate performance for editing the appearance of images in a virtual or simple scenario, their real-world applications face significant challenges. The primary cause of these challenges is the misalignment in the distribution of training and real-world data, which leads to unstable text-guided image manipulation. In this work, we propose a novel framework called TolerantGAN and tackle the new task of real-world text-guided image manipulation independent of the training data. To achieve this, we introduce two key concepts of a border smoothly connection module (BSCM) and a manipulation direction-based attention module (MDAM). BSCM smoothens the misalignment in the distribution of training and real-world data. MDAM extracts only regions highly relevant for image manipulation and assists in reconstructing unobserved objects in the training data. For in-the-wild input images of various classes, TolerantGAN robustly outperforms the state-of-the-art methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360283/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360283/</guid>
</item>
<item>
<title>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</title>
<description><p><span><big>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</big></span><br></p><p><span><small><i>Anastasia Avdeeva; Aleksei Gusev; Tseren Andzhukaev; Artem Ivanov</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343342</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Whispered speech is a quiet voice without vocalization. One of the common cases of using whispered speech is a technique that can help overcome stuttering. But whispered speech can be uncomfortable and difficult to understand in everyday communication. To address these problems, we propose a method of low-delayed whisper-to-speech voice conversion, which can be useful in real life communication of people with disordered speech. As part of our research, we study the impact of streaming Automatic Speech Recognition models on the quality of voice conversion, comparing different streaming models and methods for model adaptation to streaming settings, and showing the importance of using such models in cases of low-delayed voice conversion.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360259/</guid>
</item>
<item>
<title>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</title>
<description><p><span><big>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</big></span><br></p><p><span><small><i>Denis C. Ilie-Ablachim; Andra Băltoiu; Bogdan Dumitrescu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344313</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We propose sparse representation and dictionary learning algorithms for dictionaries whose atoms are characterized by Gaussian probability distributions around some central atoms. This extension of the space covered by the atoms permits a better characterization of localized features of the signals. The representations are computed by trading-off representation error and the probability of the actual atoms used in the representation. We propose two types of algorithms: a greedy one, similar in spirit with Orthogonal Matching Pursuit, and one based on L1-regularization. We apply our new algorithms to unsupervised anomaly detection, where the representation error is used as anomaly score. The flexibility of our approach appears to improve more the representations of the many normal signals than those of the few outliers, at least for an anomaly type called dependency, thus improving the detection quality. Comparison with both standard dictionary learning algorithms and established anomaly detection methods is favorable.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365333/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365333/</guid>
</item>
<item>
<title>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</title>
<description><p><span><big>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</big></span><br></p><p><span><small><i>Yuya Moroto; Yingrui Ye; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344079</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>There are various sentiment theories for categorizing human sentiments into several discrete sentiment categories, which means that the theory used for training sentiment prediction methods does not always match that used in the test phase. As a solution to this problem, zero-shot visual sentiment prediction methods have been proposed to predict unseen sentiments for which no images are available in the training phase. However, the training of these previous zero-shot methods relies on a single sentiment theory, which limits their ability to handle sentiments from other theories. Thus, this article proposes a more robust zero-shot visual sentiment prediction method that can handle cross-domain sentiments defined in different sentiment theories. Specifically, by focusing on the fact that sentiments are abstract concepts common to humans regardless of whether their theories are different, we incorporate knowledge distillation into our method to construct a teacher–student model that can train the implicit relationships between sentiments defined in different sentiment theories. Furthermore, to enhance sentiment discrimination capability and strengthen the implicit relationships between sentiments, we introduce a novel sentiment loss between the teacher and student models. In this way, our model becomes robust to unseen sentiments by exploiting the implicit relationships between sentiments. The contributions of this article are the introduction of knowledge distillation and a novel sentiment loss between the teacher and student models for zero-shot visual sentiment prediction, and improved performance of zero-shot visual sentiment prediction. Experiments on several open datasets demonstrate the effectiveness of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363382/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363382/</guid>
</item>
<item>
<title>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</title>
<description><p><span><big>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</big></span><br></p><p><span><small><i>Chengen Liu; Geert Leus; Elvin Isufi</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339376</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The edge flow reconstruction task consists of retreiving edge flow signals from corrupted or incomplete measurements. This is typically solved by a regularized optimization problem on higher-order networks such as simplicial complexes and the corresponding regularizers are chosen based on prior knowledge. Tailoring this prior to the setting of interest can be challenging or it may not even be possible. Thus, we consider to learn this prior knowledge via a model-based deep learning approach. We propose a new regularized optimization problem for the simplicial edge flow reconstruction task, the simplicial ElasticNet, which combines the advantages of the
$\ell _{1}$
and
$\ell _{2}$
norms. We solve the simplicial ElasticNet problem via the multi-block alternating direction method of multipliers (ADMM) algorithm and provide conditions on its convergence. By unrolling the ADMM iterative steps, we develop a model-based neural network with a low requirement on the number of training data. This unrolling network replaces the fixed parameters in the iterative algorithm by learnable weights, thus exploiting the neural network's learning capability while preserving the iterative algorithm's interpretability. We enhance this unrolling network via simplicial convolutional filters to aggregate information from the edge flow neighbors, ultimately, improving the network learning expressivity. Extensive experiments on real-world and synthetic datasets validate the proposed approaches and show considerable improvements over both baselines and traditional non-model-based neural networks.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342735/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342735/</guid>
</item>
<item>
<title>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</title>
<description><p><span><big>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</big></span><br></p><p><span><small><i>Jim Beckers; Bart Van Erp; Ziyue Zhao; Kirill Kondrashov; Bert De Vries</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337718</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Bayesian model reduction provides an efficient approach for comparing the performance of all nested sub-models of a model, without re-evaluating any of these sub-models. Until now, Bayesian model reduction has been applied mainly in the computational neuroscience community on simple models. In this paper, we formulate and apply Bayesian model reduction to perform principled pruning of Bayesian neural networks, based on variational free energy minimization. Direct application of Bayesian model reduction, however, gives rise to approximation errors. Therefore, a novel iterative pruning algorithm is presented to alleviate the problems arising with naive Bayesian model reduction, as supported experimentally on the publicly available UCI datasets for different inference algorithms. This novel parameter pruning scheme solves the shortcomings of current state-of-the-art pruning methods that are used by the signal processing community. The proposed approach has a clear stopping criterion and minimizes the same objective that is used during training. Next to these benefits, our experiments indicate better model performance in comparison to state-of-the-art pruning schemes.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334001/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334001/</guid>
</item>
<item>
<title>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</title>
<description><p><span><big>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</big></span><br></p><p><span><small><i>Anh Minh Truong; Wilfried Philips; Peter Veelaert</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340064</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Many image-based rendering (IBR) methods rely on depth estimates obtained from structured light or time-of-flight depth sensors to synthesize novel views from sparse camera networks. However, these estimates often contain missing or noisy regions, resulting in an incorrect mapping between source and target views. This situation makes the fusion process more challenging, as the visual information is misaligned, inconsistent, or missing. In this work, we first implement a lightweight network based on the transformer, which is well-known for its capability to model long-range relationships within input data, to extract spatial features from color images. These features are then used to enhance the quality of completed depth maps. Furthermore, we combine a sequential deep neural network with a spatial attention mechanism to effectively fuse the projected features from multiple source viewpoints. This approach enables us to integrate information from an arbitrary number of source viewpoints as well as improve accuracy in synthesized views. Experimental results on challenging datasets demonstrate that our method achieves superior synthesized image quality compared to state-of-the-art (SOTA) methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345792/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345792/</guid>
</item>
<item>
<title>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</title>
<description><p><span><big>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</big></span><br></p><p><span><small><i>Reza Mirzaeifard; Naveen K. D. Venkategowda; Vinay Chakravarthi Gogineni; Stefan Werner</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344395</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper investigates quantile regression in the presence of non-convex and non-smooth sparse penalties, such as the minimax concave penalty (MCP) and smoothly clipped absolute deviation (SCAD). The non-smooth and non-convex nature of these problems often leads to convergence difficulties for many algorithms. While iterative techniques such as coordinate descent and local linear approximation can facilitate convergence, the process is often slow. This sluggish pace is primarily due to the need to run these approximation techniques until full convergence at each step, a requirement we term as a
secondary convergence iteration
. To accelerate the convergence speed, we employ the alternating direction method of multipliers (ADMM) and introduce a novel single-loop smoothing ADMM algorithm with an increasing penalty parameter, named SIAD, specifically tailored for sparse-penalized quantile regression. We first delve into the convergence properties of the proposed SIAD algorithm and establish the necessary conditions for convergence. Theoretically, we confirm a convergence rate of
$o({k^{-\frac{1}{4}}})$
for the sub-gradient bound of the augmented Lagrangian, where
$k$
denotes the number of iterations. Subsequently, we provide numerical results to showcase the effectiveness of the SIAD algorithm. Our findings highlight that the SIAD method outperforms existing approaches, providing a faster and more stable solution for sparse-penalized quantile regression.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365338/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365338/</guid>
</item>
<item>
<title>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</title>
<description><p><span><big>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</big></span><br></p><p><span><small><i>Oliver Lang; Christian Hofbauer; Reinhard Feger; Mario Huemer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343308</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>A promising waveform candidate for future joint sensing and communication systems is orthogonal frequency-division multiplexing (OFDM). For such systems, supporting multiple transmit antennas requires multiplexing methods for the generation of orthogonal transmit signals, where equidistant subcarrier interleaving (ESI) is the most popular multiplexing method. In this work, we analyze a multiplexing method called Doppler-division multiplexing (DDM). This method applies a phase shift from OFDM symbol to OFDM symbol to separate signals transmitted by different Tx antennas along the velocity axis of the range-Doppler map. The main focus of this work lies on the implications of DDM on the communication task. It will be shown that for DDM, the channels observed in the communication receiver are heavily time-varying, preventing any meaningful transmission of data when not taken into account. In this work, a communication system designed to combat these time-varying channels is proposed, which includes methods for data estimation, synchronization, and channel estimation. Bit error ratio (BER) simulations demonstrate the superiority of this communications system compared to ESI-based systems.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360223/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360223/</guid>
</item>
<item>
<title>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</title>
<description><p><span><big>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</big></span><br></p><p><span><small><i>Nils L. Westhausen; Bernd T. Meyer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343320</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we introduce a causal low-latency low-complexity approach for binaural multichannel blind speaker separation in noisy reverberant conditions. The model, referred to as Group Communication Binaural Filter and Sum Network (GCBFSnet) predicts complex filters for filter-and-sum beamforming in the time-frequency domain. We apply Group Communication (GC), i.e., latent model variables are split into groups and processed with a shared sequence model with the aim of reducing the complexity of a simple model only containing one convolutional and one recurrent module. With GC we are able to reduce the size of the model by up to 83% and the complexity up to 73% compared to the model without GC, while mostly retaining performance. Even for the smallest model configuration, GCBFSnet matches the performance of a low-complexity TasNet baseline in most metrics despite the larger size and higher number of required operations of the baseline.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360275/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360275/</guid>
</item>
<item>
<title>Reverse Ordering Techniques for Attention-Based Channel Prediction</title>
<description><p><span><big>Reverse Ordering Techniques for Attention-Based Channel Prediction</big></span><br></p><p><span><small><i>Valentina Rizzello; Benedikt Böck; Michael Joham; Wolfgang Utschick</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344024</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Channel state information (CSI) is crucial for enhancing the performance of wireless systems by allowing to adjust the transmission strategies based on the current channel conditions. However, obtaining precise CSI is difficult because of the fast-changing channel conditions caused by multi-path fading. An inaccurate CSI hinders the performance of various adaptive wireless systems, highlighting the need for channel prediction techniques to effectively mitigate the drawbacks of outdated CSI. Conventional methods typically depend on assumptions regarding user velocity or require knowledge of the Doppler frequency. In contrast to existing approaches, we aim for a more robust and practical solution by training neural networks without making any assumptions about user velocity, relying solely on noisy channel observations during training. Specifically, we adapt both the sequence-to-sequence with attention (Seq2Seq-attn) and transformer models for channel prediction. Additionally, a new technique called reverse positional encoding is introduced in the transformer model to improve the robustness of the model against varying sequence lengths. Similarly, the encoder outputs of the Seq2Seq-attn model are reversed prior to the application of attention mechanisms. By means of simulations, we show that these proposed techniques enable the models to effectively capture relationships within sequences of channel snapshots without increasing the complexity. Importantly, this capability remains robust across varying sequence lengths, representing a substantial improvement over existing methodologies.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363354/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363354/</guid>
</item>
<item>
<title>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</title>
<description><p><span><big>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</big></span><br></p><p><span><small><i>Sarina Meyer; Xiaoxiao Miao; Ngoc Thang Vu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344375</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Speaker anonymization is the task of modifying a speech recording such that the original speaker cannot be identified anymore. Since the first Voice Privacy Challenge in 2020, along with the release of a framework, the popularity of this research topic is continually increasing. However, the comparison and combination of different anonymization approaches remains challenging due to the complexity of evaluation and the absence of user-friendly research frameworks. We therefore propose an efficient speaker anonymization and evaluation framework based on a modular and easily extendable structure, almost fully in Python. The framework facilitates the orchestration of several anonymization approaches in parallel and allows for interfacing between different techniques. Furthermore, we propose modifications to common evaluation methods which improves the quality of the evaluation and reduces their computation time by 65 to 95%, depending on the metric. Our code is fully open source.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365329/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365329/</guid>
</item>
<item>
<title>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</title>
<description><p><span><big>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</big></span><br></p><p><span><small><i>Alessandro Ilic Mezza; Matteo Amerena; Alberto Bernardini; Augusto Sarti</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343318</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Real-time audio communications over IP have become essential to our daily lives. Packet-switched networks, however, are inherently prone to jitter and data losses, thus creating a strong need for effective packet loss concealment (PLC) techniques. Though solutions based on deep learning have made significant progress in that direction as far as speech is concerned, extending the use of such methods to applications of Networked Music Performance (NMP) presents significant challenges, including high fidelity requirements, higher sampling rates, and stringent temporal constraints associated to the simultaneous interaction between remote musicians. In this article, we present PARCnet, a hybrid PLC method that utilizes a feed-forward neural network to estimate the time-domain residual signal of a parallel linear autoregressive model. Objective metrics and a listening test show that PARCnet provides state-of-the-art results while enabling real-time operation on CPU.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360264/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360264/</guid>
</item>
<item>
<title>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</title>
<description><p><span><big>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</big></span><br></p><p><span><small><i>Katerina Zmolikova; Michael Syskind Pedersen; Jesper Jensen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Supervised learning-based speech enhancement methods often work remarkably well in acoustic situations represented in the training corpus but generalize poorly to out-of-domain situations, i.e. situations not seen during training. This stands in the way of further improvement of these methods in realistic scenarios, as collecting paired noisy-clean recordings in the target application domain is typically not feasible. Recording noisy-only in-domain data is, though, much more practical. In this article, we tackle the problem of unsupervised domain adaptation in speech enhancement. Specifically, we propose a way to use in-domain noisy-only data in the training of a neural network to improve upon a model trained solely on out-of-domain paired data. For this, we make use of masked spectrogram prediction, a technique from self-supervised learning that aims to interpolate masked regions of a spectrogram. We hypothesize that masked spectrogram prediction encourages learning of features that represent well both speech and noise components of the noisy signals. These features can then be used to train a more robust speech enhancement system. We evaluate the proposed method on the VoiceBank-DEMAND and LibriFSD50k databases, with WSJ0-CHiME3 serving as the out-of-domain database. We show that the proposed method outperforms both the out-of-domain system and the baseline approaches, i.e. RemixIT and noisy-target training, and also combines well with the previously proposed RemixIT method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360251/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360251/</guid>
</item>
<item>
<title>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</title>
<description><p><span><big>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</big></span><br></p><p><span><small><i>Alexander Karataev; Christian Forsch; Laura Cottatellucci</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3348343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We consider a cell-free massive multiple-input multiple-output (CF-MaMIMO) communication system in the uplink transmission and propose a novel algorithm for blind or semi-blind joint channel estimation and data detection (JCD). We formulate the problem in the framework of bilinear inference and develop a solution based on the expectation propagation (EP) method for both channel estimation and data detection. We propose a new approximation of the joint a posteriori distribution of the channel and data whose representation as a factor graph enables the application of the EP approach using the message-passing technique, local low-complexity computations at the nodes, and an effective modeling of channel-data interplay. The derived algorithm, called bilinear-EP JCD, allows for a distributed implementation among access points (APs) and the central processing unit (CPU) and has polynomial complexity. Our simulation results show that it outperforms other EP-based state-of-the-art polynomial time algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10378663/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10378663/</guid>
</item>
<item>
<title>Adversarial Representation Learning for Robust Privacy Preservation in Audio</title>
<description><p><span><big>Adversarial Representation Learning for Robust Privacy Preservation in Audio</big></span><br></p><p><span><small><i>Shayan Gharib; Minh Tran; Diep Luong; Konstantinos Drossos; Tuomas Virtanen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Sound event detection systems are widely used in various applications such as surveillance and environmental monitoring where data is automatically collected, processed, and sent to a cloud for sound recognition. However, this process may inadvertently reveal sensitive information about users or their surroundings, hence raising privacy concerns. In this study, we propose a novel adversarial training method for learning representations of audio recordings that effectively prevents the detection of speech activity from the latent features of the recordings. The proposed method trains a model to generate invariant latent representations of speech-containing audio recordings that cannot be distinguished from non-speech recordings by a speech classifier. The novelty of our work is in the optimization algorithm, where the speech classifier's weights are regularly replaced with the weights of classifiers trained in a supervised manner. This increases the discrimination power of the speech classifier constantly during the adversarial training, motivating the model to generate latent representations in which speech is not distinguishable, even using new speech classifiers trained outside the adversarial training loop. The proposed method is evaluated against a baseline approach with no privacy measures and a prior adversarial training method, demonstrating a significant reduction in privacy violations compared to the baseline approach. Additionally, we show that the prior adversarial method is practically ineffective for this purpose.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10379095/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10379095/</guid>
</item>
<item>
<title>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</title>
<description><p><span><big>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</big></span><br></p><p><span><small><i>Yasaman Parhizkar; Gene Cheung; Andrew W. Eckford</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>It is a popular hypothesis in neuroscience that ganglion cells in the retina are activated by selectively detecting visual features in an observed scene. While ganglion cell firings can be predicted via data-trained deep neural nets, the n... |
http://localhost:1200/ieee/journal/4427201/earlyaccess - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Sensors Journal</title>
<link>https://ieeexplore.ieee.org/xpl/tocresult.jsp?isnumber=4427201</link>
<atom:link href="http://localhost:1200/ieee/journal/4427201/earlyaccess" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Sensors Journal - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Wed, 10 Apr 2024 12:37:09 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Torsion Sensor Based on Helical Long-Period Grating Inscribed in the Etched Double Cladding Fiber</title>
<description><p><span><big>Torsion Sensor Based on Helical Long-Period Grating Inscribed in the Etched Double Cladding Fiber</big></span><br></p><p><span><small><i>Yanping He; Yuehui Ma; Chen Jiang; Yunqi Liu</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373623</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>We proposed a high-sensitivity torsion sensor based on the helical long-period grating (HLPG) inscribed in the cladding-etched double cladding fiber (DCF). The torsion sensitivity has been greatly improved by reducing the cladding diameter of the DCF, from 0.342 nm/(rad/m) to 1.162 nm/(rad/m), which is one order of magnitude higher than that of the conventional long-period fiber grating. The cladding mode coupling characteristics during the etching process are also investigated experimentally. The DCF-HLPG has a low strain and temperature sensitivity, and the maximum strain sensitivity is only about 0.0022 nm/με. The proposed DCF-HLPG has a potential application as a high-sensitivity torsion sensor with low strain and temperature crosstalk.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466518/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466518/</guid>
</item>
<item>
<title>Deep-Learning-Based Prediction Algorithm for Fuel Cell Electric Vehicle Energy with Shift Mixup</title>
<description><p><span><big>Deep-Learning-Based Prediction Algorithm for Fuel Cell Electric Vehicle Energy with Shift Mixup</big></span><br></p><p><span><small><i>Tae-Ho Kim; Jae-Heung Cho; Young-Kwang Kim; Joon-Hyuk Chang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373078</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The automobile industry is switching from fossil-fuel-based energy sources to green energy sources. In particular, because hydrogen has a high energy density and efficiency compared to other energy sources, it is a major research field for commercial vehicles that require large amounts of energy and travel long distances on average, such as buses and trucks. In fuel cell electric vehicles (FCEVs), maintaining an adequate energy production intensity is necessary to improve energy production efficiency and prevent falling into a state of inoperability owing to a lack of energy. In this case, energy prediction becomes a significant factor. In this study, a deep-learning-based prediction method for FCEV powertrain energy is proposed. The proposed method uses only internal data, which can be obtained from the vehicle, and does not require external information regarding future routes. Additionally, we designed a model considering the vehicle data time-series characteristics, and proposed a shift mixup, which is a data augmentation method that does not compromise the vehicle’s dynamic characteristics, to address the data shortage problem. Furthermore, a pretext task-learning method that can improve model performance without external data during inference is introduced. This method includes pretext tasks designed specifically for the vehicle domain. Finally, a distance-based loss mask and contrastive learning that the representation can learn semantic information are proposed. Experimental results for the actual driving dataset show that our method improved by 28.05% and by 30.04% compared to the exponential moving average in terms of root mean squared error and mean absolute error, respectively. We demonstrate with energy management strategy (EMS) that an effective energy prediction algorithm helps sustain an optimal state of charge (SOC).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466522/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466522/</guid>
</item>
<item>
<title>Effect of substrate temperature on Cu2-xO growth for enhanced photodetector performance in visible region</title>
<description><p><span><big>Effect of substrate temperature on Cu2-xO growth for enhanced photodetector performance in visible region</big></span><br></p><p><span><small><i>Karthickraja Ramakrishnan; Y. Ashok Kumar Reddy; B. Ajitha</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373768</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The present study investigates the effect of substrate temperature (30-300°C) on the sputter-deposited Cu
2-x
O films for improved photodetector performance in the visible region. Cubic structured Cu2O phase is observed for all deposited films through structural analysis. Increasing substrate temperature causes the formation of nanospheres with higher grain size at 200°C confirmed by the morphological images. Moreover, all the deposited thin films show a strong absorption band in the visible region (526-550 nm). Among deposited films, a lower bandgap (2.02 eV) is observed for the film deposited at 200°C due to the higher surface area-to-volume ratio and increased grain size. The better crystalline nature, more copper vacancy, and formation of nanospheres with a larger grain size of the film deposited at 200°C result in higher mobility and conductivity. Further, the photodetector performance of the Au/Cu
2-x
O/Au test-devices is studied under the visible light (λ=530 nm) irradiation. Among samples, the sample grown at 200°C showed higher photocurrent (36.10 μA), I
on
/I
off
ratio (183.50), photoresponsivity (1.127 A/W), specific detectivity (22.40×10
11
Jones), and faster response times (τ
r
=92 ms and τ
f
=84 ms) even at lower incident optical power density (0.1283 mW/cm
2
). These improved experimental results signify the potential of the fabricated test-device at 200°C for the optoelectronics devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10471292/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10471292/</guid>
</item>
<item>
<title>A Cavity-Type Pressure Sensor Array with High Anti-Disturbance Performance Inspired by Fish Lateral Line Canal</title>
<description><p><span><big>A Cavity-Type Pressure Sensor Array with High Anti-Disturbance Performance Inspired by Fish Lateral Line Canal</big></span><br></p><p><span><small><i>Qian Yang; Qiao Hu; Liuhao Shan; Guangyu Jiang; Yuanji Yao; Long Tang; Zicai Zhu; Alvo Aabloo</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3381162</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Effective sensing is challenging in real underwater environment swarming with various disturbances. However, most research on underwater sensors has been conducted in still water, which may not provide sufficient insights for environments with disturbances. In addressing this issue, our paper not only proposes the development of anti-disturbance sensor units but also delves into evaluating the localization capabilities of artificial lateral line arrays in disturbed water environments. Concretely, inspired by the lateral line canal system of fish, we proposed a cavity-type pressure sensing device (denoted as a C-sensor) with notable anti-disturbance performance. On one hand, a 3D-printed resin cavity is filled with silicone oil, with flexible latex film packaging the upper and lower openings to achieve anti-disturbance capabilities akin to those of fish lateral line canals. On the other hand, employing IPMC materials based on ion sensing principles, the sensor generates electrical signals similar to sensing cilia in fish lateral line systems. The results demonstrated that the cavityless sensor (denoted as N-sensor) showed self-oscillation behavior, leading to significant errors in weak target signal detection and even sensor failure in extreme cases. However, the C-sensor exhibits remarkable sensing performance in still water and water disturbances, thereby confirming the effectiveness of its cavity-type structure in mitigating the impact of water disturbances. Furthermore, we developed two distinct artificial lateral line arrays based on the above sensors (C-array and N-array). The experimental results showed a marked 78% reduction in mean localization error for C-array compared to N-array. Additonally, the C-array also exhibited good ability in multiple dipoles detection. Such results indicate the superior applicability of the C-array within real complex underwater environments, promoting the practical application of IPMC sensors.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10486206/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10486206/</guid>
</item>
<item>
<title>Multi-sensor multiple extended objects tracking based on the message passing</title>
<description><p><span><big>Multi-sensor multiple extended objects tracking based on the message passing</big></span><br></p><p><span><small><i>Yuansheng Li; Tao Shen; Lin Gao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3384560</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Multiple extended objects (EOs) tracking has attracted a lot of attention due to the fast development of high-resolution sensors. A remarkable feature of EO, compared to the traditional point target, is that an EO normally produces more than one measurement, resulting in challenges for finding the associations among objects and measurements. In this paper, we are interested in tracking an unknown number of EOs with multiple sensors by resorting to the message passing method. The existence probability and belief of each EO are explicitly estimated, where the belief is modeled by a mixture gamma Gaussian inverse Wishart (GGIW) distribution, so as to jointly estimate the measurement rate (MR), centroid state and extension. The marginal posterior of EOs is approximately by belief, which is obtained by running the loopy sum-product algorithm (LSPA) on a suitably devised factor graph. As a result, the computational load of the proposed algorithm increases linearly with respect to the number of targets, thus admitting the scalability. Simulation experiments are carried out to verify the performance of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10495766/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10495766/</guid>
</item>
<item>
<title>Dual-Resonance Split Ring Resonator Metasurface for Terahertz Biosensing</title>
<description><p><span><big>Dual-Resonance Split Ring Resonator Metasurface for Terahertz Biosensing</big></span><br></p><p><span><small><i>Arslan Asim; Michael Cada; Yuan Ma; Alan Fine</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3376290</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this paper, the ‘terahertz gap’ has been addressed by designing a novel THz metasurface for potential use in biosensing applications. The metasurface sensor employs Surface Plasmon Resonance (SPR). It operates in the 0 - 1 THz band. Two sharp reflection dips are provided by the sensor, which serve as indicators of analyte refractive index variations. Geometrical as well as compositional parameters of the biosensor design have been studied to optimize the performance in the targeted frequency band. The sensor design shows compatibility with different metals. The performance of the metasurface with gold, copper and aluminum has been investigated. The metasurface geometry is decently resilient to fabrication tolerances. The sensor maintains its resonance conditions when the angle of incidence is changed with minor deviations in the spectral response, but the polarization state of the incident terahertz beam clearly disturbs the absorption peak. Therefore, the sensing performance is restricted to a maximum allowable incidence angle of 20 degrees and circularly polarized terahertz beams. The resonance conditions for the metasurface appear around 0.4 and 0.7 THz. Both resonances have been investigated with respect to changes in the analyte refractive index. The chosen refractive index range is 1 to 1.5. The sensor response is calibrated by plotting the resonance frequency versus the refractive index. Least squares regression technique has been used to extract a data model for sensor response. Comparison of the proposed design with contemporary works has been incorporated into the paper. The sensor provides sensitivities of 0.1614 THz/ RIU and 0.23 THz/ RIU. The electromagnetic simulations have been carried out through the Finite Element Method (FEM).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10474307/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10474307/</guid>
</item>
<item>
<title>An Intestine based Bio-Compatible Humidity Sensor for Environmental and Medical Measurements</title>
<description><p><span><big>An Intestine based Bio-Compatible Humidity Sensor for Environmental and Medical Measurements</big></span><br></p><p><span><small><i>Emrehan Yavsan; Muhammet Rojhat Kara; Mehmet Akif Erismis</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3367032</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The bio-compatible devices suitable for recycling and bio-degrading can be achieved with organic materials in nature. In this work, a bio-compatible capacitive humidity sensor is presented for reducing the amount of electronic waste and contributing to the sustainability of natural resources and the future. The sensor consists of 3 layers. The first layer is the processed intestine layer of cattle. Bio-compatibility is achieved with this layer. In addition to being a highly absorbing tissue, the intestine has been used for centuries for long-term preservation of meat based food. Correspondingly, the developed sensor is found to be more durable and long-lasting than other natural-material based humidity sensors in the literature. The other layers of the sensor are interdigitated copper electrodes and a 0.2 mm thick thin film strip. Thin film strip increases mechanical strength as well as flexibility. The developed sensor prototype was subjected to various tests in the humidity range of 20%-90%. In these tests, the hysteresis characteristic of the sensor, its response-recovery time, and its long-term stability and short-term step responses were examined. Moreover, as a possible application in medicine, the sensor can be used to detect breathing cycles. The sensor’s response and recovery times were measured as 8.72’ and 4.47’, respectively, possibly attributed to the stabilization of our test setup, while the sensor successfully detected deep, normal and fast breathing. Despite being kept in an uncontrolled environment, the sensor continued to operate consistently for breath measurements after 56 weeks, which is more than a year.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10445282/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10445282/</guid>
</item>
<item>
<title>Deep neural network assisted terahertz metasurface sensors for the detection of lung cancer biomarkers</title>
<description><p><span><big>Deep neural network assisted terahertz metasurface sensors for the detection of lung cancer biomarkers</big></span><br></p><p><span><small><i>Jie Chen; Fangrong Hu; Xiaoya Ma; Mo Yang; Shangjun Lin; An Su</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3384578</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>MicroRNA (miRNA) is closely related to various cancers, and the change in its expression level is closely related to the development and death of tumor cells. Here, we design and manufacture a terahertz (THz) metasurface sensor to realize the concentration detection and category identification of the miRNAs related to lung cancer. We established two spectral classification datasets, one contains 9 concentrations of miRNAs, and the other contains 3 categories of lung cancer-related miRNAs. And then, we used a deep neural network (DNN) algorithm to classify these spectral datasets and obtained a LOD (Limit-of-Detection) of 100 aM for miRNAs. Moreover, this method can identify the categories of miRNAs. Compared with the other five machine learning (ML) algorithms, the proposed neural network framework achieves the best classification results. This work provides a new way for the detection and identification of trace nucleic acid and biomarkers of many cancers.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10495770/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10495770/</guid>
</item>
<item>
<title>A Linearized Temperature Measurement System Based on the B-mode of SC-cut Quartz Crystal</title>
<description><p><span><big>A Linearized Temperature Measurement System Based on the B-mode of SC-cut Quartz Crystal</big></span><br></p><p><span><small><i>Zhiqi Li; Haipeng Zhai; Miao Miao; Wei Zhou</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3374763</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This article demonstrates a linearized temperature measurement system based on the B-mode of SC-cut quartz crystal, which has high precision and good linearity. The proposed system uses a digital linear phase comparison measurement method by selecting phase detection region, which achieves high resolution, and fast response for temperature measurement. Extensive error analysis and performance evaluation of the system were conducted in this study. The developed linearized temperature measurement system demonstrated excellent performance, including: wide temperature measuring range (-10°C to 80°C); providing sufficient resolution and extremely fast response time with a resolution of 0.01°C at 10 μs, a resolution of 5E-5°C at 10 ms, and a resolution of 1E-7°C at 1 s when the temperature keeps stable. It achieves extremely low temperature measurement relative error (R.E.&lt; 0.24%) and nonlinearity (N.L.&lt; 0.18%). The measurement results show that the maximum estimate of the Allan variance is 2.3E-4°C. The proposed system has wide prospects for intelligent applications and can play a crucial role in scenarios that require high-precision and fast-response temperature measurements.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472879/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472879/</guid>
</item>
<item>
<title>Mesoporous silica modified by poly(ionic liquid)s for low humidity sensing</title>
<description><p><span><big>Mesoporous silica modified by poly(ionic liquid)s for low humidity sensing</big></span><br></p><p><span><small><i>Zhiyan Ma; Yaping Song; Hongran Zhao; Sen Liu; Xi Yang; Teng Fei; Tong Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373030</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this study, imidazole-based poly(ionic liquid)s are introduced into acid-treated mesoporous silica by solid grinding and subsequent thermal polymerization methods. The purpose of treating the mesoporous silica with hydrochloric acid is to increase the number of hydroxyl groups on the surface to fix more poly(ionic liquid)s. Due to the synergism of enhanced hydrophilicity and the mesoporous structure, the synthesized composite materials exhibited excellent capacities to adsorb and transport water vapor molecules and are suitable for low humidity sensing applications. In the 7-33% relative humidity (RH) range, the response of the as-prepared sensor reached 806% with a short response/recovery time of 8/13 s.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466490/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466490/</guid>
</item>
<item>
<title>PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards</title>
<description><p><span><big>PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards</big></span><br></p><p><span><small><i>Elias Arbash; Margret Fuchs; Behnood Rasti; Sandra Lorenz; Pedram Ghamisi; Richard Gloaguen</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3380826</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Addressing the critical theme of recycling electronic waste (E-waste), this contribution is dedicated to developing advanced automated data processing pipelines as a basis for decision-making and process control. Aligning with the broader goals of the circular economy and the United Nations (UN) Sustainable Development Goals (SDG), our work leverages non-invasive analysis methods utilizing RGB and hyperspectral imaging data to provide both quantitative and qualitative insights into the E-waste stream composition for optimizing recycling efficiency. In this paper, we introduce ’PCB-Vision’; a pioneering RGB-hyperspectral printed circuit board (PCB) benchmark dataset, comprising 53 RGB images of high spatial resolution paired with their corresponding high spectral resolution hyperspectral data cubes in the visible and near-infrared (VNIR) range. Grounded in open science principles, our dataset provides a comprehensive resource for researchers through high-quality ground truths, focusing on three primary PCB components: integrated circuits (IC), capacitors, and connectors. We provide extensive statistical investigations on the proposed dataset together with the performance of several state-of-the-art (SOTA) models, including U-Net, Attention UNet, Residual U-Net, LinkNet, and DeepLabv3+. By openly sharing this multiscene benchmark dataset along with the baseline codes, we hope to foster transparent, traceable, and comparable developments of advanced data processing across various scientific communities, including, but not limited to, computer vision and remote sensing. Emphasizing our commitment to supporting a collaborative and inclusive scientific community, all materials, including code, data, ground truth, and masks, will be accessible at https://github.com/hifexplo/PCBVision.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10485259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10485259/</guid>
</item>
<item>
<title>DEEP-CARDIO: Recommendation System for Cardiovascular Disease Prediction using IOT Network</title>
<description><p><span><big>DEEP-CARDIO: Recommendation System for Cardiovascular Disease Prediction using IOT Network</big></span><br></p><p><span><small><i>A. Yashudas; Dinesh Gupta; G. C. Prashant; Amit Dua; Dokhyl AlQahtani; A. Siva Krishna Reddy</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373429</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Internet of Things (IoT) based remote healthcare applications provide fast and preventative medical services to the patients at risk. However, predicting heart disease is a complex task and diagnosis results are rarely accurate. To address this issue, a novel Recommendation System for Cardiovascular Disease Prediction Using IoT Network (DEEP-CARDIO) has been proposed for providing prior diagnosis, treatment, and dietary recommendations for cardiac diseases. Initially, the physiological data are collected from the patient’s remotely by using the four bio sensors such as ECG sensor, Pressure sensor, Pulse sensor and Glucose sensor. An Arduino controller receives the collected data from the IoT sensors to predict and diagnose the disease. A cardiovascular disease prediction model is implemented by using BiGRU (Bidirectional-Gated Recurrent Unit) attention model which diagnose the cardiovascular disease and classify into five available cardiovascular classes. The recommendation system provides physical and dietary recommendations to cardiac patients based on the classified data, via user mobile application. The performance of the DEEP-CARDIO is validated by Cloud Simulator (CloudSim) using the real-time Framingham’s and Statlog heart disease dataset. The proposed DEEP CARDIO method achieves an overall accuracy of 99.90% whereas, the MABC-SVM, HCBDA and MLbPM method achieves 86.91%, 88.65% and 93.63% respectively.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472883/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472883/</guid>
</item>
<item>
<title>AuNP-decorated textile as chemo resistive sensor for acetone detection</title>
<description><p><span><big>AuNP-decorated textile as chemo resistive sensor for acetone detection</big></span><br></p><p><span><small><i>S. Casalinuovo; D. Caschera; S. Quaranta; D. Caputo</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2023.3348693</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This paper presents the development of chemo resistive sensors for the detection of volatile organic compounds (VOCs). The proposed sensor is based on citrate-functionalized gold nanoparticles (AuNPs) serving as a sensitive layer deposited on cotton fabric. Impedance variations due to VOC/substrate interaction are used as a detection principle. Specifically, this work focuses on acetone detection after exposing the AuNP-decorated cotton to a CH3COCH3 aqueous solution. Such an interaction resulted in a reduction of the total impedance (i.e., magnitude) of the system. This behavior can be ascribed to Van der Waals forces existing between the C=O group and the citrate moieties adsorbed on the gold nanoparticles, which favor charge injection to the substrate. Response to water was also tested for comparison, assuring that the solvent interacts with the sensitive layer by a different adsorption mechanism, not influencing the overall results. Sensor selectivity was also verified by considering ethanol (representative of alcohol group). Indeed, impedance curves reflect the different type of chemical interaction between the analyte and the substrate. In addition, sensor limit of detection for acetone was found to be 1% v/v, in the considered frequency range. Furthermore, sensor performance in terms of reusability was evaluated, showing that the Au-cotton ability in VOCs detection could be restored after about 90 min with a percentage up to 97 % in the frequency of 1Hz. These results can be considered the starting point for the development of portable, sensitive and user-friendly devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10384362/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10384362/</guid>
</item>
<item>
<title>L-cysteine modified gold nanoparticles for copper (II) ion detection using etched Fiber Bragg Grating sensor</title>
<description><p><span><big>L-cysteine modified gold nanoparticles for copper (II) ion detection using etched Fiber Bragg Grating sensor</big></span><br></p><p><span><small><i>S Srivatzen; B S Kavitha; Asha Prasad; S Asokan</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3374514</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>A highly sensitive and selective optical detection of copper (II) (Cu
2+
) concentration in water has been proposed using L-cysteine functionalized gold nanoparticles (AuNPs) using an etched Fiber Bragg grating sensor (eFBG). The eFBG sensor produces a Bragg wavelength shift (ΔλB) in correspondence to the added Cu
2+
concentration. The amino acid L-Cysteine facilitates the formation of core-satellite nanoassemblies of gold in the presence of Cu
2+
. A linear wavelength shift is observed in accordance with the mediating Cu
2+
concentration. The sensor’s detection limit is found to be 1picomolar (pM) which is well within the world health organization (WHO) acceptable limit for safe drinking water and the dynamic range of the sensor lies between 1pM to 100 μM. The sensor is highly sensitive and selective to Cu
2+
with sensitivity of 230 pm 10
-1
M. The sensor has proven its capability as an ideal probe for onsite, real-time measurement of Cu
2+
in drinking water by giving a recovery percentage of &gt;90% in real water sample additional to good reproducibility (standard error&lt; 20 pm), high sensitivity, portability and specificity.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472453/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472453/</guid>
</item>
<item>
<title>AcHand: Detecting Respiratory Rate of operator via Smartphone Microphone</title>
<description><p><span><big>AcHand: Detecting Respiratory Rate of operator via Smartphone Microphone</big></span><br></p><p><span><small><i>KounKou Vincent; Lin Wang; Nan Jing; Wenyuan Liu</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3372255</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this paper, we dedicate a handgrip-free technique, called ”AcHand”, to continuously detect respiratory rates under acoustic signals sensing via a smartphone. Although the particularity of the AcHand is to follow the subtle movements of the heart during its displacement, we acknowledge specific challenges that can make the signal vulnerable to handgripfree scenarios such as 1) stochastic hand activities: fluctuations in movement can alter the trajectory of the acoustic signal to the microphones and cause random changes in the breathing curve, resulting in complex and unpredictable dynamics, and 2) clicking on the phone screen can disrupt the breathing by generating a large signal variation to create undesirable peaks-to-noisy the signal. However, by analyzing the recorded signal reflection evoked by a chirp sound stimulus, we propose a linear transformation of the Cardioreflector-Manoreflector to analyze the respiratory patterns variation and distinguish weak breathing signals under the handgrip-free. We compare the present signal with the ground truth recorded by ECG to approve the efficiency of the proposed method. Our experiments show AcHand achieves a MAE of 0.341 bpm, which is a significant improvement over the state-of-the-art devices, and enhancing detection capability across handgrip-free scenarios.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10462909/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10462909/</guid>
</item>
<item>
<title>Optimization Method for Node Deployment of Closed-Barrier Coverage in Hybrid Directional Sensor Networks</title>
<description><p><span><big>Optimization Method for Node Deployment of Closed-Barrier Coverage in Hybrid Directional Sensor Networks</big></span><br></p><p><span><small><i>Peng Wang; Yonghua Xiong; Jinhua She; Anjun Yu</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3378998</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Closed-barrier coverage is a new coverage problem that requires sensors on the barrier to form a closed loop. Due to the features of end-to-end connection, constructing closed barriers is a challenging task. In a Hybrid Directional Sensor Network, sensors can both rotate and move, making it more difficult to schedule sensors to construct closed barriers. To address this challenge, we propose an optimization method for node deployment of closed-barrier coverage (CCOM). First, selecting multiple initiators transforms the closed-barrier coverage problem into a simple linear-barrier construction problem. Then, a weighted barrier graph is used to search the optimal barrier path. Finally, design a rotation strategy and moving plan for the sensors, fully utilizing their rotation and moving capabilities to achieve closed-barrier coverage with the minimum number of sensors and energy consumption. The simulation experiment results show that compared with other advanced methods, our proposed method has better performance in terms of barrier construction success rate, minimum required number of mobile sensors, moving distance, and network lifetime.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10478814/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10478814/</guid>
</item>
<item>
<title>Enhanced Multicast Protocol for Low-power and Lossy IoT Networks</title>
<description><p><span><big>Enhanced Multicast Protocol for Low-power and Lossy IoT Networks</big></span><br></p><p><span><small><i>Issam Eddine Lakhlef; Badis Djamaa; Mustapha Reda Senouci; Abbas Bradai</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3375797</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Communication protocols in the Internet of Things (IoT) should take into account the resource-constrained nature of Low-power and Lossy Networks (LLNs). IP multicast protocols allow a packet to be routed from one source to multiple destinations in a single transmission. Hence, resources such as bandwidth, energy, and time are saved for a multitude of LLN applications, ranging from over-the-air programming and information sharing to device configuration and resource discovery. In this context, several multicast routing protocols have recently been proposed for LLNs, including the Multicast Protocol for LLNs (MPL). MPL has proven to be very reliable in many scenarios. However, the great resource consumption, especially in terms of energy and bandwidth, remains the main drawback of this protocol. In this paper, after a detailed overview, we provide an in-depth analysis of the MPL protocol to highlight its functional weaknesses. Then, we propose several improvements that touch on different areas to address such limitations. Extensive realistic simulations and experiments were performed to study the performance of the proposed improvements to MPL. The results obtained show that our proposals outperform the MPL protocol in terms of resource consumption (memory, bandwidth, and energy) while improving its performance in terms of end-to-end delay and maintaining the same reliability of data packet delivery.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10473686/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10473686/</guid>
</item>
<item>
<title>A 7.25μK Ultra-High Resolution MEMS Resonant Thermometer</title>
<description><p><span><big>A 7.25μK Ultra-High Resolution MEMS Resonant Thermometer</big></span><br></p><p><span><small><i>Zheng Wang; Liangbo Ma; Xiaorui Bie; Xingyin Xiong; Zhaoyang Zhai; Wuhao Yang; Yongjian Lu; Xudong Zou</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3370956</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This paper proposes an ultra-high-resolution MEMS resonant thermometer that exploits differences in Young’s modulus and coefficient of thermal expansion (CTE) between structural layers to increase thermal stress due to temperature variation. The increased thermal stress eventually acts on the resonator to produce a large frequency shift, and the temperature coefficient of frequency (TCF) can be increased to 17.4 times the original, which is consistent with theoretical analysis and finite element method simulations. The MEMS resonator chip is manufactured by the standard Silicon-on-insulator process and stacked on a ceramic chip carrier through multiple layers of materials. A self-sustaining oscillator, mainly composed of a packaged MEMS resonator chip and a low-noise application-specific integrated circuit (ASIC), is built to track the resonant frequency shift of the resonator. This MEMS resonant thermometer prototype demonstrates a high temperature coefficient of frequency (TCF) of -866.84ppm/K from -50°C to 40°C and exhibits good linearity. An ultra-high resolution of 7.25μK is achieved in the closed-loop experimental test. This is the best result achieved for a MEMS thermometer employing the resonant sensing paradigm to date.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10460450/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10460450/</guid>
</item>
<item>
<title>Design of MTM-based Multi-band Micro-Biosensor in Terahertz region as perfect absorber for Early-Stage Leukemia Diagnosis with sensitivity 18626373 TH...</title>
<description><p><span><big>Design of MTM-based Multi-band Micro-Biosensor in Terahertz region as perfect absorber for Early-Stage Leukemia Diagnosis with sensitivity 18626373 THz/RIU</big></span><br></p><p><span><small><i>Musa N. Hamza; Mohammad Tariqul Islam</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383522</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this article, a novel highly sensitive biosensor based on perfect metamaterial absorbers is presented. A detailed study is presented for several different models, different types of substrate materials, different types of resonant materials and substrate thicknesses showing very good sensitivity to any changes. The proposed biosensor exhibits high sensitivity to different polarization and incidence angles, ensuring its performance in terms of signal-to-noise ratio. The proposed biosensor was carefully compared with previously designed sensors and biosensors. The proposed biosensor exhibits amazing sensitivity, such as a quality factor of 41.1, a figure of merit (FOM) of 172466416 RIU-1, and a sensitivity (S) of 18626373 THz/RIU. The high sensitivity of the biosensor allows early detection of leukemia, demonstrating significant differences between leukemia and normal blood. Another interesting result of this article is the use of terahertz waves for imaging. Microwave imaging is performed for electric fields, magnetic fields, and power.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10494209/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10494209/</guid>
</item>
<item>
<title>LEPS: A lightweight and effective single-stage detector for pothole segmentation</title>
<description><p><span><big>LEPS: A lightweight and effective single-stage detector for pothole segmentation</big></span><br></p><p><span><small><i>Xiaoning Huang; Jintao Cheng; Qiuchi Xiang; Jun Dong; Jin Wu; Rui Fan; Xiaoyu Tang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2023.3330335</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Currently, the problem of potholes on urban roads is becoming increasingly severe. Identifying and locating road potholes promptly has become a major challenge for urban construction. Therefore, we proposed a lightweight and effective instance segmentation method called LEPS (Lightweight and Effective Pothole Segmentation Detector) for road pothole detection. To extract the image edge information and gradient information of potholes from the feature map more efficiently, we proposed a module that performs the convolutional super-position supplemented with a convolutional kernel to enhance spatial details for the backbone. We have designed a novel module applied to the Neck layer, improving the detection performance while reducing the parameters. To enable accurate segmentation of fine-grained features, we optimized the ProtoNet, which enables the segmentation head to generate high-quality masks for more accurate prediction. We have fully demonstrated the effectiveness of the method through a large number of comparative experiments. Our detector has excellent performance on two authoritative example datasets, URTIRI and COCO, and can successfully be applied to visual sensors to accurately detect and segment road potholes in real environments, accurately LEPS reached 0.892 and 0.648 in terms of Mask for AP50 and AP50:95, which is improved by 4.6% and 20.6% compared to the original model. These results demonstrate its strong competitiveness when compared to other models. Comprehensively, LEPS improves detection accuracy while maintaining lightweight, which allows the model to meet the practical application requirements of edge computing devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10315064/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10315064/</guid>
</item>
<item>
<title>UMHE: Unsupervised Multispectral Homography Estimation</title>
<description><p><span><big>UMHE: Unsupervised Multispectral Homography Estimation</big></span><br></p><p><span><small><i>Jeongmin Shin; Jiwon Kim; Seokjun Kwon; Namil Kim; Soonmin Hwang; Yukyung Choi</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383453</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Multispectral image alignment plays a crucial role in exploiting complementary information between different spectral images. Homography-based image alignment can be a practical solution considering a tradeoff between runtime and accuracy. Existing methods, however, have difficulty with multispectral images due to the additional spectral gap or require expensive human labels to train models. To solve these problems, this paper presents a comprehensive study on multispectral homography estimation in an unsupervised learning manner. We propose a curriculum data augmentation, an effective solution for models learning spectrum-agnostic representation by providing diverse input pairs. We also propose to use the phase congruency loss that explicitly calculates the reconstruction between images based on low-level structural information in the frequency domain. To encourage multispectral alignment research, we introduce a novel FLIR corresponding dataset that has manually labeled local correspondences between multispectral images. Our model achieves state-of-the-art alignment performance on the proposed FLIR correspondence dataset among supervised and unsupervised methods while running at
151 FPS
. Furthermore, our model shows good generalization ability on the M3FD dataset without finetuning.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10494213/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10494213/</guid>
</item>
<item>
<title>Calibration and Evaluation of a Low-Cost Optical Particulate Matter Sensor for Measurement of Lofted Lunar Dust</title>
<description><p><span><big>Calibration and Evaluation of a Low-Cost Optical Particulate Matter Sensor for Measurement of Lofted Lunar Dust</big></span><br></p><p><span><small><i>Abhay Vidwans; Jeffrey Gillis-Davis; Pratim Biswas</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3366436</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Several recent Earth-based investigations employed low-cost particulate matter sensors to address the lack of spatiotemporal resolution in air quality data. The lunar environment also has a particulate matter problem in the form of fine lofted and levitated dust particles. Natural and anthropogenic mobilized dust can cause a slew of difficulties for surface operations (deposition onto radiators, optical components, and mechanical devices). Despite the urgency of mitigating dust on the Moon and other airless bodies, the performance of low-cost sensors has not been critically evaluated for space applications. Upcoming long-term robotic and human exploration missions to the Moon necessitate a robust sensor that can monitor particulate matter levels and establish a spatially and temporally resolved global network. In this work, we calibrate two optical light-scattering particulate matter sensors against research-grade aerosol instruments for measuring the concentration of aerosolized lunar simulants. Sensors showed a stronger dependence on aerosol particle size distribution than particle composition. Vacuum testing showed a significant deviation in performance compared to atmospheric pressure, with a stronger dependency on lunar simulant. The predicted mass deposition, based on sensor output coupled with dust trajectory, was within an order of magnitude of the reference deposition. Our results demonstrate for the first time that low-cost particulate matter sensors can monitor dust concentrations with reasonable accuracy in a vacuum environment, with two caveats. First, precise calibrations must be performed with a dust simulant closely matching the particle size distribution of the target dust, and second, atmospheric pressure calibrations alone are insufficient.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10462021/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10462021/</guid>
</item>
<item>
<title>Reweighting Interacting Multiple-model Algorithm to Overcome Model Competition for Target Tracking in the Hybrid System</title>
<description><p><span><big>Reweighting Interacting Multiple-model Algorithm to Overcome Model Competition for Target Tracking in the Hybrid System</big></span><br></p><p><span><small><i>Guowei Li; Shurui Zhang; Yubing Han; Weixing Sheng; Thia Kirubarajan</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3369854</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The vicious competition of interacting multiple-model algorithm (IMM) is an inherent problem and would produce irreversible effects on IMM estimation results, especially combining with the radar system. In this paper, a novel reweighting IMM (RIMM) is proposed to overcome this issue. Firstly, the theoretical lower bound of model numbers in different situations is respectively provided through the analysis of IMM limitations. Furthermore, certificate the influence of model inaccuracy on the Kalman filter, which illustrates an effective method for reducing errors is increasing model numbers. Thirdly, the definition of model set density and the analysis of the true model space are given, and their connection establishes the standard of how to design the model set or add the model number. Finally, an effective method called RIMM is provided to overcome the competition caused by model increasing. The proposed RIMM holds strong adaptability for different model sets. The simulations of RIMM highlight the correctness and effectiveness of the proposed methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10462948/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10462948/</guid>
</item>
<item>
<title>A Stretchable and Wearable Ultrasonic Transducer Array for Bladder Volume Monitoring Application</title>
<description><p><span><big>A Stretchable and Wearable Ultrasonic Transducer Array for Bladder Volume Monitoring Application</big></span><br></p><p><span><small><i>Cong Pu; Ben Fu; Lehang Guo; Huixiong Xu; Chang Peng</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3382244</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Monitoring bladder volume is an essential application for patients with voiding dysfunction. Current wearable ultrasonic bladder monitors are rigid, bulky and cannot conform to human skin. This study presents a stretchable and wearable ultrasonic transducer array that can both conform to non-planar skin surfaces and continuously monitor bladder volume. The wearable transducer array mainly consists of 4 × 4 piezoelectric transducer elements, stretchable serpentine-based electrodes for electrical interconnection, electrically conductive adhesive as both matching and backing layers, and silicone elastomer for encapsulation. The transducer array has a center frequency of 3.9 MHz, -6 dB acoustic bandwidth of 57.6%, and up to 40% elastic stretchability. The volume of balloon-bladder phantom was estimated by a least square ellipsoid fitting method. With the true volume of balloon-bladder phantoms ranging from 10 mL to 405 mL, the proposed transducer array has a mean absolute percentage error of 9.4%, a mean absolute error of 11.9 mL and a coefficient of determination of 0.975, which demonstrates the capability of the proposed stretchable and wearable ultrasonic transducer array for bladder volume monitoring application.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10489841/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10489841/</guid>
</item>
<item>
<title>Intelligent fault diagnosis of bearing using multiwavelet perception kernel convolutional neural network</title>
<description><p><span><big>Intelligent fault diagnosis of bearing using multiwavelet perception kernel convolutional neural network</big></span><br></p><p><span><small><i>Yuanyuan Zhou; Hang Wang; Yongbin Liu; Xianzeng Liu; Zheng Cao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3370564</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Strong background noise characteristics of vibration signals cause issues with poor identification capability of features by fault diagnostic models. To address this issue, a method is proposed for intelligent fault diagnosis of bearing using multiwavelet perception kernel (MPK) and feature attention convolutional neural network (FA-CNN). First, four multiwavelet perception kernels are constructed to decompose the vibration signals in full-band multilevel. Second, improved multiwavelet information entropy (IMIE) of the frequency band components is calculated. The calculated component entropies of the corresponding frequency bands are integrated to construct frequency band clusters (FBC) from low to high frequencies. Third, joint approximate diagonalization eigen (JADE) is introduced to perform feature fusion for every FBC to eliminate redundant information, and fused features from low to high frequencies are obtained as original inputs. The FA-CNN bearing fault diagnosis framework is constructed for intelligent fault diagnosis of bearings. Finally, the effectiveness of the proposed method is verified by two cases. The results show that the proposed method has high fault feature recognition capability.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10458914/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10458914/</guid>
</item>
<item>
<title>Improved multipath mitigation using multiple trend-surface hemispherical map in GNSS precise point positioning</title>
<description><p><span><big>Improved multipath mitigation using multiple trend-surface hemispherical map in GNSS precise point positioning</big></span><br></p><p><span><small><i>Haijun Yuan; Zhetao Zhang; Xiufeng He; Jinwen Zeng; Hao Wang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3374458</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Global Navigation Satellite System (GNSS) Precise Point Positioning (PPP) is highly appreciated as a positioning technology; however, multipath restricts its accuracy and reliability. In this study, we proposed a Multiple Trend-surface Multipath Hemispherical Map (MT-MHM), where discrepant multipath fading frequency or orbit orientation of each satellite in each azimuth and elevation grid is considered. In a typical static scenario, consecutive 8-day observations are collected to extract multipath and conduct PPP experiments. Compared with traditional multipath hemispherical map based on trend-surface fitting, MT-MHM has better modelling performance of residuals and improves standard deviations of residuals for all satellites. Besides, the positioning errors and required convergence epochs of PPP are reduced by using MT-MHM. For the positioning accuracy of all solutions, MT-MHM exhibits 50.8, 15.2, and 27.9% improvements compared with that without multipath correction in the east, north, and up directions, respectively. In conclusion, our proposed MT-MHM exhibits better performance in terms of residual reduction, convergence time, and positioning accuracy in PPP.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472420/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472420/</guid>
</item>
<item>
<title>Design optimization and characterization of a 3D-printed tactile sensor for tissue palpation</title>
<description><p><span><big>Design optimization and characterization of a 3D-printed tactile sensor for tissue palpation</big></span><br></p><p><span><small><i>D. Lo Presti; L. Zoboli; A. Addabbo; D. Bianchi; A. Dimo; C. Massaroni; V. Altomare; A. Grasso; A. Gizzi; E. Schena</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3369337</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Manual palpation is a crucial medical procedure that relies on surface examination to detect internal tissue abnormalities, heavily reliant on healthcare professionals’ expertise and tactile sensitivity. To tackle these issues, smart palpation systems based on electrical or optical sensors have been developed to perform quantitative tactile measurements, crucial for identifying various solid tumors, including breast and prostate cancer by assessing tissue mechanical properties. In this context, fiber Bragg gratings (FBGs) are emerging as a promising technology due to their advantages (e.g., high metrological properties, multiplexing capacity, and easy packaging) making them ideal for tactile sensing. This study explores the benefits of FBG and 3D printing to develop a tactile sensor for tissue palpation. First, an optimization of the design of the sensing core of a previously developed probe was conducted through finite element analysis. The novel structure addresses the primary limitation of the previous solution, where non-uniform strain distribution on the encapsulated FBG causes compression on the grating with high risk of bending and breakage. In contrast, the modeled geometry ensures FBG elongation during tissue palpation. A 3D printing and characterization of the proposed solution was carried out to investigate the response of the enclosed FBG when pushed against different materials showing promising results in discriminating tissues according to their mechanical properties: the more rigid the indented substrate the higher the sensor output. This property will be fundamental for enhancing early tumor detection through superficial tissue palpation, advancing the efficacy of prevention measures.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10455966/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10455966/</guid>
</item>
<item>
<title>Deep Reinforcement Learning-based Joint Sequence Scheduling and Trajectory Planning in Wireless Rechargeable Sensor Networks</title>
<description><p><span><big>Deep Reinforcement Learning-based Joint Sequence Scheduling and Trajectory Planning in Wireless Rechargeable Sensor Networks</big></span><br></p><p><span><small><i>Chengpeng Jiang; Wencong Chen; Ziyang Wang; Wendong Xiao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373664</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Mobile charging has become a popular and efficient method for replenishing energy. This is done with a mobile charger (MC) and wireless energy transfer technology (WET), which helps to alleviate the issue of energy constraints in wireless rechargeable sensor networks (WRSNs). Notably, designing mobile charging scheduling schemes is essential for improving charging performance. Most current studies assume that the networks are obstacle-free. Unlike the existing studies, this paper focuses on joint sequence scheduling and trajectory planning problems (JSSTP), which assumes that the network has multiple static obstacles. To address this issue, we propose a novel deep reinforcement learning-based JSSTP (DRL-JSSTP) that enables the MC to avoid obstacles and reach the charging target to charge the sensors fully. This approach maximizes energy usage efficiency and sensor survival rate while satisfying MC energy capacity constraints. DRL-JSSTP includes a charging target selector and a trajectory planner, which determine the index of the next charging target and plan the movement trajectory to avoid obstacles, respectively. We adopt 1-D convolutional neural networks to extract feature information about the environment state and gated recurrent units to predict the charging decisions. Simulation results demonstrate that DRL-JSSTP outperforms existing approaches, achieving higher energy usage efficiency and sensor survival rate.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466524/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466524/</guid>
</item>
<item>
<title>A Simultaneous Wirel |
|
Successfully generated as following: http://localhost:1200/ieee/journal/8782710 - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Open Journal of Signal Processing</title>
<link>https://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=8782710</link>
<atom:link href="http://localhost:1200/ieee/journal/8782710" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Open Journal of Signal Processing - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Wed, 10 Apr 2024 14:39:57 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Synthbuster: Towards Detection of Diffusion Model Generated Images</title>
<description><p><span><big>Synthbuster: Towards Detection of Diffusion Model Generated Images</big></span><br></p><p><span><small><i>Quentin Bammey</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337714</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Synthetically-generated images are getting increasingly popular. Diffusion models have advanced to the stage where even non-experts can generate photo-realistic images from a simple text prompt. They expand creative horizons but also open a Pandora's box of potential disinformation risks. In this context, the present corpus of synthetic image detection techniques, primarily focusing on older generative models like Generative Adversarial Networks, finds itself ill-equipped to deal with this emerging trend. Recognizing this challenge, we introduce a method specifically designed to detect synthetic images produced by diffusion models. Our approach capitalizes on the inherent frequency artefacts left behind during the diffusion process. Spectral analysis is used to highlight the artefacts in the Fourier transform of a residual image, which are used to distinguish real from fake images. The proposed method can detect diffusion-model-generated images even under mild
jpeg
compression, and generalizes relatively well to unknown models. By pioneering this novel approach, we aim to fortify forensic methodologies and ignite further research into the detection of AI-generated images.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334046/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334046/</guid>
</item>
<item>
<title>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</title>
<description><p><span><big>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</big></span><br></p><p><span><small><i>Yanbin Zou; Jingna Fan; Zekai Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we consider using time-of-arrival (TOA) measurements from a single moving receiver to locate a moving target at constant velocity that emits a periodic signal with unknown signal period. First, we give the TOA measurement model and deduce the Cram
$\acute{\text{e}}$
r-Rao lower bounds (CRLB). Then, we formulate a nonlinear least squares (NLS) problem to estimate the unknown parameters. We use semidefinite programming (SDP) techniques to relax the nonconvex NLS problem. However, it is shown that the SDP localization algorithm cannot provide a high-quality solution. Subsequently, we develop a fixed point iteration (FPI) method to improve the performance of the SDP algorithm. In addition, we also consider the presence of receiver position errors and develop the corresponding localization algorithm. Numerical simulations are conducted to demonstrate the localization performance of the proposed algorithms by comparing them with the CRLB.
Index Term
-Fixed point iteration (FPI), semidefinite programming (SDP), single moving receiver, target localization, time-of-arrival (TOA).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336384/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336384/</guid>
</item>
<item>
<title>The Neural-SRP Method for Universal Robust Multi-Source Tracking</title>
<description><p><span><big>The Neural-SRP Method for Universal Robust Multi-Source Tracking</big></span><br></p><p><span><small><i>Eric Grinstein; Christopher M. Hicks; Toon van Waterschoot; Mike Brookes; Patrick A. Naylor</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340057</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Neural networks have achieved state-of-the-art performance on the task of acoustic Direction-of-Arrival (DOA) estimation using microphone arrays. Neural models can be classified as end-to-end or hybrid, each class showing advantages and disadvantages. This work introduces Neural-SRP, an end-to-end neural network architecture for DOA estimation inspired by the classical Steered Response Power (SRP) method, which overcomes limitations of current neural models. We evaluate the architecture on multiple scenarios, namely, multi-source DOA tracking and single-source DOA tracking under the presence of directional and diffuse noise. The experiments demonstrate that our proposed method compares favourably in terms of computational and localization performance with established neural methods on various recorded and simulated benchmark datasets.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345765/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345765/</guid>
</item>
<item>
<title>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</title>
<description><p><span><big>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</big></span><br></p><p><span><small><i>Alexander Venus; Erik Leitinger; Stefan Tertinek; Klaus Witrisal</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper presents a neural-enhanced probabilistic model and corresponding factor graph-based sum-product algorithm for robust localization and tracking in multipath-prone environments. The introduced hybrid probabilistic model consists of physics-based and data-driven measurement models capturing the information contained in both, the line-of-sight (LOS) component as well as in multipath components (NLOS components). The physics-based and data-driven models are embedded in a joint Bayesian framework allowing to derive from first principles a factor graph-based algorithm that fuses the information of these models. The proposed algorithm uses radio signal measurements from multiple base stations to robustly estimate the mobile agent's position together with all model parameters. It provides high localization accuracy by exploiting the position-related information of the LOS component via the physics-based model and robustness by exploiting the geometric imprint of multipath components independent of the propagation channel via the data-driven model. In a challenging numerical experiment involving obstructed LOS situations to all anchors, we show that the proposed sequential algorithm significantly outperforms state-of-the-art methods and attains the posterior Cramér-Rao lower bound even with training data limited to local regions.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336409/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336409/</guid>
</item>
<item>
<title>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</title>
<description><p><span><big>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</big></span><br></p><p><span><small><i>AMIT EFRAIM; Joseph M. Francos</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In solving a model fitting problem, the existence of outliers in the set of measurements can have a devastating effect on the solution accuracy. Traditionally, in order to overcome this problem, robust point cloud registration algorithms are composed of transformation hypothesis generation, followed by hypothesis evaluation aimed at selecting the best hypothesized estimate. Hypotheses evaluation is commonly performed using the sample consensus criterion. However, since this criterion accounts only for the consensus size, it fails when the maximal sample consensus is incorrect. We propose a new hypothesis evaluation approach, generalizing the sample consensus approach, where instead of the sample consensus, the transformation that maximizes the point clouds feature correlation is selected as the best hypothesis. The feature vector at each point contains information such as on local geometry and semantic context. Utilizing this information in the hypotheses evaluation and selection procedure allows for a correct decision even when the hypothesis yielding the maximal sample consensus is false. Consequently, the probability of selecting the correct model increases. We show both mathematically and empirically that substituting the sample consensus criterion with the proposed point cloud feature correlation hypothesis test (PC-FCHT) lowers the probability of large registration errors, compared to using the special case of sample consensus. The proposed PC-FCHT is applicable to any algorithm that follows the hypothesis generation and evaluation scheme, potentially improving the success rates of a wide family of point cloud registration algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345750/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345750/</guid>
</item>
<item>
<title>Joint PAPR and OBP Reduction for NC-OFDM Systems</title>
<description><p><span><big>Joint PAPR and OBP Reduction for NC-OFDM Systems</big></span><br></p><p><span><small><i>Hsuan-Fu Wang; Fang-Biau Ueng; Bo-Heng Yeh</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3329757</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The spectrum resource is always a critical issue for wireless communications since it directly impacts the data rate and capacity. However, the problem of spectrum resource scarcity always exists. Moreover, spectrum resource scarcity becomes more severe as new communication technologies and wireless applications sprout. Noncontiguous orthogonal frequency division multiplexing (NC-OFDM) is a multicarrier method for bandwidth utilization. Unfortunately, this system has two fatal defects: high peak-to-average power ratio (PAPR) and considerable out-of-band power (OBP), which are detrimental to the system's performance. To solve these two problems, we propose a convex optimization-based method for joint PAPR and OBP reduction in NC-OFDM Systems. The strategy is to permit the secondary user to utilize the unoccupied spectrum of the primary user with dynamic spectrum sharing (DSS) based on a cognitive radio network (CRN). To this end, a flexible system operating over noncontiguous bands and DSS scenarios is necessary. The simulation results have shown that our method could effectively improve the overall performance and outperform other schemes, i.e., projections onto convex sets (POCS) and alternating projections onto convex and non-convex sets (APOCNCS), without harming the transmission of the primary system. The collaboration between secondary and primary systems is viable with the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10305259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10305259/</guid>
</item>
<item>
<title>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</title>
<description><p><span><big>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</big></span><br></p><p><span><small><i>Gal Itzhak; Israel Cohen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339433</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article presents a Kronecker-product (KP) beamforming approach incorporating sparse concentric circular arrays (SCCAs). The locations of the microphones on the SCCA are optimized concerning the broadband array directivity over a wide range of direction-of-arrival (DOA) deviations of a desired signal. A maximum directivity factor (MDF) sub-beamformer is derived accordingly with the optimal locations. Then, we propose two global beamformers obtained as a Kronecker product of a uniform linear array (ULA) and the SCCA sub-beamformer. The global beamformers differ by the type of the ULA, which is designed either as an MDF sub-beamformer along the
$\mathsf {x}$
-axis or as a maximum white noise gain sub-beamformer along the
$\mathsf {y}$
-axis. We analyze the performance of the proposed beamformers in terms of the directivity factor, the white noise gain, and their spatial beampatterns. Compared to traditional beamformers, the proposed beamformers exhibit considerably larger tolerance to DOA deviations concerning both the azimuth and elevation angles. Experimental results with speech signals in noisy and reverberant environments demonstrate that the proposed approach outperforms traditional beamformers regarding the perceptual evaluation of speech quality (PESQ) and short-time objective intelligibility (STOI) scores when the desired speech signals deviate from the nominal DOA.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342869/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342869/</guid>
</item>
<item>
<title>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</title>
<description><p><span><big>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</big></span><br></p><p><span><small><i>Karn N. Watcharasupat; Chih-Wei Wu; Yiwei Ding; Iroro Orife; Aaron J. Hipple; Phillip A. Williams; Scott Kramer; Alexander Lerch; William Wolcott</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339428</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Cinematic audio source separation is a relatively new subtask of audio source separation, with the aim of extracting the dialogue, music, and effects stems from their mixture. In this work, we developed a model generalizing the Bandsplit RNN for any complete or overcomplete partitions of the frequency axis. Psychoacoustically motivated frequency scales were used to inform the band definitions which are now defined with redundancy for more reliable feature extraction. A loss function motivated by the signal-to-noise ratio and the sparsity-promoting property of the 1-norm was proposed. We additionally exploit the information-sharing property of a common-encoder setup to reduce computational complexity during both training and inference, improve separation performance for hard-to-generalize classes of sounds, and allow flexibility during inference time with detachable decoders. Our best model sets the state of the art on the Divide and Remaster dataset with performance above the ideal ratio mask for the dialogue stem.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342812/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342812/</guid>
</item>
<item>
<title>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</title>
<description><p><span><big>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</big></span><br></p><p><span><small><i>Miguel Ferrer; María de Diego; Alberto Gonzalez</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340106</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The LMS algorithm is widely employed in adaptive systems due to its robustness, simplicity, and reasonable performance. However, it is well known that this algorithm suffers from a slow convergence speed when dealing with colored reference signals. Numerous variants and alternative algorithms have been proposed to address this issue, though all of them entail an increase in computational cost. Among the proposed alternatives, the affine projection algorithm stands out. This algorithm has the peculiarity of starting from
$N$
data vectors of the reference signal. It transforms these vectors into as many data vectors suitably normalized in energy and mutually orthogonal. In this work, we propose a version of the LMS algorithm that, similar to the affine projection algorithm, starts from
$N$
data vectors of the reference signal but corrects them by using only a scalar factor that functions as a convergence step. Our goal is to align the behavior of this algorithm with the behavior of the affine projection algorithm without significantly increasing the computational cost of the LMS.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345730/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345730/</guid>
</item>
<item>
<title>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</title>
<description><p><span><big>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</big></span><br></p><p><span><small><i>Zongyao Li; Ren Togo; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340616</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article proposes a method for transferring knowledge of semantic segmentation from a labeled source domain to an unlabeled target domain without using the source-domain data. Such a problem is called source-data-free domain adaptation, in which a pre-trained source-domain model and the unlabeled target-domain data are used to transfer the label knowledge across the domains. Like most previous methods, our method uses pseudo labels for distilling and transferring the source-domain knowledge. On the basis of the pseudo-label learning, our method improves the domain adaptation performance in two innovative ways: 1) reducing the domain differences by source-data-free style transfer and 2) exploring the style diversity within the target domain by style modification. To this end, we introduce two additional modules: 1) an inter-domain style transfer module which aligns the feature statistics of the source and target domains before producing the pseudo labels thereby improving the pseudo labels' accuracy, and 2) an intra-domain style modification module which modifies the image styles within the target domain for learning intra-domain style-invariant features. Our method with the two modules outperforms previous source-data-free domain adaptation methods in two commonly used benchmarks. Moreover, our method is well compatible with the previous methods for further improvement.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10356748/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10356748/</guid>
</item>
<item>
<title>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</title>
<description><p><span><big>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</big></span><br></p><p><span><small><i>Tobias Kabzinski; Peter Jax</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337721</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Quasi-continuous system identification of time-variant linear acoustic systems can be applied in various audio signal processing applications when numerous acoustic transfer functions must be measured. A prominent application is measuring head-related transfer functions. We treat the underlying multiple-input-multiple-output (MIMO) system identification problem in a state-space model as a joint estimation problem for states, representing impulse responses, and state-space model parameters using the expectation maximization (EM) algorithm. We address limitations of prior work by imposing different model structures, especially for dependencies within a (transformed) state vector. This results in block diagonal matrix structures, for which we derive M-step update rules. Making assumptions about this model structure and choosing a block size for a given application define the computational complexity. In examples, we found that applying this framework yields improvements of up to 10 dB in relative system distance in comparison to a conventional method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334061/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334061/</guid>
</item>
<item>
<title>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</title>
<description><p><span><big>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</big></span><br></p><p><span><small><i>Damir Rakhimov; Martin Haardt</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337729</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this paper, we present an analytical performance assessment of 2-D Tensor ESPRIT in terms of physical parameters. We show that the error in the
$r$
-mode depends only on two components, irrespective of the dimensionality of the problem. We obtain analytical expressions in closed form for the mean squared error (MSE) in each dimension as a function of the signal-to-noise (SNR) ratio, the array steering matrices, the number of antennas, the number of snapshots, the selection matrices, and the signal correlation. The derived expressions allow a better understanding of the difference in performance between the tensor and the matrix versions of the ESPRIT algorithm. The simulation results confirm the coincidence between the presented analytical expression and the curves obtained via Monte Carlo trials. We analyze the behavior of each of the two error components in different scenarios.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334446/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334446/</guid>
</item>
<item>
<title>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</title>
<description><p><span><big>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</big></span><br></p><p><span><small><i>Yaman Kındap; Simon Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343341</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this work we study linear vector stochastic differential equation (SDE) models driven by the generalised hyperbolic (GH) Lévy process for inference in continuous-time non-Gaussian filtering problems. The GH family of stochastic processes offers a flexible framework for modelling of non-Gaussian, heavy-tailed characteristics and includes the normal inverse-Gaussian, variance-gamma and Student-t processes as special cases. We present continuous-time simulation methods for the solution of vector SDE models driven by GH processes and novel inference methodologies using a variant of sequential Markov chain Monte Carlo (MCMC). As an example a particular formulation of Langevin dynamics is studied within this framework. The model is applied to both a synthetically generated data set and a real-world financial series to demonstrate its capabilities.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360268/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360268/</guid>
</item>
<item>
<title>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</title>
<description><p><span><big>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</big></span><br></p><p><span><small><i>James M. Cozens; Simon J. Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344048</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper proposes a probabilistic approach for extracting time-varying and irregular time signature information from polyphonic audio extracts, subsequently providing beat and bar line positions given inferred time signature divisions. This is achieved via dynamically evaluating the beat tempo as a function of time through finding an optimal compromise in beat and bar alignment in the time and tempo domains. Time signature divisions are determined based on a new representation, termed the Metrogram, that presents time-varying information regarding rhythmic and metric periodicities in the Tempogram. Our methodology is characterised by its ability to provide a distribution over metric interpretations, offering insights into the diverse ways music can be rhythmically perceived. Results indicate high-level accuracy for a variety of polyphonic extracts containing irregular, complex, irrational, and time-varying time signatures. Accuracy rivalling state-of-the-art methodologies is also reported in a beat tracking task performed on the standard Ballroom Dataset. The paper offers insights into the field of dynamic time signature recognition and beat tracking, offering a valuable and versatile resource for the analysis, composition, and performance of music.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363392/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363392/</guid>
</item>
<item>
<title>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</title>
<description><p><span><big>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</big></span><br></p><p><span><small><i>Yuto Watanabe; Ren Togo; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343335</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Although text-guided image manipulation approaches have demonstrated highly accurate performance for editing the appearance of images in a virtual or simple scenario, their real-world applications face significant challenges. The primary cause of these challenges is the misalignment in the distribution of training and real-world data, which leads to unstable text-guided image manipulation. In this work, we propose a novel framework called TolerantGAN and tackle the new task of real-world text-guided image manipulation independent of the training data. To achieve this, we introduce two key concepts of a border smoothly connection module (BSCM) and a manipulation direction-based attention module (MDAM). BSCM smoothens the misalignment in the distribution of training and real-world data. MDAM extracts only regions highly relevant for image manipulation and assists in reconstructing unobserved objects in the training data. For in-the-wild input images of various classes, TolerantGAN robustly outperforms the state-of-the-art methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360283/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360283/</guid>
</item>
<item>
<title>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</title>
<description><p><span><big>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</big></span><br></p><p><span><small><i>Anastasia Avdeeva; Aleksei Gusev; Tseren Andzhukaev; Artem Ivanov</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343342</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Whispered speech is a quiet voice without vocalization. One of the common cases of using whispered speech is a technique that can help overcome stuttering. But whispered speech can be uncomfortable and difficult to understand in everyday communication. To address these problems, we propose a method of low-delayed whisper-to-speech voice conversion, which can be useful in real life communication of people with disordered speech. As part of our research, we study the impact of streaming Automatic Speech Recognition models on the quality of voice conversion, comparing different streaming models and methods for model adaptation to streaming settings, and showing the importance of using such models in cases of low-delayed voice conversion.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360259/</guid>
</item>
<item>
<title>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</title>
<description><p><span><big>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</big></span><br></p><p><span><small><i>Denis C. Ilie-Ablachim; Andra Băltoiu; Bogdan Dumitrescu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344313</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We propose sparse representation and dictionary learning algorithms for dictionaries whose atoms are characterized by Gaussian probability distributions around some central atoms. This extension of the space covered by the atoms permits a better characterization of localized features of the signals. The representations are computed by trading-off representation error and the probability of the actual atoms used in the representation. We propose two types of algorithms: a greedy one, similar in spirit with Orthogonal Matching Pursuit, and one based on L1-regularization. We apply our new algorithms to unsupervised anomaly detection, where the representation error is used as anomaly score. The flexibility of our approach appears to improve more the representations of the many normal signals than those of the few outliers, at least for an anomaly type called dependency, thus improving the detection quality. Comparison with both standard dictionary learning algorithms and established anomaly detection methods is favorable.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365333/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365333/</guid>
</item>
<item>
<title>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</title>
<description><p><span><big>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</big></span><br></p><p><span><small><i>Yuya Moroto; Yingrui Ye; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344079</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>There are various sentiment theories for categorizing human sentiments into several discrete sentiment categories, which means that the theory used for training sentiment prediction methods does not always match that used in the test phase. As a solution to this problem, zero-shot visual sentiment prediction methods have been proposed to predict unseen sentiments for which no images are available in the training phase. However, the training of these previous zero-shot methods relies on a single sentiment theory, which limits their ability to handle sentiments from other theories. Thus, this article proposes a more robust zero-shot visual sentiment prediction method that can handle cross-domain sentiments defined in different sentiment theories. Specifically, by focusing on the fact that sentiments are abstract concepts common to humans regardless of whether their theories are different, we incorporate knowledge distillation into our method to construct a teacher–student model that can train the implicit relationships between sentiments defined in different sentiment theories. Furthermore, to enhance sentiment discrimination capability and strengthen the implicit relationships between sentiments, we introduce a novel sentiment loss between the teacher and student models. In this way, our model becomes robust to unseen sentiments by exploiting the implicit relationships between sentiments. The contributions of this article are the introduction of knowledge distillation and a novel sentiment loss between the teacher and student models for zero-shot visual sentiment prediction, and improved performance of zero-shot visual sentiment prediction. Experiments on several open datasets demonstrate the effectiveness of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363382/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363382/</guid>
</item>
<item>
<title>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</title>
<description><p><span><big>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</big></span><br></p><p><span><small><i>Chengen Liu; Geert Leus; Elvin Isufi</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339376</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The edge flow reconstruction task consists of retreiving edge flow signals from corrupted or incomplete measurements. This is typically solved by a regularized optimization problem on higher-order networks such as simplicial complexes and the corresponding regularizers are chosen based on prior knowledge. Tailoring this prior to the setting of interest can be challenging or it may not even be possible. Thus, we consider to learn this prior knowledge via a model-based deep learning approach. We propose a new regularized optimization problem for the simplicial edge flow reconstruction task, the simplicial ElasticNet, which combines the advantages of the
$\ell _{1}$
and
$\ell _{2}$
norms. We solve the simplicial ElasticNet problem via the multi-block alternating direction method of multipliers (ADMM) algorithm and provide conditions on its convergence. By unrolling the ADMM iterative steps, we develop a model-based neural network with a low requirement on the number of training data. This unrolling network replaces the fixed parameters in the iterative algorithm by learnable weights, thus exploiting the neural network's learning capability while preserving the iterative algorithm's interpretability. We enhance this unrolling network via simplicial convolutional filters to aggregate information from the edge flow neighbors, ultimately, improving the network learning expressivity. Extensive experiments on real-world and synthetic datasets validate the proposed approaches and show considerable improvements over both baselines and traditional non-model-based neural networks.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342735/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342735/</guid>
</item>
<item>
<title>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</title>
<description><p><span><big>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</big></span><br></p><p><span><small><i>Jim Beckers; Bart Van Erp; Ziyue Zhao; Kirill Kondrashov; Bert De Vries</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337718</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Bayesian model reduction provides an efficient approach for comparing the performance of all nested sub-models of a model, without re-evaluating any of these sub-models. Until now, Bayesian model reduction has been applied mainly in the computational neuroscience community on simple models. In this paper, we formulate and apply Bayesian model reduction to perform principled pruning of Bayesian neural networks, based on variational free energy minimization. Direct application of Bayesian model reduction, however, gives rise to approximation errors. Therefore, a novel iterative pruning algorithm is presented to alleviate the problems arising with naive Bayesian model reduction, as supported experimentally on the publicly available UCI datasets for different inference algorithms. This novel parameter pruning scheme solves the shortcomings of current state-of-the-art pruning methods that are used by the signal processing community. The proposed approach has a clear stopping criterion and minimizes the same objective that is used during training. Next to these benefits, our experiments indicate better model performance in comparison to state-of-the-art pruning schemes.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334001/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334001/</guid>
</item>
<item>
<title>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</title>
<description><p><span><big>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</big></span><br></p><p><span><small><i>Anh Minh Truong; Wilfried Philips; Peter Veelaert</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340064</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Many image-based rendering (IBR) methods rely on depth estimates obtained from structured light or time-of-flight depth sensors to synthesize novel views from sparse camera networks. However, these estimates often contain missing or noisy regions, resulting in an incorrect mapping between source and target views. This situation makes the fusion process more challenging, as the visual information is misaligned, inconsistent, or missing. In this work, we first implement a lightweight network based on the transformer, which is well-known for its capability to model long-range relationships within input data, to extract spatial features from color images. These features are then used to enhance the quality of completed depth maps. Furthermore, we combine a sequential deep neural network with a spatial attention mechanism to effectively fuse the projected features from multiple source viewpoints. This approach enables us to integrate information from an arbitrary number of source viewpoints as well as improve accuracy in synthesized views. Experimental results on challenging datasets demonstrate that our method achieves superior synthesized image quality compared to state-of-the-art (SOTA) methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345792/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345792/</guid>
</item>
<item>
<title>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</title>
<description><p><span><big>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</big></span><br></p><p><span><small><i>Reza Mirzaeifard; Naveen K. D. Venkategowda; Vinay Chakravarthi Gogineni; Stefan Werner</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344395</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper investigates quantile regression in the presence of non-convex and non-smooth sparse penalties, such as the minimax concave penalty (MCP) and smoothly clipped absolute deviation (SCAD). The non-smooth and non-convex nature of these problems often leads to convergence difficulties for many algorithms. While iterative techniques such as coordinate descent and local linear approximation can facilitate convergence, the process is often slow. This sluggish pace is primarily due to the need to run these approximation techniques until full convergence at each step, a requirement we term as a
secondary convergence iteration
. To accelerate the convergence speed, we employ the alternating direction method of multipliers (ADMM) and introduce a novel single-loop smoothing ADMM algorithm with an increasing penalty parameter, named SIAD, specifically tailored for sparse-penalized quantile regression. We first delve into the convergence properties of the proposed SIAD algorithm and establish the necessary conditions for convergence. Theoretically, we confirm a convergence rate of
$o({k^{-\frac{1}{4}}})$
for the sub-gradient bound of the augmented Lagrangian, where
$k$
denotes the number of iterations. Subsequently, we provide numerical results to showcase the effectiveness of the SIAD algorithm. Our findings highlight that the SIAD method outperforms existing approaches, providing a faster and more stable solution for sparse-penalized quantile regression.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365338/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365338/</guid>
</item>
<item>
<title>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</title>
<description><p><span><big>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</big></span><br></p><p><span><small><i>Oliver Lang; Christian Hofbauer; Reinhard Feger; Mario Huemer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343308</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>A promising waveform candidate for future joint sensing and communication systems is orthogonal frequency-division multiplexing (OFDM). For such systems, supporting multiple transmit antennas requires multiplexing methods for the generation of orthogonal transmit signals, where equidistant subcarrier interleaving (ESI) is the most popular multiplexing method. In this work, we analyze a multiplexing method called Doppler-division multiplexing (DDM). This method applies a phase shift from OFDM symbol to OFDM symbol to separate signals transmitted by different Tx antennas along the velocity axis of the range-Doppler map. The main focus of this work lies on the implications of DDM on the communication task. It will be shown that for DDM, the channels observed in the communication receiver are heavily time-varying, preventing any meaningful transmission of data when not taken into account. In this work, a communication system designed to combat these time-varying channels is proposed, which includes methods for data estimation, synchronization, and channel estimation. Bit error ratio (BER) simulations demonstrate the superiority of this communications system compared to ESI-based systems.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360223/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360223/</guid>
</item>
<item>
<title>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</title>
<description><p><span><big>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</big></span><br></p><p><span><small><i>Nils L. Westhausen; Bernd T. Meyer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343320</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we introduce a causal low-latency low-complexity approach for binaural multichannel blind speaker separation in noisy reverberant conditions. The model, referred to as Group Communication Binaural Filter and Sum Network (GCBFSnet) predicts complex filters for filter-and-sum beamforming in the time-frequency domain. We apply Group Communication (GC), i.e., latent model variables are split into groups and processed with a shared sequence model with the aim of reducing the complexity of a simple model only containing one convolutional and one recurrent module. With GC we are able to reduce the size of the model by up to 83% and the complexity up to 73% compared to the model without GC, while mostly retaining performance. Even for the smallest model configuration, GCBFSnet matches the performance of a low-complexity TasNet baseline in most metrics despite the larger size and higher number of required operations of the baseline.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360275/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360275/</guid>
</item>
<item>
<title>Reverse Ordering Techniques for Attention-Based Channel Prediction</title>
<description><p><span><big>Reverse Ordering Techniques for Attention-Based Channel Prediction</big></span><br></p><p><span><small><i>Valentina Rizzello; Benedikt Böck; Michael Joham; Wolfgang Utschick</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344024</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Channel state information (CSI) is crucial for enhancing the performance of wireless systems by allowing to adjust the transmission strategies based on the current channel conditions. However, obtaining precise CSI is difficult because of the fast-changing channel conditions caused by multi-path fading. An inaccurate CSI hinders the performance of various adaptive wireless systems, highlighting the need for channel prediction techniques to effectively mitigate the drawbacks of outdated CSI. Conventional methods typically depend on assumptions regarding user velocity or require knowledge of the Doppler frequency. In contrast to existing approaches, we aim for a more robust and practical solution by training neural networks without making any assumptions about user velocity, relying solely on noisy channel observations during training. Specifically, we adapt both the sequence-to-sequence with attention (Seq2Seq-attn) and transformer models for channel prediction. Additionally, a new technique called reverse positional encoding is introduced in the transformer model to improve the robustness of the model against varying sequence lengths. Similarly, the encoder outputs of the Seq2Seq-attn model are reversed prior to the application of attention mechanisms. By means of simulations, we show that these proposed techniques enable the models to effectively capture relationships within sequences of channel snapshots without increasing the complexity. Importantly, this capability remains robust across varying sequence lengths, representing a substantial improvement over existing methodologies.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363354/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363354/</guid>
</item>
<item>
<title>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</title>
<description><p><span><big>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</big></span><br></p><p><span><small><i>Sarina Meyer; Xiaoxiao Miao; Ngoc Thang Vu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344375</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Speaker anonymization is the task of modifying a speech recording such that the original speaker cannot be identified anymore. Since the first Voice Privacy Challenge in 2020, along with the release of a framework, the popularity of this research topic is continually increasing. However, the comparison and combination of different anonymization approaches remains challenging due to the complexity of evaluation and the absence of user-friendly research frameworks. We therefore propose an efficient speaker anonymization and evaluation framework based on a modular and easily extendable structure, almost fully in Python. The framework facilitates the orchestration of several anonymization approaches in parallel and allows for interfacing between different techniques. Furthermore, we propose modifications to common evaluation methods which improves the quality of the evaluation and reduces their computation time by 65 to 95%, depending on the metric. Our code is fully open source.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365329/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365329/</guid>
</item>
<item>
<title>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</title>
<description><p><span><big>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</big></span><br></p><p><span><small><i>Alessandro Ilic Mezza; Matteo Amerena; Alberto Bernardini; Augusto Sarti</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343318</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Real-time audio communications over IP have become essential to our daily lives. Packet-switched networks, however, are inherently prone to jitter and data losses, thus creating a strong need for effective packet loss concealment (PLC) techniques. Though solutions based on deep learning have made significant progress in that direction as far as speech is concerned, extending the use of such methods to applications of Networked Music Performance (NMP) presents significant challenges, including high fidelity requirements, higher sampling rates, and stringent temporal constraints associated to the simultaneous interaction between remote musicians. In this article, we present PARCnet, a hybrid PLC method that utilizes a feed-forward neural network to estimate the time-domain residual signal of a parallel linear autoregressive model. Objective metrics and a listening test show that PARCnet provides state-of-the-art results while enabling real-time operation on CPU.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360264/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360264/</guid>
</item>
<item>
<title>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</title>
<description><p><span><big>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</big></span><br></p><p><span><small><i>Katerina Zmolikova; Michael Syskind Pedersen; Jesper Jensen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Supervised learning-based speech enhancement methods often work remarkably well in acoustic situations represented in the training corpus but generalize poorly to out-of-domain situations, i.e. situations not seen during training. This stands in the way of further improvement of these methods in realistic scenarios, as collecting paired noisy-clean recordings in the target application domain is typically not feasible. Recording noisy-only in-domain data is, though, much more practical. In this article, we tackle the problem of unsupervised domain adaptation in speech enhancement. Specifically, we propose a way to use in-domain noisy-only data in the training of a neural network to improve upon a model trained solely on out-of-domain paired data. For this, we make use of masked spectrogram prediction, a technique from self-supervised learning that aims to interpolate masked regions of a spectrogram. We hypothesize that masked spectrogram prediction encourages learning of features that represent well both speech and noise components of the noisy signals. These features can then be used to train a more robust speech enhancement system. We evaluate the proposed method on the VoiceBank-DEMAND and LibriFSD50k databases, with WSJ0-CHiME3 serving as the out-of-domain database. We show that the proposed method outperforms both the out-of-domain system and the baseline approaches, i.e. RemixIT and noisy-target training, and also combines well with the previously proposed RemixIT method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360251/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360251/</guid>
</item>
<item>
<title>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</title>
<description><p><span><big>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</big></span><br></p><p><span><small><i>Alexander Karataev; Christian Forsch; Laura Cottatellucci</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3348343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We consider a cell-free massive multiple-input multiple-output (CF-MaMIMO) communication system in the uplink transmission and propose a novel algorithm for blind or semi-blind joint channel estimation and data detection (JCD). We formulate the problem in the framework of bilinear inference and develop a solution based on the expectation propagation (EP) method for both channel estimation and data detection. We propose a new approximation of the joint a posteriori distribution of the channel and data whose representation as a factor graph enables the application of the EP approach using the message-passing technique, local low-complexity computations at the nodes, and an effective modeling of channel-data interplay. The derived algorithm, called bilinear-EP JCD, allows for a distributed implementation among access points (APs) and the central processing unit (CPU) and has polynomial complexity. Our simulation results show that it outperforms other EP-based state-of-the-art polynomial time algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10378663/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10378663/</guid>
</item>
<item>
<title>Adversarial Representation Learning for Robust Privacy Preservation in Audio</title>
<description><p><span><big>Adversarial Representation Learning for Robust Privacy Preservation in Audio</big></span><br></p><p><span><small><i>Shayan Gharib; Minh Tran; Diep Luong; Konstantinos Drossos; Tuomas Virtanen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Sound event detection systems are widely used in various applications such as surveillance and environmental monitoring where data is automatically collected, processed, and sent to a cloud for sound recognition. However, this process may inadvertently reveal sensitive information about users or their surroundings, hence raising privacy concerns. In this study, we propose a novel adversarial training method for learning representations of audio recordings that effectively prevents the detection of speech activity from the latent features of the recordings. The proposed method trains a model to generate invariant latent representations of speech-containing audio recordings that cannot be distinguished from non-speech recordings by a speech classifier. The novelty of our work is in the optimization algorithm, where the speech classifier's weights are regularly replaced with the weights of classifiers trained in a supervised manner. This increases the discrimination power of the speech classifier constantly during the adversarial training, motivating the model to generate latent representations in which speech is not distinguishable, even using new speech classifiers trained outside the adversarial training loop. The proposed method is evaluated against a baseline approach with no privacy measures and a prior adversarial training method, demonstrating a significant reduction in privacy violations compared to the baseline approach. Additionally, we show that the prior adversarial method is practically ineffective for this purpose.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10379095/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10379095/</guid>
</item>
<item>
<title>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</title>
<description><p><span><big>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</big></span><br></p><p><span><small><i>Yasaman Parhizkar; Gene Cheung; Andrew W. Eckford</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>It is a popular hypothesis in neuroscience that ganglion cells in the retina are activated by selectively detecting visual features in an observed scene. While ganglion cell firings can be predicted via data-trained deep neural nets, the n... |
http://localhost:1200/ieee/journal/4427201/earlyaccess - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Sensors Journal</title>
<link>https://ieeexplore.ieee.org/xpl/tocresult.jsp?isnumber=4427201</link>
<atom:link href="http://localhost:1200/ieee/journal/4427201/earlyaccess" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Sensors Journal - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Wed, 10 Apr 2024 14:40:03 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Torsion Sensor Based on Helical Long-Period Grating Inscribed in the Etched Double Cladding Fiber</title>
<description><p><span><big>Torsion Sensor Based on Helical Long-Period Grating Inscribed in the Etched Double Cladding Fiber</big></span><br></p><p><span><small><i>Yanping He; Yuehui Ma; Chen Jiang; Yunqi Liu</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373623</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>We proposed a high-sensitivity torsion sensor based on the helical long-period grating (HLPG) inscribed in the cladding-etched double cladding fiber (DCF). The torsion sensitivity has been greatly improved by reducing the cladding diameter of the DCF, from 0.342 nm/(rad/m) to 1.162 nm/(rad/m), which is one order of magnitude higher than that of the conventional long-period fiber grating. The cladding mode coupling characteristics during the etching process are also investigated experimentally. The DCF-HLPG has a low strain and temperature sensitivity, and the maximum strain sensitivity is only about 0.0022 nm/με. The proposed DCF-HLPG has a potential application as a high-sensitivity torsion sensor with low strain and temperature crosstalk.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466518/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466518/</guid>
</item>
<item>
<title>Deep-Learning-Based Prediction Algorithm for Fuel Cell Electric Vehicle Energy with Shift Mixup</title>
<description><p><span><big>Deep-Learning-Based Prediction Algorithm for Fuel Cell Electric Vehicle Energy with Shift Mixup</big></span><br></p><p><span><small><i>Tae-Ho Kim; Jae-Heung Cho; Young-Kwang Kim; Joon-Hyuk Chang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373078</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The automobile industry is switching from fossil-fuel-based energy sources to green energy sources. In particular, because hydrogen has a high energy density and efficiency compared to other energy sources, it is a major research field for commercial vehicles that require large amounts of energy and travel long distances on average, such as buses and trucks. In fuel cell electric vehicles (FCEVs), maintaining an adequate energy production intensity is necessary to improve energy production efficiency and prevent falling into a state of inoperability owing to a lack of energy. In this case, energy prediction becomes a significant factor. In this study, a deep-learning-based prediction method for FCEV powertrain energy is proposed. The proposed method uses only internal data, which can be obtained from the vehicle, and does not require external information regarding future routes. Additionally, we designed a model considering the vehicle data time-series characteristics, and proposed a shift mixup, which is a data augmentation method that does not compromise the vehicle’s dynamic characteristics, to address the data shortage problem. Furthermore, a pretext task-learning method that can improve model performance without external data during inference is introduced. This method includes pretext tasks designed specifically for the vehicle domain. Finally, a distance-based loss mask and contrastive learning that the representation can learn semantic information are proposed. Experimental results for the actual driving dataset show that our method improved by 28.05% and by 30.04% compared to the exponential moving average in terms of root mean squared error and mean absolute error, respectively. We demonstrate with energy management strategy (EMS) that an effective energy prediction algorithm helps sustain an optimal state of charge (SOC).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466522/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466522/</guid>
</item>
<item>
<title>Effect of substrate temperature on Cu2-xO growth for enhanced photodetector performance in visible region</title>
<description><p><span><big>Effect of substrate temperature on Cu2-xO growth for enhanced photodetector performance in visible region</big></span><br></p><p><span><small><i>Karthickraja Ramakrishnan; Y. Ashok Kumar Reddy; B. Ajitha</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373768</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The present study investigates the effect of substrate temperature (30-300°C) on the sputter-deposited Cu
2-x
O films for improved photodetector performance in the visible region. Cubic structured Cu2O phase is observed for all deposited films through structural analysis. Increasing substrate temperature causes the formation of nanospheres with higher grain size at 200°C confirmed by the morphological images. Moreover, all the deposited thin films show a strong absorption band in the visible region (526-550 nm). Among deposited films, a lower bandgap (2.02 eV) is observed for the film deposited at 200°C due to the higher surface area-to-volume ratio and increased grain size. The better crystalline nature, more copper vacancy, and formation of nanospheres with a larger grain size of the film deposited at 200°C result in higher mobility and conductivity. Further, the photodetector performance of the Au/Cu
2-x
O/Au test-devices is studied under the visible light (λ=530 nm) irradiation. Among samples, the sample grown at 200°C showed higher photocurrent (36.10 μA), I
on
/I
off
ratio (183.50), photoresponsivity (1.127 A/W), specific detectivity (22.40×10
11
Jones), and faster response times (τ
r
=92 ms and τ
f
=84 ms) even at lower incident optical power density (0.1283 mW/cm
2
). These improved experimental results signify the potential of the fabricated test-device at 200°C for the optoelectronics devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10471292/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10471292/</guid>
</item>
<item>
<title>A Cavity-Type Pressure Sensor Array with High Anti-Disturbance Performance Inspired by Fish Lateral Line Canal</title>
<description><p><span><big>A Cavity-Type Pressure Sensor Array with High Anti-Disturbance Performance Inspired by Fish Lateral Line Canal</big></span><br></p><p><span><small><i>Qian Yang; Qiao Hu; Liuhao Shan; Guangyu Jiang; Yuanji Yao; Long Tang; Zicai Zhu; Alvo Aabloo</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3381162</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Effective sensing is challenging in real underwater environment swarming with various disturbances. However, most research on underwater sensors has been conducted in still water, which may not provide sufficient insights for environments with disturbances. In addressing this issue, our paper not only proposes the development of anti-disturbance sensor units but also delves into evaluating the localization capabilities of artificial lateral line arrays in disturbed water environments. Concretely, inspired by the lateral line canal system of fish, we proposed a cavity-type pressure sensing device (denoted as a C-sensor) with notable anti-disturbance performance. On one hand, a 3D-printed resin cavity is filled with silicone oil, with flexible latex film packaging the upper and lower openings to achieve anti-disturbance capabilities akin to those of fish lateral line canals. On the other hand, employing IPMC materials based on ion sensing principles, the sensor generates electrical signals similar to sensing cilia in fish lateral line systems. The results demonstrated that the cavityless sensor (denoted as N-sensor) showed self-oscillation behavior, leading to significant errors in weak target signal detection and even sensor failure in extreme cases. However, the C-sensor exhibits remarkable sensing performance in still water and water disturbances, thereby confirming the effectiveness of its cavity-type structure in mitigating the impact of water disturbances. Furthermore, we developed two distinct artificial lateral line arrays based on the above sensors (C-array and N-array). The experimental results showed a marked 78% reduction in mean localization error for C-array compared to N-array. Additonally, the C-array also exhibited good ability in multiple dipoles detection. Such results indicate the superior applicability of the C-array within real complex underwater environments, promoting the practical application of IPMC sensors.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10486206/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10486206/</guid>
</item>
<item>
<title>Multi-sensor multiple extended objects tracking based on the message passing</title>
<description><p><span><big>Multi-sensor multiple extended objects tracking based on the message passing</big></span><br></p><p><span><small><i>Yuansheng Li; Tao Shen; Lin Gao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3384560</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Multiple extended objects (EOs) tracking has attracted a lot of attention due to the fast development of high-resolution sensors. A remarkable feature of EO, compared to the traditional point target, is that an EO normally produces more than one measurement, resulting in challenges for finding the associations among objects and measurements. In this paper, we are interested in tracking an unknown number of EOs with multiple sensors by resorting to the message passing method. The existence probability and belief of each EO are explicitly estimated, where the belief is modeled by a mixture gamma Gaussian inverse Wishart (GGIW) distribution, so as to jointly estimate the measurement rate (MR), centroid state and extension. The marginal posterior of EOs is approximately by belief, which is obtained by running the loopy sum-product algorithm (LSPA) on a suitably devised factor graph. As a result, the computational load of the proposed algorithm increases linearly with respect to the number of targets, thus admitting the scalability. Simulation experiments are carried out to verify the performance of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10495766/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10495766/</guid>
</item>
<item>
<title>Dual-Resonance Split Ring Resonator Metasurface for Terahertz Biosensing</title>
<description><p><span><big>Dual-Resonance Split Ring Resonator Metasurface for Terahertz Biosensing</big></span><br></p><p><span><small><i>Arslan Asim; Michael Cada; Yuan Ma; Alan Fine</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3376290</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this paper, the ‘terahertz gap’ has been addressed by designing a novel THz metasurface for potential use in biosensing applications. The metasurface sensor employs Surface Plasmon Resonance (SPR). It operates in the 0 - 1 THz band. Two sharp reflection dips are provided by the sensor, which serve as indicators of analyte refractive index variations. Geometrical as well as compositional parameters of the biosensor design have been studied to optimize the performance in the targeted frequency band. The sensor design shows compatibility with different metals. The performance of the metasurface with gold, copper and aluminum has been investigated. The metasurface geometry is decently resilient to fabrication tolerances. The sensor maintains its resonance conditions when the angle of incidence is changed with minor deviations in the spectral response, but the polarization state of the incident terahertz beam clearly disturbs the absorption peak. Therefore, the sensing performance is restricted to a maximum allowable incidence angle of 20 degrees and circularly polarized terahertz beams. The resonance conditions for the metasurface appear around 0.4 and 0.7 THz. Both resonances have been investigated with respect to changes in the analyte refractive index. The chosen refractive index range is 1 to 1.5. The sensor response is calibrated by plotting the resonance frequency versus the refractive index. Least squares regression technique has been used to extract a data model for sensor response. Comparison of the proposed design with contemporary works has been incorporated into the paper. The sensor provides sensitivities of 0.1614 THz/ RIU and 0.23 THz/ RIU. The electromagnetic simulations have been carried out through the Finite Element Method (FEM).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10474307/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10474307/</guid>
</item>
<item>
<title>An Intestine based Bio-Compatible Humidity Sensor for Environmental and Medical Measurements</title>
<description><p><span><big>An Intestine based Bio-Compatible Humidity Sensor for Environmental and Medical Measurements</big></span><br></p><p><span><small><i>Emrehan Yavsan; Muhammet Rojhat Kara; Mehmet Akif Erismis</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3367032</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The bio-compatible devices suitable for recycling and bio-degrading can be achieved with organic materials in nature. In this work, a bio-compatible capacitive humidity sensor is presented for reducing the amount of electronic waste and contributing to the sustainability of natural resources and the future. The sensor consists of 3 layers. The first layer is the processed intestine layer of cattle. Bio-compatibility is achieved with this layer. In addition to being a highly absorbing tissue, the intestine has been used for centuries for long-term preservation of meat based food. Correspondingly, the developed sensor is found to be more durable and long-lasting than other natural-material based humidity sensors in the literature. The other layers of the sensor are interdigitated copper electrodes and a 0.2 mm thick thin film strip. Thin film strip increases mechanical strength as well as flexibility. The developed sensor prototype was subjected to various tests in the humidity range of 20%-90%. In these tests, the hysteresis characteristic of the sensor, its response-recovery time, and its long-term stability and short-term step responses were examined. Moreover, as a possible application in medicine, the sensor can be used to detect breathing cycles. The sensor’s response and recovery times were measured as 8.72’ and 4.47’, respectively, possibly attributed to the stabilization of our test setup, while the sensor successfully detected deep, normal and fast breathing. Despite being kept in an uncontrolled environment, the sensor continued to operate consistently for breath measurements after 56 weeks, which is more than a year.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10445282/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10445282/</guid>
</item>
<item>
<title>Deep neural network assisted terahertz metasurface sensors for the detection of lung cancer biomarkers</title>
<description><p><span><big>Deep neural network assisted terahertz metasurface sensors for the detection of lung cancer biomarkers</big></span><br></p><p><span><small><i>Jie Chen; Fangrong Hu; Xiaoya Ma; Mo Yang; Shangjun Lin; An Su</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3384578</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>MicroRNA (miRNA) is closely related to various cancers, and the change in its expression level is closely related to the development and death of tumor cells. Here, we design and manufacture a terahertz (THz) metasurface sensor to realize the concentration detection and category identification of the miRNAs related to lung cancer. We established two spectral classification datasets, one contains 9 concentrations of miRNAs, and the other contains 3 categories of lung cancer-related miRNAs. And then, we used a deep neural network (DNN) algorithm to classify these spectral datasets and obtained a LOD (Limit-of-Detection) of 100 aM for miRNAs. Moreover, this method can identify the categories of miRNAs. Compared with the other five machine learning (ML) algorithms, the proposed neural network framework achieves the best classification results. This work provides a new way for the detection and identification of trace nucleic acid and biomarkers of many cancers.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10495770/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10495770/</guid>
</item>
<item>
<title>A Linearized Temperature Measurement System Based on the B-mode of SC-cut Quartz Crystal</title>
<description><p><span><big>A Linearized Temperature Measurement System Based on the B-mode of SC-cut Quartz Crystal</big></span><br></p><p><span><small><i>Zhiqi Li; Haipeng Zhai; Miao Miao; Wei Zhou</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3374763</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This article demonstrates a linearized temperature measurement system based on the B-mode of SC-cut quartz crystal, which has high precision and good linearity. The proposed system uses a digital linear phase comparison measurement method by selecting phase detection region, which achieves high resolution, and fast response for temperature measurement. Extensive error analysis and performance evaluation of the system were conducted in this study. The developed linearized temperature measurement system demonstrated excellent performance, including: wide temperature measuring range (-10°C to 80°C); providing sufficient resolution and extremely fast response time with a resolution of 0.01°C at 10 μs, a resolution of 5E-5°C at 10 ms, and a resolution of 1E-7°C at 1 s when the temperature keeps stable. It achieves extremely low temperature measurement relative error (R.E.&lt; 0.24%) and nonlinearity (N.L.&lt; 0.18%). The measurement results show that the maximum estimate of the Allan variance is 2.3E-4°C. The proposed system has wide prospects for intelligent applications and can play a crucial role in scenarios that require high-precision and fast-response temperature measurements.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472879/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472879/</guid>
</item>
<item>
<title>Mesoporous silica modified by poly(ionic liquid)s for low humidity sensing</title>
<description><p><span><big>Mesoporous silica modified by poly(ionic liquid)s for low humidity sensing</big></span><br></p><p><span><small><i>Zhiyan Ma; Yaping Song; Hongran Zhao; Sen Liu; Xi Yang; Teng Fei; Tong Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373030</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this study, imidazole-based poly(ionic liquid)s are introduced into acid-treated mesoporous silica by solid grinding and subsequent thermal polymerization methods. The purpose of treating the mesoporous silica with hydrochloric acid is to increase the number of hydroxyl groups on the surface to fix more poly(ionic liquid)s. Due to the synergism of enhanced hydrophilicity and the mesoporous structure, the synthesized composite materials exhibited excellent capacities to adsorb and transport water vapor molecules and are suitable for low humidity sensing applications. In the 7-33% relative humidity (RH) range, the response of the as-prepared sensor reached 806% with a short response/recovery time of 8/13 s.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466490/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466490/</guid>
</item>
<item>
<title>PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards</title>
<description><p><span><big>PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards</big></span><br></p><p><span><small><i>Elias Arbash; Margret Fuchs; Behnood Rasti; Sandra Lorenz; Pedram Ghamisi; Richard Gloaguen</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3380826</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Addressing the critical theme of recycling electronic waste (E-waste), this contribution is dedicated to developing advanced automated data processing pipelines as a basis for decision-making and process control. Aligning with the broader goals of the circular economy and the United Nations (UN) Sustainable Development Goals (SDG), our work leverages non-invasive analysis methods utilizing RGB and hyperspectral imaging data to provide both quantitative and qualitative insights into the E-waste stream composition for optimizing recycling efficiency. In this paper, we introduce ’PCB-Vision’; a pioneering RGB-hyperspectral printed circuit board (PCB) benchmark dataset, comprising 53 RGB images of high spatial resolution paired with their corresponding high spectral resolution hyperspectral data cubes in the visible and near-infrared (VNIR) range. Grounded in open science principles, our dataset provides a comprehensive resource for researchers through high-quality ground truths, focusing on three primary PCB components: integrated circuits (IC), capacitors, and connectors. We provide extensive statistical investigations on the proposed dataset together with the performance of several state-of-the-art (SOTA) models, including U-Net, Attention UNet, Residual U-Net, LinkNet, and DeepLabv3+. By openly sharing this multiscene benchmark dataset along with the baseline codes, we hope to foster transparent, traceable, and comparable developments of advanced data processing across various scientific communities, including, but not limited to, computer vision and remote sensing. Emphasizing our commitment to supporting a collaborative and inclusive scientific community, all materials, including code, data, ground truth, and masks, will be accessible at https://github.com/hifexplo/PCBVision.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10485259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10485259/</guid>
</item>
<item>
<title>DEEP-CARDIO: Recommendation System for Cardiovascular Disease Prediction using IOT Network</title>
<description><p><span><big>DEEP-CARDIO: Recommendation System for Cardiovascular Disease Prediction using IOT Network</big></span><br></p><p><span><small><i>A. Yashudas; Dinesh Gupta; G. C. Prashant; Amit Dua; Dokhyl AlQahtani; A. Siva Krishna Reddy</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373429</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Internet of Things (IoT) based remote healthcare applications provide fast and preventative medical services to the patients at risk. However, predicting heart disease is a complex task and diagnosis results are rarely accurate. To address this issue, a novel Recommendation System for Cardiovascular Disease Prediction Using IoT Network (DEEP-CARDIO) has been proposed for providing prior diagnosis, treatment, and dietary recommendations for cardiac diseases. Initially, the physiological data are collected from the patient’s remotely by using the four bio sensors such as ECG sensor, Pressure sensor, Pulse sensor and Glucose sensor. An Arduino controller receives the collected data from the IoT sensors to predict and diagnose the disease. A cardiovascular disease prediction model is implemented by using BiGRU (Bidirectional-Gated Recurrent Unit) attention model which diagnose the cardiovascular disease and classify into five available cardiovascular classes. The recommendation system provides physical and dietary recommendations to cardiac patients based on the classified data, via user mobile application. The performance of the DEEP-CARDIO is validated by Cloud Simulator (CloudSim) using the real-time Framingham’s and Statlog heart disease dataset. The proposed DEEP CARDIO method achieves an overall accuracy of 99.90% whereas, the MABC-SVM, HCBDA and MLbPM method achieves 86.91%, 88.65% and 93.63% respectively.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472883/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472883/</guid>
</item>
<item>
<title>AuNP-decorated textile as chemo resistive sensor for acetone detection</title>
<description><p><span><big>AuNP-decorated textile as chemo resistive sensor for acetone detection</big></span><br></p><p><span><small><i>S. Casalinuovo; D. Caschera; S. Quaranta; D. Caputo</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2023.3348693</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This paper presents the development of chemo resistive sensors for the detection of volatile organic compounds (VOCs). The proposed sensor is based on citrate-functionalized gold nanoparticles (AuNPs) serving as a sensitive layer deposited on cotton fabric. Impedance variations due to VOC/substrate interaction are used as a detection principle. Specifically, this work focuses on acetone detection after exposing the AuNP-decorated cotton to a CH3COCH3 aqueous solution. Such an interaction resulted in a reduction of the total impedance (i.e., magnitude) of the system. This behavior can be ascribed to Van der Waals forces existing between the C=O group and the citrate moieties adsorbed on the gold nanoparticles, which favor charge injection to the substrate. Response to water was also tested for comparison, assuring that the solvent interacts with the sensitive layer by a different adsorption mechanism, not influencing the overall results. Sensor selectivity was also verified by considering ethanol (representative of alcohol group). Indeed, impedance curves reflect the different type of chemical interaction between the analyte and the substrate. In addition, sensor limit of detection for acetone was found to be 1% v/v, in the considered frequency range. Furthermore, sensor performance in terms of reusability was evaluated, showing that the Au-cotton ability in VOCs detection could be restored after about 90 min with a percentage up to 97 % in the frequency of 1Hz. These results can be considered the starting point for the development of portable, sensitive and user-friendly devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10384362/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10384362/</guid>
</item>
<item>
<title>L-cysteine modified gold nanoparticles for copper (II) ion detection using etched Fiber Bragg Grating sensor</title>
<description><p><span><big>L-cysteine modified gold nanoparticles for copper (II) ion detection using etched Fiber Bragg Grating sensor</big></span><br></p><p><span><small><i>S Srivatzen; B S Kavitha; Asha Prasad; S Asokan</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3374514</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>A highly sensitive and selective optical detection of copper (II) (Cu
2+
) concentration in water has been proposed using L-cysteine functionalized gold nanoparticles (AuNPs) using an etched Fiber Bragg grating sensor (eFBG). The eFBG sensor produces a Bragg wavelength shift (ΔλB) in correspondence to the added Cu
2+
concentration. The amino acid L-Cysteine facilitates the formation of core-satellite nanoassemblies of gold in the presence of Cu
2+
. A linear wavelength shift is observed in accordance with the mediating Cu
2+
concentration. The sensor’s detection limit is found to be 1picomolar (pM) which is well within the world health organization (WHO) acceptable limit for safe drinking water and the dynamic range of the sensor lies between 1pM to 100 μM. The sensor is highly sensitive and selective to Cu
2+
with sensitivity of 230 pm 10
-1
M. The sensor has proven its capability as an ideal probe for onsite, real-time measurement of Cu
2+
in drinking water by giving a recovery percentage of &gt;90% in real water sample additional to good reproducibility (standard error&lt; 20 pm), high sensitivity, portability and specificity.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472453/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472453/</guid>
</item>
<item>
<title>AcHand: Detecting Respiratory Rate of operator via Smartphone Microphone</title>
<description><p><span><big>AcHand: Detecting Respiratory Rate of operator via Smartphone Microphone</big></span><br></p><p><span><small><i>KounKou Vincent; Lin Wang; Nan Jing; Wenyuan Liu</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3372255</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this paper, we dedicate a handgrip-free technique, called ”AcHand”, to continuously detect respiratory rates under acoustic signals sensing via a smartphone. Although the particularity of the AcHand is to follow the subtle movements of the heart during its displacement, we acknowledge specific challenges that can make the signal vulnerable to handgripfree scenarios such as 1) stochastic hand activities: fluctuations in movement can alter the trajectory of the acoustic signal to the microphones and cause random changes in the breathing curve, resulting in complex and unpredictable dynamics, and 2) clicking on the phone screen can disrupt the breathing by generating a large signal variation to create undesirable peaks-to-noisy the signal. However, by analyzing the recorded signal reflection evoked by a chirp sound stimulus, we propose a linear transformation of the Cardioreflector-Manoreflector to analyze the respiratory patterns variation and distinguish weak breathing signals under the handgrip-free. We compare the present signal with the ground truth recorded by ECG to approve the efficiency of the proposed method. Our experiments show AcHand achieves a MAE of 0.341 bpm, which is a significant improvement over the state-of-the-art devices, and enhancing detection capability across handgrip-free scenarios.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10462909/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10462909/</guid>
</item>
<item>
<title>Optimization Method for Node Deployment of Closed-Barrier Coverage in Hybrid Directional Sensor Networks</title>
<description><p><span><big>Optimization Method for Node Deployment of Closed-Barrier Coverage in Hybrid Directional Sensor Networks</big></span><br></p><p><span><small><i>Peng Wang; Yonghua Xiong; Jinhua She; Anjun Yu</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3378998</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Closed-barrier coverage is a new coverage problem that requires sensors on the barrier to form a closed loop. Due to the features of end-to-end connection, constructing closed barriers is a challenging task. In a Hybrid Directional Sensor Network, sensors can both rotate and move, making it more difficult to schedule sensors to construct closed barriers. To address this challenge, we propose an optimization method for node deployment of closed-barrier coverage (CCOM). First, selecting multiple initiators transforms the closed-barrier coverage problem into a simple linear-barrier construction problem. Then, a weighted barrier graph is used to search the optimal barrier path. Finally, design a rotation strategy and moving plan for the sensors, fully utilizing their rotation and moving capabilities to achieve closed-barrier coverage with the minimum number of sensors and energy consumption. The simulation experiment results show that compared with other advanced methods, our proposed method has better performance in terms of barrier construction success rate, minimum required number of mobile sensors, moving distance, and network lifetime.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10478814/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10478814/</guid>
</item>
<item>
<title>Enhanced Multicast Protocol for Low-power and Lossy IoT Networks</title>
<description><p><span><big>Enhanced Multicast Protocol for Low-power and Lossy IoT Networks</big></span><br></p><p><span><small><i>Issam Eddine Lakhlef; Badis Djamaa; Mustapha Reda Senouci; Abbas Bradai</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3375797</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Communication protocols in the Internet of Things (IoT) should take into account the resource-constrained nature of Low-power and Lossy Networks (LLNs). IP multicast protocols allow a packet to be routed from one source to multiple destinations in a single transmission. Hence, resources such as bandwidth, energy, and time are saved for a multitude of LLN applications, ranging from over-the-air programming and information sharing to device configuration and resource discovery. In this context, several multicast routing protocols have recently been proposed for LLNs, including the Multicast Protocol for LLNs (MPL). MPL has proven to be very reliable in many scenarios. However, the great resource consumption, especially in terms of energy and bandwidth, remains the main drawback of this protocol. In this paper, after a detailed overview, we provide an in-depth analysis of the MPL protocol to highlight its functional weaknesses. Then, we propose several improvements that touch on different areas to address such limitations. Extensive realistic simulations and experiments were performed to study the performance of the proposed improvements to MPL. The results obtained show that our proposals outperform the MPL protocol in terms of resource consumption (memory, bandwidth, and energy) while improving its performance in terms of end-to-end delay and maintaining the same reliability of data packet delivery.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10473686/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10473686/</guid>
</item>
<item>
<title>A 7.25μK Ultra-High Resolution MEMS Resonant Thermometer</title>
<description><p><span><big>A 7.25μK Ultra-High Resolution MEMS Resonant Thermometer</big></span><br></p><p><span><small><i>Zheng Wang; Liangbo Ma; Xiaorui Bie; Xingyin Xiong; Zhaoyang Zhai; Wuhao Yang; Yongjian Lu; Xudong Zou</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3370956</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This paper proposes an ultra-high-resolution MEMS resonant thermometer that exploits differences in Young’s modulus and coefficient of thermal expansion (CTE) between structural layers to increase thermal stress due to temperature variation. The increased thermal stress eventually acts on the resonator to produce a large frequency shift, and the temperature coefficient of frequency (TCF) can be increased to 17.4 times the original, which is consistent with theoretical analysis and finite element method simulations. The MEMS resonator chip is manufactured by the standard Silicon-on-insulator process and stacked on a ceramic chip carrier through multiple layers of materials. A self-sustaining oscillator, mainly composed of a packaged MEMS resonator chip and a low-noise application-specific integrated circuit (ASIC), is built to track the resonant frequency shift of the resonator. This MEMS resonant thermometer prototype demonstrates a high temperature coefficient of frequency (TCF) of -866.84ppm/K from -50°C to 40°C and exhibits good linearity. An ultra-high resolution of 7.25μK is achieved in the closed-loop experimental test. This is the best result achieved for a MEMS thermometer employing the resonant sensing paradigm to date.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10460450/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10460450/</guid>
</item>
<item>
<title>Design of MTM-based Multi-band Micro-Biosensor in Terahertz region as perfect absorber for Early-Stage Leukemia Diagnosis with sensitivity 18626373 TH...</title>
<description><p><span><big>Design of MTM-based Multi-band Micro-Biosensor in Terahertz region as perfect absorber for Early-Stage Leukemia Diagnosis with sensitivity 18626373 THz/RIU</big></span><br></p><p><span><small><i>Musa N. Hamza; Mohammad Tariqul Islam</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383522</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this article, a novel highly sensitive biosensor based on perfect metamaterial absorbers is presented. A detailed study is presented for several different models, different types of substrate materials, different types of resonant materials and substrate thicknesses showing very good sensitivity to any changes. The proposed biosensor exhibits high sensitivity to different polarization and incidence angles, ensuring its performance in terms of signal-to-noise ratio. The proposed biosensor was carefully compared with previously designed sensors and biosensors. The proposed biosensor exhibits amazing sensitivity, such as a quality factor of 41.1, a figure of merit (FOM) of 172466416 RIU-1, and a sensitivity (S) of 18626373 THz/RIU. The high sensitivity of the biosensor allows early detection of leukemia, demonstrating significant differences between leukemia and normal blood. Another interesting result of this article is the use of terahertz waves for imaging. Microwave imaging is performed for electric fields, magnetic fields, and power.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10494209/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10494209/</guid>
</item>
<item>
<title>LEPS: A lightweight and effective single-stage detector for pothole segmentation</title>
<description><p><span><big>LEPS: A lightweight and effective single-stage detector for pothole segmentation</big></span><br></p><p><span><small><i>Xiaoning Huang; Jintao Cheng; Qiuchi Xiang; Jun Dong; Jin Wu; Rui Fan; Xiaoyu Tang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2023.3330335</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Currently, the problem of potholes on urban roads is becoming increasingly severe. Identifying and locating road potholes promptly has become a major challenge for urban construction. Therefore, we proposed a lightweight and effective instance segmentation method called LEPS (Lightweight and Effective Pothole Segmentation Detector) for road pothole detection. To extract the image edge information and gradient information of potholes from the feature map more efficiently, we proposed a module that performs the convolutional super-position supplemented with a convolutional kernel to enhance spatial details for the backbone. We have designed a novel module applied to the Neck layer, improving the detection performance while reducing the parameters. To enable accurate segmentation of fine-grained features, we optimized the ProtoNet, which enables the segmentation head to generate high-quality masks for more accurate prediction. We have fully demonstrated the effectiveness of the method through a large number of comparative experiments. Our detector has excellent performance on two authoritative example datasets, URTIRI and COCO, and can successfully be applied to visual sensors to accurately detect and segment road potholes in real environments, accurately LEPS reached 0.892 and 0.648 in terms of Mask for AP50 and AP50:95, which is improved by 4.6% and 20.6% compared to the original model. These results demonstrate its strong competitiveness when compared to other models. Comprehensively, LEPS improves detection accuracy while maintaining lightweight, which allows the model to meet the practical application requirements of edge computing devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10315064/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10315064/</guid>
</item>
<item>
<title>UMHE: Unsupervised Multispectral Homography Estimation</title>
<description><p><span><big>UMHE: Unsupervised Multispectral Homography Estimation</big></span><br></p><p><span><small><i>Jeongmin Shin; Jiwon Kim; Seokjun Kwon; Namil Kim; Soonmin Hwang; Yukyung Choi</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383453</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Multispectral image alignment plays a crucial role in exploiting complementary information between different spectral images. Homography-based image alignment can be a practical solution considering a tradeoff between runtime and accuracy. Existing methods, however, have difficulty with multispectral images due to the additional spectral gap or require expensive human labels to train models. To solve these problems, this paper presents a comprehensive study on multispectral homography estimation in an unsupervised learning manner. We propose a curriculum data augmentation, an effective solution for models learning spectrum-agnostic representation by providing diverse input pairs. We also propose to use the phase congruency loss that explicitly calculates the reconstruction between images based on low-level structural information in the frequency domain. To encourage multispectral alignment research, we introduce a novel FLIR corresponding dataset that has manually labeled local correspondences between multispectral images. Our model achieves state-of-the-art alignment performance on the proposed FLIR correspondence dataset among supervised and unsupervised methods while running at
151 FPS
. Furthermore, our model shows good generalization ability on the M3FD dataset without finetuning.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10494213/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10494213/</guid>
</item>
<item>
<title>Calibration and Evaluation of a Low-Cost Optical Particulate Matter Sensor for Measurement of Lofted Lunar Dust</title>
<description><p><span><big>Calibration and Evaluation of a Low-Cost Optical Particulate Matter Sensor for Measurement of Lofted Lunar Dust</big></span><br></p><p><span><small><i>Abhay Vidwans; Jeffrey Gillis-Davis; Pratim Biswas</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3366436</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Several recent Earth-based investigations employed low-cost particulate matter sensors to address the lack of spatiotemporal resolution in air quality data. The lunar environment also has a particulate matter problem in the form of fine lofted and levitated dust particles. Natural and anthropogenic mobilized dust can cause a slew of difficulties for surface operations (deposition onto radiators, optical components, and mechanical devices). Despite the urgency of mitigating dust on the Moon and other airless bodies, the performance of low-cost sensors has not been critically evaluated for space applications. Upcoming long-term robotic and human exploration missions to the Moon necessitate a robust sensor that can monitor particulate matter levels and establish a spatially and temporally resolved global network. In this work, we calibrate two optical light-scattering particulate matter sensors against research-grade aerosol instruments for measuring the concentration of aerosolized lunar simulants. Sensors showed a stronger dependence on aerosol particle size distribution than particle composition. Vacuum testing showed a significant deviation in performance compared to atmospheric pressure, with a stronger dependency on lunar simulant. The predicted mass deposition, based on sensor output coupled with dust trajectory, was within an order of magnitude of the reference deposition. Our results demonstrate for the first time that low-cost particulate matter sensors can monitor dust concentrations with reasonable accuracy in a vacuum environment, with two caveats. First, precise calibrations must be performed with a dust simulant closely matching the particle size distribution of the target dust, and second, atmospheric pressure calibrations alone are insufficient.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10462021/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10462021/</guid>
</item>
<item>
<title>Reweighting Interacting Multiple-model Algorithm to Overcome Model Competition for Target Tracking in the Hybrid System</title>
<description><p><span><big>Reweighting Interacting Multiple-model Algorithm to Overcome Model Competition for Target Tracking in the Hybrid System</big></span><br></p><p><span><small><i>Guowei Li; Shurui Zhang; Yubing Han; Weixing Sheng; Thia Kirubarajan</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3369854</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The vicious competition of interacting multiple-model algorithm (IMM) is an inherent problem and would produce irreversible effects on IMM estimation results, especially combining with the radar system. In this paper, a novel reweighting IMM (RIMM) is proposed to overcome this issue. Firstly, the theoretical lower bound of model numbers in different situations is respectively provided through the analysis of IMM limitations. Furthermore, certificate the influence of model inaccuracy on the Kalman filter, which illustrates an effective method for reducing errors is increasing model numbers. Thirdly, the definition of model set density and the analysis of the true model space are given, and their connection establishes the standard of how to design the model set or add the model number. Finally, an effective method called RIMM is provided to overcome the competition caused by model increasing. The proposed RIMM holds strong adaptability for different model sets. The simulations of RIMM highlight the correctness and effectiveness of the proposed methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10462948/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10462948/</guid>
</item>
<item>
<title>A Stretchable and Wearable Ultrasonic Transducer Array for Bladder Volume Monitoring Application</title>
<description><p><span><big>A Stretchable and Wearable Ultrasonic Transducer Array for Bladder Volume Monitoring Application</big></span><br></p><p><span><small><i>Cong Pu; Ben Fu; Lehang Guo; Huixiong Xu; Chang Peng</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3382244</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Monitoring bladder volume is an essential application for patients with voiding dysfunction. Current wearable ultrasonic bladder monitors are rigid, bulky and cannot conform to human skin. This study presents a stretchable and wearable ultrasonic transducer array that can both conform to non-planar skin surfaces and continuously monitor bladder volume. The wearable transducer array mainly consists of 4 × 4 piezoelectric transducer elements, stretchable serpentine-based electrodes for electrical interconnection, electrically conductive adhesive as both matching and backing layers, and silicone elastomer for encapsulation. The transducer array has a center frequency of 3.9 MHz, -6 dB acoustic bandwidth of 57.6%, and up to 40% elastic stretchability. The volume of balloon-bladder phantom was estimated by a least square ellipsoid fitting method. With the true volume of balloon-bladder phantoms ranging from 10 mL to 405 mL, the proposed transducer array has a mean absolute percentage error of 9.4%, a mean absolute error of 11.9 mL and a coefficient of determination of 0.975, which demonstrates the capability of the proposed stretchable and wearable ultrasonic transducer array for bladder volume monitoring application.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10489841/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10489841/</guid>
</item>
<item>
<title>Intelligent fault diagnosis of bearing using multiwavelet perception kernel convolutional neural network</title>
<description><p><span><big>Intelligent fault diagnosis of bearing using multiwavelet perception kernel convolutional neural network</big></span><br></p><p><span><small><i>Yuanyuan Zhou; Hang Wang; Yongbin Liu; Xianzeng Liu; Zheng Cao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3370564</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Strong background noise characteristics of vibration signals cause issues with poor identification capability of features by fault diagnostic models. To address this issue, a method is proposed for intelligent fault diagnosis of bearing using multiwavelet perception kernel (MPK) and feature attention convolutional neural network (FA-CNN). First, four multiwavelet perception kernels are constructed to decompose the vibration signals in full-band multilevel. Second, improved multiwavelet information entropy (IMIE) of the frequency band components is calculated. The calculated component entropies of the corresponding frequency bands are integrated to construct frequency band clusters (FBC) from low to high frequencies. Third, joint approximate diagonalization eigen (JADE) is introduced to perform feature fusion for every FBC to eliminate redundant information, and fused features from low to high frequencies are obtained as original inputs. The FA-CNN bearing fault diagnosis framework is constructed for intelligent fault diagnosis of bearings. Finally, the effectiveness of the proposed method is verified by two cases. The results show that the proposed method has high fault feature recognition capability.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10458914/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10458914/</guid>
</item>
<item>
<title>Improved multipath mitigation using multiple trend-surface hemispherical map in GNSS precise point positioning</title>
<description><p><span><big>Improved multipath mitigation using multiple trend-surface hemispherical map in GNSS precise point positioning</big></span><br></p><p><span><small><i>Haijun Yuan; Zhetao Zhang; Xiufeng He; Jinwen Zeng; Hao Wang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3374458</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Global Navigation Satellite System (GNSS) Precise Point Positioning (PPP) is highly appreciated as a positioning technology; however, multipath restricts its accuracy and reliability. In this study, we proposed a Multiple Trend-surface Multipath Hemispherical Map (MT-MHM), where discrepant multipath fading frequency or orbit orientation of each satellite in each azimuth and elevation grid is considered. In a typical static scenario, consecutive 8-day observations are collected to extract multipath and conduct PPP experiments. Compared with traditional multipath hemispherical map based on trend-surface fitting, MT-MHM has better modelling performance of residuals and improves standard deviations of residuals for all satellites. Besides, the positioning errors and required convergence epochs of PPP are reduced by using MT-MHM. For the positioning accuracy of all solutions, MT-MHM exhibits 50.8, 15.2, and 27.9% improvements compared with that without multipath correction in the east, north, and up directions, respectively. In conclusion, our proposed MT-MHM exhibits better performance in terms of residual reduction, convergence time, and positioning accuracy in PPP.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472420/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472420/</guid>
</item>
<item>
<title>Design optimization and characterization of a 3D-printed tactile sensor for tissue palpation</title>
<description><p><span><big>Design optimization and characterization of a 3D-printed tactile sensor for tissue palpation</big></span><br></p><p><span><small><i>D. Lo Presti; L. Zoboli; A. Addabbo; D. Bianchi; A. Dimo; C. Massaroni; V. Altomare; A. Grasso; A. Gizzi; E. Schena</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3369337</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Manual palpation is a crucial medical procedure that relies on surface examination to detect internal tissue abnormalities, heavily reliant on healthcare professionals’ expertise and tactile sensitivity. To tackle these issues, smart palpation systems based on electrical or optical sensors have been developed to perform quantitative tactile measurements, crucial for identifying various solid tumors, including breast and prostate cancer by assessing tissue mechanical properties. In this context, fiber Bragg gratings (FBGs) are emerging as a promising technology due to their advantages (e.g., high metrological properties, multiplexing capacity, and easy packaging) making them ideal for tactile sensing. This study explores the benefits of FBG and 3D printing to develop a tactile sensor for tissue palpation. First, an optimization of the design of the sensing core of a previously developed probe was conducted through finite element analysis. The novel structure addresses the primary limitation of the previous solution, where non-uniform strain distribution on the encapsulated FBG causes compression on the grating with high risk of bending and breakage. In contrast, the modeled geometry ensures FBG elongation during tissue palpation. A 3D printing and characterization of the proposed solution was carried out to investigate the response of the enclosed FBG when pushed against different materials showing promising results in discriminating tissues according to their mechanical properties: the more rigid the indented substrate the higher the sensor output. This property will be fundamental for enhancing early tumor detection through superficial tissue palpation, advancing the efficacy of prevention measures.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10455966/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10455966/</guid>
</item>
<item>
<title>Deep Reinforcement Learning-based Joint Sequence Scheduling and Trajectory Planning in Wireless Rechargeable Sensor Networks</title>
<description><p><span><big>Deep Reinforcement Learning-based Joint Sequence Scheduling and Trajectory Planning in Wireless Rechargeable Sensor Networks</big></span><br></p><p><span><small><i>Chengpeng Jiang; Wencong Chen; Ziyang Wang; Wendong Xiao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373664</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Mobile charging has become a popular and efficient method for replenishing energy. This is done with a mobile charger (MC) and wireless energy transfer technology (WET), which helps to alleviate the issue of energy constraints in wireless rechargeable sensor networks (WRSNs). Notably, designing mobile charging scheduling schemes is essential for improving charging performance. Most current studies assume that the networks are obstacle-free. Unlike the existing studies, this paper focuses on joint sequence scheduling and trajectory planning problems (JSSTP), which assumes that the network has multiple static obstacles. To address this issue, we propose a novel deep reinforcement learning-based JSSTP (DRL-JSSTP) that enables the MC to avoid obstacles and reach the charging target to charge the sensors fully. This approach maximizes energy usage efficiency and sensor survival rate while satisfying MC energy capacity constraints. DRL-JSSTP includes a charging target selector and a trajectory planner, which determine the index of the next charging target and plan the movement trajectory to avoid obstacles, respectively. We adopt 1-D convolutional neural networks to extract feature information about the environment state and gated recurrent units to predict the charging decisions. Simulation results demonstrate that DRL-JSSTP outperforms existing approaches, achieving higher energy usage efficiency and sensor survival rate.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466524/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466524/</guid>
</item>
<item>
<title>A Simultaneous Wirel |
Derekmini
changed the title
got to ofetch in /ieee/journal.ts and /ieee/earlyaccess.tsgot to ofetch in /ieee/journal.ts, /ieee/author.ts and /ieee/earlyaccess.ts
|
Successfully generated as following: http://localhost:1200/ieee/journal/8782710 - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Open Journal of Signal Processing</title>
<link>https://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=8782710</link>
<atom:link href="http://localhost:1200/ieee/journal/8782710" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Open Journal of Signal Processing - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Wed, 10 Apr 2024 15:20:05 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Synthbuster: Towards Detection of Diffusion Model Generated Images</title>
<description><p><span><big>Synthbuster: Towards Detection of Diffusion Model Generated Images</big></span><br></p><p><span><small><i>Quentin Bammey</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337714</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Synthetically-generated images are getting increasingly popular. Diffusion models have advanced to the stage where even non-experts can generate photo-realistic images from a simple text prompt. They expand creative horizons but also open a Pandora's box of potential disinformation risks. In this context, the present corpus of synthetic image detection techniques, primarily focusing on older generative models like Generative Adversarial Networks, finds itself ill-equipped to deal with this emerging trend. Recognizing this challenge, we introduce a method specifically designed to detect synthetic images produced by diffusion models. Our approach capitalizes on the inherent frequency artefacts left behind during the diffusion process. Spectral analysis is used to highlight the artefacts in the Fourier transform of a residual image, which are used to distinguish real from fake images. The proposed method can detect diffusion-model-generated images even under mild
jpeg
compression, and generalizes relatively well to unknown models. By pioneering this novel approach, we aim to fortify forensic methodologies and ignite further research into the detection of AI-generated images.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334046/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334046/</guid>
</item>
<item>
<title>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</title>
<description><p><span><big>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</big></span><br></p><p><span><small><i>Yanbin Zou; Jingna Fan; Zekai Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we consider using time-of-arrival (TOA) measurements from a single moving receiver to locate a moving target at constant velocity that emits a periodic signal with unknown signal period. First, we give the TOA measurement model and deduce the Cram
$\acute{\text{e}}$
r-Rao lower bounds (CRLB). Then, we formulate a nonlinear least squares (NLS) problem to estimate the unknown parameters. We use semidefinite programming (SDP) techniques to relax the nonconvex NLS problem. However, it is shown that the SDP localization algorithm cannot provide a high-quality solution. Subsequently, we develop a fixed point iteration (FPI) method to improve the performance of the SDP algorithm. In addition, we also consider the presence of receiver position errors and develop the corresponding localization algorithm. Numerical simulations are conducted to demonstrate the localization performance of the proposed algorithms by comparing them with the CRLB.
Index Term
-Fixed point iteration (FPI), semidefinite programming (SDP), single moving receiver, target localization, time-of-arrival (TOA).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336384/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336384/</guid>
</item>
<item>
<title>The Neural-SRP Method for Universal Robust Multi-Source Tracking</title>
<description><p><span><big>The Neural-SRP Method for Universal Robust Multi-Source Tracking</big></span><br></p><p><span><small><i>Eric Grinstein; Christopher M. Hicks; Toon van Waterschoot; Mike Brookes; Patrick A. Naylor</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340057</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Neural networks have achieved state-of-the-art performance on the task of acoustic Direction-of-Arrival (DOA) estimation using microphone arrays. Neural models can be classified as end-to-end or hybrid, each class showing advantages and disadvantages. This work introduces Neural-SRP, an end-to-end neural network architecture for DOA estimation inspired by the classical Steered Response Power (SRP) method, which overcomes limitations of current neural models. We evaluate the architecture on multiple scenarios, namely, multi-source DOA tracking and single-source DOA tracking under the presence of directional and diffuse noise. The experiments demonstrate that our proposed method compares favourably in terms of computational and localization performance with established neural methods on various recorded and simulated benchmark datasets.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345765/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345765/</guid>
</item>
<item>
<title>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</title>
<description><p><span><big>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</big></span><br></p><p><span><small><i>Alexander Venus; Erik Leitinger; Stefan Tertinek; Klaus Witrisal</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper presents a neural-enhanced probabilistic model and corresponding factor graph-based sum-product algorithm for robust localization and tracking in multipath-prone environments. The introduced hybrid probabilistic model consists of physics-based and data-driven measurement models capturing the information contained in both, the line-of-sight (LOS) component as well as in multipath components (NLOS components). The physics-based and data-driven models are embedded in a joint Bayesian framework allowing to derive from first principles a factor graph-based algorithm that fuses the information of these models. The proposed algorithm uses radio signal measurements from multiple base stations to robustly estimate the mobile agent's position together with all model parameters. It provides high localization accuracy by exploiting the position-related information of the LOS component via the physics-based model and robustness by exploiting the geometric imprint of multipath components independent of the propagation channel via the data-driven model. In a challenging numerical experiment involving obstructed LOS situations to all anchors, we show that the proposed sequential algorithm significantly outperforms state-of-the-art methods and attains the posterior Cramér-Rao lower bound even with training data limited to local regions.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336409/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336409/</guid>
</item>
<item>
<title>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</title>
<description><p><span><big>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</big></span><br></p><p><span><small><i>AMIT EFRAIM; Joseph M. Francos</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In solving a model fitting problem, the existence of outliers in the set of measurements can have a devastating effect on the solution accuracy. Traditionally, in order to overcome this problem, robust point cloud registration algorithms are composed of transformation hypothesis generation, followed by hypothesis evaluation aimed at selecting the best hypothesized estimate. Hypotheses evaluation is commonly performed using the sample consensus criterion. However, since this criterion accounts only for the consensus size, it fails when the maximal sample consensus is incorrect. We propose a new hypothesis evaluation approach, generalizing the sample consensus approach, where instead of the sample consensus, the transformation that maximizes the point clouds feature correlation is selected as the best hypothesis. The feature vector at each point contains information such as on local geometry and semantic context. Utilizing this information in the hypotheses evaluation and selection procedure allows for a correct decision even when the hypothesis yielding the maximal sample consensus is false. Consequently, the probability of selecting the correct model increases. We show both mathematically and empirically that substituting the sample consensus criterion with the proposed point cloud feature correlation hypothesis test (PC-FCHT) lowers the probability of large registration errors, compared to using the special case of sample consensus. The proposed PC-FCHT is applicable to any algorithm that follows the hypothesis generation and evaluation scheme, potentially improving the success rates of a wide family of point cloud registration algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345750/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345750/</guid>
</item>
<item>
<title>Joint PAPR and OBP Reduction for NC-OFDM Systems</title>
<description><p><span><big>Joint PAPR and OBP Reduction for NC-OFDM Systems</big></span><br></p><p><span><small><i>Hsuan-Fu Wang; Fang-Biau Ueng; Bo-Heng Yeh</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3329757</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The spectrum resource is always a critical issue for wireless communications since it directly impacts the data rate and capacity. However, the problem of spectrum resource scarcity always exists. Moreover, spectrum resource scarcity becomes more severe as new communication technologies and wireless applications sprout. Noncontiguous orthogonal frequency division multiplexing (NC-OFDM) is a multicarrier method for bandwidth utilization. Unfortunately, this system has two fatal defects: high peak-to-average power ratio (PAPR) and considerable out-of-band power (OBP), which are detrimental to the system's performance. To solve these two problems, we propose a convex optimization-based method for joint PAPR and OBP reduction in NC-OFDM Systems. The strategy is to permit the secondary user to utilize the unoccupied spectrum of the primary user with dynamic spectrum sharing (DSS) based on a cognitive radio network (CRN). To this end, a flexible system operating over noncontiguous bands and DSS scenarios is necessary. The simulation results have shown that our method could effectively improve the overall performance and outperform other schemes, i.e., projections onto convex sets (POCS) and alternating projections onto convex and non-convex sets (APOCNCS), without harming the transmission of the primary system. The collaboration between secondary and primary systems is viable with the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10305259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10305259/</guid>
</item>
<item>
<title>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</title>
<description><p><span><big>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</big></span><br></p><p><span><small><i>Gal Itzhak; Israel Cohen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339433</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article presents a Kronecker-product (KP) beamforming approach incorporating sparse concentric circular arrays (SCCAs). The locations of the microphones on the SCCA are optimized concerning the broadband array directivity over a wide range of direction-of-arrival (DOA) deviations of a desired signal. A maximum directivity factor (MDF) sub-beamformer is derived accordingly with the optimal locations. Then, we propose two global beamformers obtained as a Kronecker product of a uniform linear array (ULA) and the SCCA sub-beamformer. The global beamformers differ by the type of the ULA, which is designed either as an MDF sub-beamformer along the
$\mathsf {x}$
-axis or as a maximum white noise gain sub-beamformer along the
$\mathsf {y}$
-axis. We analyze the performance of the proposed beamformers in terms of the directivity factor, the white noise gain, and their spatial beampatterns. Compared to traditional beamformers, the proposed beamformers exhibit considerably larger tolerance to DOA deviations concerning both the azimuth and elevation angles. Experimental results with speech signals in noisy and reverberant environments demonstrate that the proposed approach outperforms traditional beamformers regarding the perceptual evaluation of speech quality (PESQ) and short-time objective intelligibility (STOI) scores when the desired speech signals deviate from the nominal DOA.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342869/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342869/</guid>
</item>
<item>
<title>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</title>
<description><p><span><big>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</big></span><br></p><p><span><small><i>Karn N. Watcharasupat; Chih-Wei Wu; Yiwei Ding; Iroro Orife; Aaron J. Hipple; Phillip A. Williams; Scott Kramer; Alexander Lerch; William Wolcott</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339428</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Cinematic audio source separation is a relatively new subtask of audio source separation, with the aim of extracting the dialogue, music, and effects stems from their mixture. In this work, we developed a model generalizing the Bandsplit RNN for any complete or overcomplete partitions of the frequency axis. Psychoacoustically motivated frequency scales were used to inform the band definitions which are now defined with redundancy for more reliable feature extraction. A loss function motivated by the signal-to-noise ratio and the sparsity-promoting property of the 1-norm was proposed. We additionally exploit the information-sharing property of a common-encoder setup to reduce computational complexity during both training and inference, improve separation performance for hard-to-generalize classes of sounds, and allow flexibility during inference time with detachable decoders. Our best model sets the state of the art on the Divide and Remaster dataset with performance above the ideal ratio mask for the dialogue stem.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342812/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342812/</guid>
</item>
<item>
<title>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</title>
<description><p><span><big>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</big></span><br></p><p><span><small><i>Miguel Ferrer; María de Diego; Alberto Gonzalez</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340106</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The LMS algorithm is widely employed in adaptive systems due to its robustness, simplicity, and reasonable performance. However, it is well known that this algorithm suffers from a slow convergence speed when dealing with colored reference signals. Numerous variants and alternative algorithms have been proposed to address this issue, though all of them entail an increase in computational cost. Among the proposed alternatives, the affine projection algorithm stands out. This algorithm has the peculiarity of starting from
$N$
data vectors of the reference signal. It transforms these vectors into as many data vectors suitably normalized in energy and mutually orthogonal. In this work, we propose a version of the LMS algorithm that, similar to the affine projection algorithm, starts from
$N$
data vectors of the reference signal but corrects them by using only a scalar factor that functions as a convergence step. Our goal is to align the behavior of this algorithm with the behavior of the affine projection algorithm without significantly increasing the computational cost of the LMS.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345730/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345730/</guid>
</item>
<item>
<title>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</title>
<description><p><span><big>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</big></span><br></p><p><span><small><i>Zongyao Li; Ren Togo; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340616</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article proposes a method for transferring knowledge of semantic segmentation from a labeled source domain to an unlabeled target domain without using the source-domain data. Such a problem is called source-data-free domain adaptation, in which a pre-trained source-domain model and the unlabeled target-domain data are used to transfer the label knowledge across the domains. Like most previous methods, our method uses pseudo labels for distilling and transferring the source-domain knowledge. On the basis of the pseudo-label learning, our method improves the domain adaptation performance in two innovative ways: 1) reducing the domain differences by source-data-free style transfer and 2) exploring the style diversity within the target domain by style modification. To this end, we introduce two additional modules: 1) an inter-domain style transfer module which aligns the feature statistics of the source and target domains before producing the pseudo labels thereby improving the pseudo labels' accuracy, and 2) an intra-domain style modification module which modifies the image styles within the target domain for learning intra-domain style-invariant features. Our method with the two modules outperforms previous source-data-free domain adaptation methods in two commonly used benchmarks. Moreover, our method is well compatible with the previous methods for further improvement.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10356748/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10356748/</guid>
</item>
<item>
<title>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</title>
<description><p><span><big>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</big></span><br></p><p><span><small><i>Tobias Kabzinski; Peter Jax</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337721</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Quasi-continuous system identification of time-variant linear acoustic systems can be applied in various audio signal processing applications when numerous acoustic transfer functions must be measured. A prominent application is measuring head-related transfer functions. We treat the underlying multiple-input-multiple-output (MIMO) system identification problem in a state-space model as a joint estimation problem for states, representing impulse responses, and state-space model parameters using the expectation maximization (EM) algorithm. We address limitations of prior work by imposing different model structures, especially for dependencies within a (transformed) state vector. This results in block diagonal matrix structures, for which we derive M-step update rules. Making assumptions about this model structure and choosing a block size for a given application define the computational complexity. In examples, we found that applying this framework yields improvements of up to 10 dB in relative system distance in comparison to a conventional method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334061/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334061/</guid>
</item>
<item>
<title>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</title>
<description><p><span><big>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</big></span><br></p><p><span><small><i>Damir Rakhimov; Martin Haardt</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337729</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this paper, we present an analytical performance assessment of 2-D Tensor ESPRIT in terms of physical parameters. We show that the error in the
$r$
-mode depends only on two components, irrespective of the dimensionality of the problem. We obtain analytical expressions in closed form for the mean squared error (MSE) in each dimension as a function of the signal-to-noise (SNR) ratio, the array steering matrices, the number of antennas, the number of snapshots, the selection matrices, and the signal correlation. The derived expressions allow a better understanding of the difference in performance between the tensor and the matrix versions of the ESPRIT algorithm. The simulation results confirm the coincidence between the presented analytical expression and the curves obtained via Monte Carlo trials. We analyze the behavior of each of the two error components in different scenarios.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334446/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334446/</guid>
</item>
<item>
<title>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</title>
<description><p><span><big>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</big></span><br></p><p><span><small><i>Yaman Kındap; Simon Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343341</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this work we study linear vector stochastic differential equation (SDE) models driven by the generalised hyperbolic (GH) Lévy process for inference in continuous-time non-Gaussian filtering problems. The GH family of stochastic processes offers a flexible framework for modelling of non-Gaussian, heavy-tailed characteristics and includes the normal inverse-Gaussian, variance-gamma and Student-t processes as special cases. We present continuous-time simulation methods for the solution of vector SDE models driven by GH processes and novel inference methodologies using a variant of sequential Markov chain Monte Carlo (MCMC). As an example a particular formulation of Langevin dynamics is studied within this framework. The model is applied to both a synthetically generated data set and a real-world financial series to demonstrate its capabilities.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360268/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360268/</guid>
</item>
<item>
<title>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</title>
<description><p><span><big>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</big></span><br></p><p><span><small><i>James M. Cozens; Simon J. Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344048</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper proposes a probabilistic approach for extracting time-varying and irregular time signature information from polyphonic audio extracts, subsequently providing beat and bar line positions given inferred time signature divisions. This is achieved via dynamically evaluating the beat tempo as a function of time through finding an optimal compromise in beat and bar alignment in the time and tempo domains. Time signature divisions are determined based on a new representation, termed the Metrogram, that presents time-varying information regarding rhythmic and metric periodicities in the Tempogram. Our methodology is characterised by its ability to provide a distribution over metric interpretations, offering insights into the diverse ways music can be rhythmically perceived. Results indicate high-level accuracy for a variety of polyphonic extracts containing irregular, complex, irrational, and time-varying time signatures. Accuracy rivalling state-of-the-art methodologies is also reported in a beat tracking task performed on the standard Ballroom Dataset. The paper offers insights into the field of dynamic time signature recognition and beat tracking, offering a valuable and versatile resource for the analysis, composition, and performance of music.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363392/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363392/</guid>
</item>
<item>
<title>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</title>
<description><p><span><big>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</big></span><br></p><p><span><small><i>Yuto Watanabe; Ren Togo; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343335</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Although text-guided image manipulation approaches have demonstrated highly accurate performance for editing the appearance of images in a virtual or simple scenario, their real-world applications face significant challenges. The primary cause of these challenges is the misalignment in the distribution of training and real-world data, which leads to unstable text-guided image manipulation. In this work, we propose a novel framework called TolerantGAN and tackle the new task of real-world text-guided image manipulation independent of the training data. To achieve this, we introduce two key concepts of a border smoothly connection module (BSCM) and a manipulation direction-based attention module (MDAM). BSCM smoothens the misalignment in the distribution of training and real-world data. MDAM extracts only regions highly relevant for image manipulation and assists in reconstructing unobserved objects in the training data. For in-the-wild input images of various classes, TolerantGAN robustly outperforms the state-of-the-art methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360283/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360283/</guid>
</item>
<item>
<title>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</title>
<description><p><span><big>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</big></span><br></p><p><span><small><i>Anastasia Avdeeva; Aleksei Gusev; Tseren Andzhukaev; Artem Ivanov</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343342</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Whispered speech is a quiet voice without vocalization. One of the common cases of using whispered speech is a technique that can help overcome stuttering. But whispered speech can be uncomfortable and difficult to understand in everyday communication. To address these problems, we propose a method of low-delayed whisper-to-speech voice conversion, which can be useful in real life communication of people with disordered speech. As part of our research, we study the impact of streaming Automatic Speech Recognition models on the quality of voice conversion, comparing different streaming models and methods for model adaptation to streaming settings, and showing the importance of using such models in cases of low-delayed voice conversion.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360259/</guid>
</item>
<item>
<title>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</title>
<description><p><span><big>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</big></span><br></p><p><span><small><i>Denis C. Ilie-Ablachim; Andra Băltoiu; Bogdan Dumitrescu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344313</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We propose sparse representation and dictionary learning algorithms for dictionaries whose atoms are characterized by Gaussian probability distributions around some central atoms. This extension of the space covered by the atoms permits a better characterization of localized features of the signals. The representations are computed by trading-off representation error and the probability of the actual atoms used in the representation. We propose two types of algorithms: a greedy one, similar in spirit with Orthogonal Matching Pursuit, and one based on L1-regularization. We apply our new algorithms to unsupervised anomaly detection, where the representation error is used as anomaly score. The flexibility of our approach appears to improve more the representations of the many normal signals than those of the few outliers, at least for an anomaly type called dependency, thus improving the detection quality. Comparison with both standard dictionary learning algorithms and established anomaly detection methods is favorable.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365333/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365333/</guid>
</item>
<item>
<title>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</title>
<description><p><span><big>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</big></span><br></p><p><span><small><i>Yuya Moroto; Yingrui Ye; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344079</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>There are various sentiment theories for categorizing human sentiments into several discrete sentiment categories, which means that the theory used for training sentiment prediction methods does not always match that used in the test phase. As a solution to this problem, zero-shot visual sentiment prediction methods have been proposed to predict unseen sentiments for which no images are available in the training phase. However, the training of these previous zero-shot methods relies on a single sentiment theory, which limits their ability to handle sentiments from other theories. Thus, this article proposes a more robust zero-shot visual sentiment prediction method that can handle cross-domain sentiments defined in different sentiment theories. Specifically, by focusing on the fact that sentiments are abstract concepts common to humans regardless of whether their theories are different, we incorporate knowledge distillation into our method to construct a teacher–student model that can train the implicit relationships between sentiments defined in different sentiment theories. Furthermore, to enhance sentiment discrimination capability and strengthen the implicit relationships between sentiments, we introduce a novel sentiment loss between the teacher and student models. In this way, our model becomes robust to unseen sentiments by exploiting the implicit relationships between sentiments. The contributions of this article are the introduction of knowledge distillation and a novel sentiment loss between the teacher and student models for zero-shot visual sentiment prediction, and improved performance of zero-shot visual sentiment prediction. Experiments on several open datasets demonstrate the effectiveness of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363382/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363382/</guid>
</item>
<item>
<title>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</title>
<description><p><span><big>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</big></span><br></p><p><span><small><i>Chengen Liu; Geert Leus; Elvin Isufi</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339376</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The edge flow reconstruction task consists of retreiving edge flow signals from corrupted or incomplete measurements. This is typically solved by a regularized optimization problem on higher-order networks such as simplicial complexes and the corresponding regularizers are chosen based on prior knowledge. Tailoring this prior to the setting of interest can be challenging or it may not even be possible. Thus, we consider to learn this prior knowledge via a model-based deep learning approach. We propose a new regularized optimization problem for the simplicial edge flow reconstruction task, the simplicial ElasticNet, which combines the advantages of the
$\ell _{1}$
and
$\ell _{2}$
norms. We solve the simplicial ElasticNet problem via the multi-block alternating direction method of multipliers (ADMM) algorithm and provide conditions on its convergence. By unrolling the ADMM iterative steps, we develop a model-based neural network with a low requirement on the number of training data. This unrolling network replaces the fixed parameters in the iterative algorithm by learnable weights, thus exploiting the neural network's learning capability while preserving the iterative algorithm's interpretability. We enhance this unrolling network via simplicial convolutional filters to aggregate information from the edge flow neighbors, ultimately, improving the network learning expressivity. Extensive experiments on real-world and synthetic datasets validate the proposed approaches and show considerable improvements over both baselines and traditional non-model-based neural networks.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342735/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342735/</guid>
</item>
<item>
<title>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</title>
<description><p><span><big>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</big></span><br></p><p><span><small><i>Jim Beckers; Bart Van Erp; Ziyue Zhao; Kirill Kondrashov; Bert De Vries</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337718</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Bayesian model reduction provides an efficient approach for comparing the performance of all nested sub-models of a model, without re-evaluating any of these sub-models. Until now, Bayesian model reduction has been applied mainly in the computational neuroscience community on simple models. In this paper, we formulate and apply Bayesian model reduction to perform principled pruning of Bayesian neural networks, based on variational free energy minimization. Direct application of Bayesian model reduction, however, gives rise to approximation errors. Therefore, a novel iterative pruning algorithm is presented to alleviate the problems arising with naive Bayesian model reduction, as supported experimentally on the publicly available UCI datasets for different inference algorithms. This novel parameter pruning scheme solves the shortcomings of current state-of-the-art pruning methods that are used by the signal processing community. The proposed approach has a clear stopping criterion and minimizes the same objective that is used during training. Next to these benefits, our experiments indicate better model performance in comparison to state-of-the-art pruning schemes.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334001/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334001/</guid>
</item>
<item>
<title>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</title>
<description><p><span><big>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</big></span><br></p><p><span><small><i>Anh Minh Truong; Wilfried Philips; Peter Veelaert</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340064</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Many image-based rendering (IBR) methods rely on depth estimates obtained from structured light or time-of-flight depth sensors to synthesize novel views from sparse camera networks. However, these estimates often contain missing or noisy regions, resulting in an incorrect mapping between source and target views. This situation makes the fusion process more challenging, as the visual information is misaligned, inconsistent, or missing. In this work, we first implement a lightweight network based on the transformer, which is well-known for its capability to model long-range relationships within input data, to extract spatial features from color images. These features are then used to enhance the quality of completed depth maps. Furthermore, we combine a sequential deep neural network with a spatial attention mechanism to effectively fuse the projected features from multiple source viewpoints. This approach enables us to integrate information from an arbitrary number of source viewpoints as well as improve accuracy in synthesized views. Experimental results on challenging datasets demonstrate that our method achieves superior synthesized image quality compared to state-of-the-art (SOTA) methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345792/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345792/</guid>
</item>
<item>
<title>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</title>
<description><p><span><big>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</big></span><br></p><p><span><small><i>Reza Mirzaeifard; Naveen K. D. Venkategowda; Vinay Chakravarthi Gogineni; Stefan Werner</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344395</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper investigates quantile regression in the presence of non-convex and non-smooth sparse penalties, such as the minimax concave penalty (MCP) and smoothly clipped absolute deviation (SCAD). The non-smooth and non-convex nature of these problems often leads to convergence difficulties for many algorithms. While iterative techniques such as coordinate descent and local linear approximation can facilitate convergence, the process is often slow. This sluggish pace is primarily due to the need to run these approximation techniques until full convergence at each step, a requirement we term as a
secondary convergence iteration
. To accelerate the convergence speed, we employ the alternating direction method of multipliers (ADMM) and introduce a novel single-loop smoothing ADMM algorithm with an increasing penalty parameter, named SIAD, specifically tailored for sparse-penalized quantile regression. We first delve into the convergence properties of the proposed SIAD algorithm and establish the necessary conditions for convergence. Theoretically, we confirm a convergence rate of
$o({k^{-\frac{1}{4}}})$
for the sub-gradient bound of the augmented Lagrangian, where
$k$
denotes the number of iterations. Subsequently, we provide numerical results to showcase the effectiveness of the SIAD algorithm. Our findings highlight that the SIAD method outperforms existing approaches, providing a faster and more stable solution for sparse-penalized quantile regression.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365338/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365338/</guid>
</item>
<item>
<title>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</title>
<description><p><span><big>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</big></span><br></p><p><span><small><i>Oliver Lang; Christian Hofbauer; Reinhard Feger; Mario Huemer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343308</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>A promising waveform candidate for future joint sensing and communication systems is orthogonal frequency-division multiplexing (OFDM). For such systems, supporting multiple transmit antennas requires multiplexing methods for the generation of orthogonal transmit signals, where equidistant subcarrier interleaving (ESI) is the most popular multiplexing method. In this work, we analyze a multiplexing method called Doppler-division multiplexing (DDM). This method applies a phase shift from OFDM symbol to OFDM symbol to separate signals transmitted by different Tx antennas along the velocity axis of the range-Doppler map. The main focus of this work lies on the implications of DDM on the communication task. It will be shown that for DDM, the channels observed in the communication receiver are heavily time-varying, preventing any meaningful transmission of data when not taken into account. In this work, a communication system designed to combat these time-varying channels is proposed, which includes methods for data estimation, synchronization, and channel estimation. Bit error ratio (BER) simulations demonstrate the superiority of this communications system compared to ESI-based systems.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360223/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360223/</guid>
</item>
<item>
<title>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</title>
<description><p><span><big>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</big></span><br></p><p><span><small><i>Nils L. Westhausen; Bernd T. Meyer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343320</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we introduce a causal low-latency low-complexity approach for binaural multichannel blind speaker separation in noisy reverberant conditions. The model, referred to as Group Communication Binaural Filter and Sum Network (GCBFSnet) predicts complex filters for filter-and-sum beamforming in the time-frequency domain. We apply Group Communication (GC), i.e., latent model variables are split into groups and processed with a shared sequence model with the aim of reducing the complexity of a simple model only containing one convolutional and one recurrent module. With GC we are able to reduce the size of the model by up to 83% and the complexity up to 73% compared to the model without GC, while mostly retaining performance. Even for the smallest model configuration, GCBFSnet matches the performance of a low-complexity TasNet baseline in most metrics despite the larger size and higher number of required operations of the baseline.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360275/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360275/</guid>
</item>
<item>
<title>Reverse Ordering Techniques for Attention-Based Channel Prediction</title>
<description><p><span><big>Reverse Ordering Techniques for Attention-Based Channel Prediction</big></span><br></p><p><span><small><i>Valentina Rizzello; Benedikt Böck; Michael Joham; Wolfgang Utschick</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344024</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Channel state information (CSI) is crucial for enhancing the performance of wireless systems by allowing to adjust the transmission strategies based on the current channel conditions. However, obtaining precise CSI is difficult because of the fast-changing channel conditions caused by multi-path fading. An inaccurate CSI hinders the performance of various adaptive wireless systems, highlighting the need for channel prediction techniques to effectively mitigate the drawbacks of outdated CSI. Conventional methods typically depend on assumptions regarding user velocity or require knowledge of the Doppler frequency. In contrast to existing approaches, we aim for a more robust and practical solution by training neural networks without making any assumptions about user velocity, relying solely on noisy channel observations during training. Specifically, we adapt both the sequence-to-sequence with attention (Seq2Seq-attn) and transformer models for channel prediction. Additionally, a new technique called reverse positional encoding is introduced in the transformer model to improve the robustness of the model against varying sequence lengths. Similarly, the encoder outputs of the Seq2Seq-attn model are reversed prior to the application of attention mechanisms. By means of simulations, we show that these proposed techniques enable the models to effectively capture relationships within sequences of channel snapshots without increasing the complexity. Importantly, this capability remains robust across varying sequence lengths, representing a substantial improvement over existing methodologies.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363354/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363354/</guid>
</item>
<item>
<title>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</title>
<description><p><span><big>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</big></span><br></p><p><span><small><i>Sarina Meyer; Xiaoxiao Miao; Ngoc Thang Vu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344375</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Speaker anonymization is the task of modifying a speech recording such that the original speaker cannot be identified anymore. Since the first Voice Privacy Challenge in 2020, along with the release of a framework, the popularity of this research topic is continually increasing. However, the comparison and combination of different anonymization approaches remains challenging due to the complexity of evaluation and the absence of user-friendly research frameworks. We therefore propose an efficient speaker anonymization and evaluation framework based on a modular and easily extendable structure, almost fully in Python. The framework facilitates the orchestration of several anonymization approaches in parallel and allows for interfacing between different techniques. Furthermore, we propose modifications to common evaluation methods which improves the quality of the evaluation and reduces their computation time by 65 to 95%, depending on the metric. Our code is fully open source.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365329/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365329/</guid>
</item>
<item>
<title>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</title>
<description><p><span><big>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</big></span><br></p><p><span><small><i>Alessandro Ilic Mezza; Matteo Amerena; Alberto Bernardini; Augusto Sarti</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343318</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Real-time audio communications over IP have become essential to our daily lives. Packet-switched networks, however, are inherently prone to jitter and data losses, thus creating a strong need for effective packet loss concealment (PLC) techniques. Though solutions based on deep learning have made significant progress in that direction as far as speech is concerned, extending the use of such methods to applications of Networked Music Performance (NMP) presents significant challenges, including high fidelity requirements, higher sampling rates, and stringent temporal constraints associated to the simultaneous interaction between remote musicians. In this article, we present PARCnet, a hybrid PLC method that utilizes a feed-forward neural network to estimate the time-domain residual signal of a parallel linear autoregressive model. Objective metrics and a listening test show that PARCnet provides state-of-the-art results while enabling real-time operation on CPU.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360264/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360264/</guid>
</item>
<item>
<title>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</title>
<description><p><span><big>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</big></span><br></p><p><span><small><i>Katerina Zmolikova; Michael Syskind Pedersen; Jesper Jensen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Supervised learning-based speech enhancement methods often work remarkably well in acoustic situations represented in the training corpus but generalize poorly to out-of-domain situations, i.e. situations not seen during training. This stands in the way of further improvement of these methods in realistic scenarios, as collecting paired noisy-clean recordings in the target application domain is typically not feasible. Recording noisy-only in-domain data is, though, much more practical. In this article, we tackle the problem of unsupervised domain adaptation in speech enhancement. Specifically, we propose a way to use in-domain noisy-only data in the training of a neural network to improve upon a model trained solely on out-of-domain paired data. For this, we make use of masked spectrogram prediction, a technique from self-supervised learning that aims to interpolate masked regions of a spectrogram. We hypothesize that masked spectrogram prediction encourages learning of features that represent well both speech and noise components of the noisy signals. These features can then be used to train a more robust speech enhancement system. We evaluate the proposed method on the VoiceBank-DEMAND and LibriFSD50k databases, with WSJ0-CHiME3 serving as the out-of-domain database. We show that the proposed method outperforms both the out-of-domain system and the baseline approaches, i.e. RemixIT and noisy-target training, and also combines well with the previously proposed RemixIT method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360251/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360251/</guid>
</item>
<item>
<title>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</title>
<description><p><span><big>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</big></span><br></p><p><span><small><i>Alexander Karataev; Christian Forsch; Laura Cottatellucci</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3348343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We consider a cell-free massive multiple-input multiple-output (CF-MaMIMO) communication system in the uplink transmission and propose a novel algorithm for blind or semi-blind joint channel estimation and data detection (JCD). We formulate the problem in the framework of bilinear inference and develop a solution based on the expectation propagation (EP) method for both channel estimation and data detection. We propose a new approximation of the joint a posteriori distribution of the channel and data whose representation as a factor graph enables the application of the EP approach using the message-passing technique, local low-complexity computations at the nodes, and an effective modeling of channel-data interplay. The derived algorithm, called bilinear-EP JCD, allows for a distributed implementation among access points (APs) and the central processing unit (CPU) and has polynomial complexity. Our simulation results show that it outperforms other EP-based state-of-the-art polynomial time algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10378663/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10378663/</guid>
</item>
<item>
<title>Adversarial Representation Learning for Robust Privacy Preservation in Audio</title>
<description><p><span><big>Adversarial Representation Learning for Robust Privacy Preservation in Audio</big></span><br></p><p><span><small><i>Shayan Gharib; Minh Tran; Diep Luong; Konstantinos Drossos; Tuomas Virtanen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Sound event detection systems are widely used in various applications such as surveillance and environmental monitoring where data is automatically collected, processed, and sent to a cloud for sound recognition. However, this process may inadvertently reveal sensitive information about users or their surroundings, hence raising privacy concerns. In this study, we propose a novel adversarial training method for learning representations of audio recordings that effectively prevents the detection of speech activity from the latent features of the recordings. The proposed method trains a model to generate invariant latent representations of speech-containing audio recordings that cannot be distinguished from non-speech recordings by a speech classifier. The novelty of our work is in the optimization algorithm, where the speech classifier's weights are regularly replaced with the weights of classifiers trained in a supervised manner. This increases the discrimination power of the speech classifier constantly during the adversarial training, motivating the model to generate latent representations in which speech is not distinguishable, even using new speech classifiers trained outside the adversarial training loop. The proposed method is evaluated against a baseline approach with no privacy measures and a prior adversarial training method, demonstrating a significant reduction in privacy violations compared to the baseline approach. Additionally, we show that the prior adversarial method is practically ineffective for this purpose.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10379095/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10379095/</guid>
</item>
<item>
<title>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</title>
<description><p><span><big>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</big></span><br></p><p><span><small><i>Yasaman Parhizkar; Gene Cheung; Andrew W. Eckford</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>It is a popular hypothesis in neuroscience that ganglion cells in the retina are activated by selectively detecting visual features in an observed scene. While ganglion cell firings can be predicted via data-trained deep neural nets, the n... |
http://localhost:1200/ieee/journal/4427201/earlyaccess - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Sensors Journal</title>
<link>https://ieeexplore.ieee.org/xpl/tocresult.jsp?isnumber=4427201</link>
<atom:link href="http://localhost:1200/ieee/journal/4427201/earlyaccess" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Sensors Journal - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Wed, 10 Apr 2024 15:20:09 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Torsion Sensor Based on Helical Long-Period Grating Inscribed in the Etched Double Cladding Fiber</title>
<description><p><span><big>Torsion Sensor Based on Helical Long-Period Grating Inscribed in the Etched Double Cladding Fiber</big></span><br></p><p><span><small><i>Yanping He; Yuehui Ma; Chen Jiang; Yunqi Liu</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373623</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>We proposed a high-sensitivity torsion sensor based on the helical long-period grating (HLPG) inscribed in the cladding-etched double cladding fiber (DCF). The torsion sensitivity has been greatly improved by reducing the cladding diameter of the DCF, from 0.342 nm/(rad/m) to 1.162 nm/(rad/m), which is one order of magnitude higher than that of the conventional long-period fiber grating. The cladding mode coupling characteristics during the etching process are also investigated experimentally. The DCF-HLPG has a low strain and temperature sensitivity, and the maximum strain sensitivity is only about 0.0022 nm/με. The proposed DCF-HLPG has a potential application as a high-sensitivity torsion sensor with low strain and temperature crosstalk.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466518/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466518/</guid>
</item>
<item>
<title>Deep-Learning-Based Prediction Algorithm for Fuel Cell Electric Vehicle Energy with Shift Mixup</title>
<description><p><span><big>Deep-Learning-Based Prediction Algorithm for Fuel Cell Electric Vehicle Energy with Shift Mixup</big></span><br></p><p><span><small><i>Tae-Ho Kim; Jae-Heung Cho; Young-Kwang Kim; Joon-Hyuk Chang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373078</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The automobile industry is switching from fossil-fuel-based energy sources to green energy sources. In particular, because hydrogen has a high energy density and efficiency compared to other energy sources, it is a major research field for commercial vehicles that require large amounts of energy and travel long distances on average, such as buses and trucks. In fuel cell electric vehicles (FCEVs), maintaining an adequate energy production intensity is necessary to improve energy production efficiency and prevent falling into a state of inoperability owing to a lack of energy. In this case, energy prediction becomes a significant factor. In this study, a deep-learning-based prediction method for FCEV powertrain energy is proposed. The proposed method uses only internal data, which can be obtained from the vehicle, and does not require external information regarding future routes. Additionally, we designed a model considering the vehicle data time-series characteristics, and proposed a shift mixup, which is a data augmentation method that does not compromise the vehicle’s dynamic characteristics, to address the data shortage problem. Furthermore, a pretext task-learning method that can improve model performance without external data during inference is introduced. This method includes pretext tasks designed specifically for the vehicle domain. Finally, a distance-based loss mask and contrastive learning that the representation can learn semantic information are proposed. Experimental results for the actual driving dataset show that our method improved by 28.05% and by 30.04% compared to the exponential moving average in terms of root mean squared error and mean absolute error, respectively. We demonstrate with energy management strategy (EMS) that an effective energy prediction algorithm helps sustain an optimal state of charge (SOC).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466522/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466522/</guid>
</item>
<item>
<title>Effect of substrate temperature on Cu2-xO growth for enhanced photodetector performance in visible region</title>
<description><p><span><big>Effect of substrate temperature on Cu2-xO growth for enhanced photodetector performance in visible region</big></span><br></p><p><span><small><i>Karthickraja Ramakrishnan; Y. Ashok Kumar Reddy; B. Ajitha</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373768</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The present study investigates the effect of substrate temperature (30-300°C) on the sputter-deposited Cu
2-x
O films for improved photodetector performance in the visible region. Cubic structured Cu2O phase is observed for all deposited films through structural analysis. Increasing substrate temperature causes the formation of nanospheres with higher grain size at 200°C confirmed by the morphological images. Moreover, all the deposited thin films show a strong absorption band in the visible region (526-550 nm). Among deposited films, a lower bandgap (2.02 eV) is observed for the film deposited at 200°C due to the higher surface area-to-volume ratio and increased grain size. The better crystalline nature, more copper vacancy, and formation of nanospheres with a larger grain size of the film deposited at 200°C result in higher mobility and conductivity. Further, the photodetector performance of the Au/Cu
2-x
O/Au test-devices is studied under the visible light (λ=530 nm) irradiation. Among samples, the sample grown at 200°C showed higher photocurrent (36.10 μA), I
on
/I
off
ratio (183.50), photoresponsivity (1.127 A/W), specific detectivity (22.40×10
11
Jones), and faster response times (τ
r
=92 ms and τ
f
=84 ms) even at lower incident optical power density (0.1283 mW/cm
2
). These improved experimental results signify the potential of the fabricated test-device at 200°C for the optoelectronics devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10471292/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10471292/</guid>
</item>
<item>
<title>A Cavity-Type Pressure Sensor Array with High Anti-Disturbance Performance Inspired by Fish Lateral Line Canal</title>
<description><p><span><big>A Cavity-Type Pressure Sensor Array with High Anti-Disturbance Performance Inspired by Fish Lateral Line Canal</big></span><br></p><p><span><small><i>Qian Yang; Qiao Hu; Liuhao Shan; Guangyu Jiang; Yuanji Yao; Long Tang; Zicai Zhu; Alvo Aabloo</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3381162</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Effective sensing is challenging in real underwater environment swarming with various disturbances. However, most research on underwater sensors has been conducted in still water, which may not provide sufficient insights for environments with disturbances. In addressing this issue, our paper not only proposes the development of anti-disturbance sensor units but also delves into evaluating the localization capabilities of artificial lateral line arrays in disturbed water environments. Concretely, inspired by the lateral line canal system of fish, we proposed a cavity-type pressure sensing device (denoted as a C-sensor) with notable anti-disturbance performance. On one hand, a 3D-printed resin cavity is filled with silicone oil, with flexible latex film packaging the upper and lower openings to achieve anti-disturbance capabilities akin to those of fish lateral line canals. On the other hand, employing IPMC materials based on ion sensing principles, the sensor generates electrical signals similar to sensing cilia in fish lateral line systems. The results demonstrated that the cavityless sensor (denoted as N-sensor) showed self-oscillation behavior, leading to significant errors in weak target signal detection and even sensor failure in extreme cases. However, the C-sensor exhibits remarkable sensing performance in still water and water disturbances, thereby confirming the effectiveness of its cavity-type structure in mitigating the impact of water disturbances. Furthermore, we developed two distinct artificial lateral line arrays based on the above sensors (C-array and N-array). The experimental results showed a marked 78% reduction in mean localization error for C-array compared to N-array. Additonally, the C-array also exhibited good ability in multiple dipoles detection. Such results indicate the superior applicability of the C-array within real complex underwater environments, promoting the practical application of IPMC sensors.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10486206/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10486206/</guid>
</item>
<item>
<title>Multi-sensor multiple extended objects tracking based on the message passing</title>
<description><p><span><big>Multi-sensor multiple extended objects tracking based on the message passing</big></span><br></p><p><span><small><i>Yuansheng Li; Tao Shen; Lin Gao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3384560</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Multiple extended objects (EOs) tracking has attracted a lot of attention due to the fast development of high-resolution sensors. A remarkable feature of EO, compared to the traditional point target, is that an EO normally produces more than one measurement, resulting in challenges for finding the associations among objects and measurements. In this paper, we are interested in tracking an unknown number of EOs with multiple sensors by resorting to the message passing method. The existence probability and belief of each EO are explicitly estimated, where the belief is modeled by a mixture gamma Gaussian inverse Wishart (GGIW) distribution, so as to jointly estimate the measurement rate (MR), centroid state and extension. The marginal posterior of EOs is approximately by belief, which is obtained by running the loopy sum-product algorithm (LSPA) on a suitably devised factor graph. As a result, the computational load of the proposed algorithm increases linearly with respect to the number of targets, thus admitting the scalability. Simulation experiments are carried out to verify the performance of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10495766/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10495766/</guid>
</item>
<item>
<title>Dual-Resonance Split Ring Resonator Metasurface for Terahertz Biosensing</title>
<description><p><span><big>Dual-Resonance Split Ring Resonator Metasurface for Terahertz Biosensing</big></span><br></p><p><span><small><i>Arslan Asim; Michael Cada; Yuan Ma; Alan Fine</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3376290</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this paper, the ‘terahertz gap’ has been addressed by designing a novel THz metasurface for potential use in biosensing applications. The metasurface sensor employs Surface Plasmon Resonance (SPR). It operates in the 0 - 1 THz band. Two sharp reflection dips are provided by the sensor, which serve as indicators of analyte refractive index variations. Geometrical as well as compositional parameters of the biosensor design have been studied to optimize the performance in the targeted frequency band. The sensor design shows compatibility with different metals. The performance of the metasurface with gold, copper and aluminum has been investigated. The metasurface geometry is decently resilient to fabrication tolerances. The sensor maintains its resonance conditions when the angle of incidence is changed with minor deviations in the spectral response, but the polarization state of the incident terahertz beam clearly disturbs the absorption peak. Therefore, the sensing performance is restricted to a maximum allowable incidence angle of 20 degrees and circularly polarized terahertz beams. The resonance conditions for the metasurface appear around 0.4 and 0.7 THz. Both resonances have been investigated with respect to changes in the analyte refractive index. The chosen refractive index range is 1 to 1.5. The sensor response is calibrated by plotting the resonance frequency versus the refractive index. Least squares regression technique has been used to extract a data model for sensor response. Comparison of the proposed design with contemporary works has been incorporated into the paper. The sensor provides sensitivities of 0.1614 THz/ RIU and 0.23 THz/ RIU. The electromagnetic simulations have been carried out through the Finite Element Method (FEM).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10474307/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10474307/</guid>
</item>
<item>
<title>An Intestine based Bio-Compatible Humidity Sensor for Environmental and Medical Measurements</title>
<description><p><span><big>An Intestine based Bio-Compatible Humidity Sensor for Environmental and Medical Measurements</big></span><br></p><p><span><small><i>Emrehan Yavsan; Muhammet Rojhat Kara; Mehmet Akif Erismis</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3367032</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The bio-compatible devices suitable for recycling and bio-degrading can be achieved with organic materials in nature. In this work, a bio-compatible capacitive humidity sensor is presented for reducing the amount of electronic waste and contributing to the sustainability of natural resources and the future. The sensor consists of 3 layers. The first layer is the processed intestine layer of cattle. Bio-compatibility is achieved with this layer. In addition to being a highly absorbing tissue, the intestine has been used for centuries for long-term preservation of meat based food. Correspondingly, the developed sensor is found to be more durable and long-lasting than other natural-material based humidity sensors in the literature. The other layers of the sensor are interdigitated copper electrodes and a 0.2 mm thick thin film strip. Thin film strip increases mechanical strength as well as flexibility. The developed sensor prototype was subjected to various tests in the humidity range of 20%-90%. In these tests, the hysteresis characteristic of the sensor, its response-recovery time, and its long-term stability and short-term step responses were examined. Moreover, as a possible application in medicine, the sensor can be used to detect breathing cycles. The sensor’s response and recovery times were measured as 8.72’ and 4.47’, respectively, possibly attributed to the stabilization of our test setup, while the sensor successfully detected deep, normal and fast breathing. Despite being kept in an uncontrolled environment, the sensor continued to operate consistently for breath measurements after 56 weeks, which is more than a year.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10445282/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10445282/</guid>
</item>
<item>
<title>Deep neural network assisted terahertz metasurface sensors for the detection of lung cancer biomarkers</title>
<description><p><span><big>Deep neural network assisted terahertz metasurface sensors for the detection of lung cancer biomarkers</big></span><br></p><p><span><small><i>Jie Chen; Fangrong Hu; Xiaoya Ma; Mo Yang; Shangjun Lin; An Su</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3384578</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>MicroRNA (miRNA) is closely related to various cancers, and the change in its expression level is closely related to the development and death of tumor cells. Here, we design and manufacture a terahertz (THz) metasurface sensor to realize the concentration detection and category identification of the miRNAs related to lung cancer. We established two spectral classification datasets, one contains 9 concentrations of miRNAs, and the other contains 3 categories of lung cancer-related miRNAs. And then, we used a deep neural network (DNN) algorithm to classify these spectral datasets and obtained a LOD (Limit-of-Detection) of 100 aM for miRNAs. Moreover, this method can identify the categories of miRNAs. Compared with the other five machine learning (ML) algorithms, the proposed neural network framework achieves the best classification results. This work provides a new way for the detection and identification of trace nucleic acid and biomarkers of many cancers.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10495770/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10495770/</guid>
</item>
<item>
<title>A Linearized Temperature Measurement System Based on the B-mode of SC-cut Quartz Crystal</title>
<description><p><span><big>A Linearized Temperature Measurement System Based on the B-mode of SC-cut Quartz Crystal</big></span><br></p><p><span><small><i>Zhiqi Li; Haipeng Zhai; Miao Miao; Wei Zhou</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3374763</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This article demonstrates a linearized temperature measurement system based on the B-mode of SC-cut quartz crystal, which has high precision and good linearity. The proposed system uses a digital linear phase comparison measurement method by selecting phase detection region, which achieves high resolution, and fast response for temperature measurement. Extensive error analysis and performance evaluation of the system were conducted in this study. The developed linearized temperature measurement system demonstrated excellent performance, including: wide temperature measuring range (-10°C to 80°C); providing sufficient resolution and extremely fast response time with a resolution of 0.01°C at 10 μs, a resolution of 5E-5°C at 10 ms, and a resolution of 1E-7°C at 1 s when the temperature keeps stable. It achieves extremely low temperature measurement relative error (R.E.&lt; 0.24%) and nonlinearity (N.L.&lt; 0.18%). The measurement results show that the maximum estimate of the Allan variance is 2.3E-4°C. The proposed system has wide prospects for intelligent applications and can play a crucial role in scenarios that require high-precision and fast-response temperature measurements.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472879/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472879/</guid>
</item>
<item>
<title>Mesoporous silica modified by poly(ionic liquid)s for low humidity sensing</title>
<description><p><span><big>Mesoporous silica modified by poly(ionic liquid)s for low humidity sensing</big></span><br></p><p><span><small><i>Zhiyan Ma; Yaping Song; Hongran Zhao; Sen Liu; Xi Yang; Teng Fei; Tong Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373030</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this study, imidazole-based poly(ionic liquid)s are introduced into acid-treated mesoporous silica by solid grinding and subsequent thermal polymerization methods. The purpose of treating the mesoporous silica with hydrochloric acid is to increase the number of hydroxyl groups on the surface to fix more poly(ionic liquid)s. Due to the synergism of enhanced hydrophilicity and the mesoporous structure, the synthesized composite materials exhibited excellent capacities to adsorb and transport water vapor molecules and are suitable for low humidity sensing applications. In the 7-33% relative humidity (RH) range, the response of the as-prepared sensor reached 806% with a short response/recovery time of 8/13 s.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466490/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466490/</guid>
</item>
<item>
<title>PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards</title>
<description><p><span><big>PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards</big></span><br></p><p><span><small><i>Elias Arbash; Margret Fuchs; Behnood Rasti; Sandra Lorenz; Pedram Ghamisi; Richard Gloaguen</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3380826</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Addressing the critical theme of recycling electronic waste (E-waste), this contribution is dedicated to developing advanced automated data processing pipelines as a basis for decision-making and process control. Aligning with the broader goals of the circular economy and the United Nations (UN) Sustainable Development Goals (SDG), our work leverages non-invasive analysis methods utilizing RGB and hyperspectral imaging data to provide both quantitative and qualitative insights into the E-waste stream composition for optimizing recycling efficiency. In this paper, we introduce ’PCB-Vision’; a pioneering RGB-hyperspectral printed circuit board (PCB) benchmark dataset, comprising 53 RGB images of high spatial resolution paired with their corresponding high spectral resolution hyperspectral data cubes in the visible and near-infrared (VNIR) range. Grounded in open science principles, our dataset provides a comprehensive resource for researchers through high-quality ground truths, focusing on three primary PCB components: integrated circuits (IC), capacitors, and connectors. We provide extensive statistical investigations on the proposed dataset together with the performance of several state-of-the-art (SOTA) models, including U-Net, Attention UNet, Residual U-Net, LinkNet, and DeepLabv3+. By openly sharing this multiscene benchmark dataset along with the baseline codes, we hope to foster transparent, traceable, and comparable developments of advanced data processing across various scientific communities, including, but not limited to, computer vision and remote sensing. Emphasizing our commitment to supporting a collaborative and inclusive scientific community, all materials, including code, data, ground truth, and masks, will be accessible at https://github.com/hifexplo/PCBVision.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10485259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10485259/</guid>
</item>
<item>
<title>DEEP-CARDIO: Recommendation System for Cardiovascular Disease Prediction using IOT Network</title>
<description><p><span><big>DEEP-CARDIO: Recommendation System for Cardiovascular Disease Prediction using IOT Network</big></span><br></p><p><span><small><i>A. Yashudas; Dinesh Gupta; G. C. Prashant; Amit Dua; Dokhyl AlQahtani; A. Siva Krishna Reddy</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373429</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Internet of Things (IoT) based remote healthcare applications provide fast and preventative medical services to the patients at risk. However, predicting heart disease is a complex task and diagnosis results are rarely accurate. To address this issue, a novel Recommendation System for Cardiovascular Disease Prediction Using IoT Network (DEEP-CARDIO) has been proposed for providing prior diagnosis, treatment, and dietary recommendations for cardiac diseases. Initially, the physiological data are collected from the patient’s remotely by using the four bio sensors such as ECG sensor, Pressure sensor, Pulse sensor and Glucose sensor. An Arduino controller receives the collected data from the IoT sensors to predict and diagnose the disease. A cardiovascular disease prediction model is implemented by using BiGRU (Bidirectional-Gated Recurrent Unit) attention model which diagnose the cardiovascular disease and classify into five available cardiovascular classes. The recommendation system provides physical and dietary recommendations to cardiac patients based on the classified data, via user mobile application. The performance of the DEEP-CARDIO is validated by Cloud Simulator (CloudSim) using the real-time Framingham’s and Statlog heart disease dataset. The proposed DEEP CARDIO method achieves an overall accuracy of 99.90% whereas, the MABC-SVM, HCBDA and MLbPM method achieves 86.91%, 88.65% and 93.63% respectively.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472883/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472883/</guid>
</item>
<item>
<title>AuNP-decorated textile as chemo resistive sensor for acetone detection</title>
<description><p><span><big>AuNP-decorated textile as chemo resistive sensor for acetone detection</big></span><br></p><p><span><small><i>S. Casalinuovo; D. Caschera; S. Quaranta; D. Caputo</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2023.3348693</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This paper presents the development of chemo resistive sensors for the detection of volatile organic compounds (VOCs). The proposed sensor is based on citrate-functionalized gold nanoparticles (AuNPs) serving as a sensitive layer deposited on cotton fabric. Impedance variations due to VOC/substrate interaction are used as a detection principle. Specifically, this work focuses on acetone detection after exposing the AuNP-decorated cotton to a CH3COCH3 aqueous solution. Such an interaction resulted in a reduction of the total impedance (i.e., magnitude) of the system. This behavior can be ascribed to Van der Waals forces existing between the C=O group and the citrate moieties adsorbed on the gold nanoparticles, which favor charge injection to the substrate. Response to water was also tested for comparison, assuring that the solvent interacts with the sensitive layer by a different adsorption mechanism, not influencing the overall results. Sensor selectivity was also verified by considering ethanol (representative of alcohol group). Indeed, impedance curves reflect the different type of chemical interaction between the analyte and the substrate. In addition, sensor limit of detection for acetone was found to be 1% v/v, in the considered frequency range. Furthermore, sensor performance in terms of reusability was evaluated, showing that the Au-cotton ability in VOCs detection could be restored after about 90 min with a percentage up to 97 % in the frequency of 1Hz. These results can be considered the starting point for the development of portable, sensitive and user-friendly devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10384362/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10384362/</guid>
</item>
<item>
<title>L-cysteine modified gold nanoparticles for copper (II) ion detection using etched Fiber Bragg Grating sensor</title>
<description><p><span><big>L-cysteine modified gold nanoparticles for copper (II) ion detection using etched Fiber Bragg Grating sensor</big></span><br></p><p><span><small><i>S Srivatzen; B S Kavitha; Asha Prasad; S Asokan</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3374514</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>A highly sensitive and selective optical detection of copper (II) (Cu
2+
) concentration in water has been proposed using L-cysteine functionalized gold nanoparticles (AuNPs) using an etched Fiber Bragg grating sensor (eFBG). The eFBG sensor produces a Bragg wavelength shift (ΔλB) in correspondence to the added Cu
2+
concentration. The amino acid L-Cysteine facilitates the formation of core-satellite nanoassemblies of gold in the presence of Cu
2+
. A linear wavelength shift is observed in accordance with the mediating Cu
2+
concentration. The sensor’s detection limit is found to be 1picomolar (pM) which is well within the world health organization (WHO) acceptable limit for safe drinking water and the dynamic range of the sensor lies between 1pM to 100 μM. The sensor is highly sensitive and selective to Cu
2+
with sensitivity of 230 pm 10
-1
M. The sensor has proven its capability as an ideal probe for onsite, real-time measurement of Cu
2+
in drinking water by giving a recovery percentage of &gt;90% in real water sample additional to good reproducibility (standard error&lt; 20 pm), high sensitivity, portability and specificity.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472453/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472453/</guid>
</item>
<item>
<title>AcHand: Detecting Respiratory Rate of operator via Smartphone Microphone</title>
<description><p><span><big>AcHand: Detecting Respiratory Rate of operator via Smartphone Microphone</big></span><br></p><p><span><small><i>KounKou Vincent; Lin Wang; Nan Jing; Wenyuan Liu</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3372255</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this paper, we dedicate a handgrip-free technique, called ”AcHand”, to continuously detect respiratory rates under acoustic signals sensing via a smartphone. Although the particularity of the AcHand is to follow the subtle movements of the heart during its displacement, we acknowledge specific challenges that can make the signal vulnerable to handgripfree scenarios such as 1) stochastic hand activities: fluctuations in movement can alter the trajectory of the acoustic signal to the microphones and cause random changes in the breathing curve, resulting in complex and unpredictable dynamics, and 2) clicking on the phone screen can disrupt the breathing by generating a large signal variation to create undesirable peaks-to-noisy the signal. However, by analyzing the recorded signal reflection evoked by a chirp sound stimulus, we propose a linear transformation of the Cardioreflector-Manoreflector to analyze the respiratory patterns variation and distinguish weak breathing signals under the handgrip-free. We compare the present signal with the ground truth recorded by ECG to approve the efficiency of the proposed method. Our experiments show AcHand achieves a MAE of 0.341 bpm, which is a significant improvement over the state-of-the-art devices, and enhancing detection capability across handgrip-free scenarios.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10462909/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10462909/</guid>
</item>
<item>
<title>Optimization Method for Node Deployment of Closed-Barrier Coverage in Hybrid Directional Sensor Networks</title>
<description><p><span><big>Optimization Method for Node Deployment of Closed-Barrier Coverage in Hybrid Directional Sensor Networks</big></span><br></p><p><span><small><i>Peng Wang; Yonghua Xiong; Jinhua She; Anjun Yu</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3378998</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Closed-barrier coverage is a new coverage problem that requires sensors on the barrier to form a closed loop. Due to the features of end-to-end connection, constructing closed barriers is a challenging task. In a Hybrid Directional Sensor Network, sensors can both rotate and move, making it more difficult to schedule sensors to construct closed barriers. To address this challenge, we propose an optimization method for node deployment of closed-barrier coverage (CCOM). First, selecting multiple initiators transforms the closed-barrier coverage problem into a simple linear-barrier construction problem. Then, a weighted barrier graph is used to search the optimal barrier path. Finally, design a rotation strategy and moving plan for the sensors, fully utilizing their rotation and moving capabilities to achieve closed-barrier coverage with the minimum number of sensors and energy consumption. The simulation experiment results show that compared with other advanced methods, our proposed method has better performance in terms of barrier construction success rate, minimum required number of mobile sensors, moving distance, and network lifetime.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10478814/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10478814/</guid>
</item>
<item>
<title>Enhanced Multicast Protocol for Low-power and Lossy IoT Networks</title>
<description><p><span><big>Enhanced Multicast Protocol for Low-power and Lossy IoT Networks</big></span><br></p><p><span><small><i>Issam Eddine Lakhlef; Badis Djamaa; Mustapha Reda Senouci; Abbas Bradai</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3375797</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Communication protocols in the Internet of Things (IoT) should take into account the resource-constrained nature of Low-power and Lossy Networks (LLNs). IP multicast protocols allow a packet to be routed from one source to multiple destinations in a single transmission. Hence, resources such as bandwidth, energy, and time are saved for a multitude of LLN applications, ranging from over-the-air programming and information sharing to device configuration and resource discovery. In this context, several multicast routing protocols have recently been proposed for LLNs, including the Multicast Protocol for LLNs (MPL). MPL has proven to be very reliable in many scenarios. However, the great resource consumption, especially in terms of energy and bandwidth, remains the main drawback of this protocol. In this paper, after a detailed overview, we provide an in-depth analysis of the MPL protocol to highlight its functional weaknesses. Then, we propose several improvements that touch on different areas to address such limitations. Extensive realistic simulations and experiments were performed to study the performance of the proposed improvements to MPL. The results obtained show that our proposals outperform the MPL protocol in terms of resource consumption (memory, bandwidth, and energy) while improving its performance in terms of end-to-end delay and maintaining the same reliability of data packet delivery.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10473686/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10473686/</guid>
</item>
<item>
<title>A 7.25μK Ultra-High Resolution MEMS Resonant Thermometer</title>
<description><p><span><big>A 7.25μK Ultra-High Resolution MEMS Resonant Thermometer</big></span><br></p><p><span><small><i>Zheng Wang; Liangbo Ma; Xiaorui Bie; Xingyin Xiong; Zhaoyang Zhai; Wuhao Yang; Yongjian Lu; Xudong Zou</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3370956</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This paper proposes an ultra-high-resolution MEMS resonant thermometer that exploits differences in Young’s modulus and coefficient of thermal expansion (CTE) between structural layers to increase thermal stress due to temperature variation. The increased thermal stress eventually acts on the resonator to produce a large frequency shift, and the temperature coefficient of frequency (TCF) can be increased to 17.4 times the original, which is consistent with theoretical analysis and finite element method simulations. The MEMS resonator chip is manufactured by the standard Silicon-on-insulator process and stacked on a ceramic chip carrier through multiple layers of materials. A self-sustaining oscillator, mainly composed of a packaged MEMS resonator chip and a low-noise application-specific integrated circuit (ASIC), is built to track the resonant frequency shift of the resonator. This MEMS resonant thermometer prototype demonstrates a high temperature coefficient of frequency (TCF) of -866.84ppm/K from -50°C to 40°C and exhibits good linearity. An ultra-high resolution of 7.25μK is achieved in the closed-loop experimental test. This is the best result achieved for a MEMS thermometer employing the resonant sensing paradigm to date.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10460450/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10460450/</guid>
</item>
<item>
<title>Design of MTM-based Multi-band Micro-Biosensor in Terahertz region as perfect absorber for Early-Stage Leukemia Diagnosis with sensitivity 18626373 TH...</title>
<description><p><span><big>Design of MTM-based Multi-band Micro-Biosensor in Terahertz region as perfect absorber for Early-Stage Leukemia Diagnosis with sensitivity 18626373 THz/RIU</big></span><br></p><p><span><small><i>Musa N. Hamza; Mohammad Tariqul Islam</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383522</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this article, a novel highly sensitive biosensor based on perfect metamaterial absorbers is presented. A detailed study is presented for several different models, different types of substrate materials, different types of resonant materials and substrate thicknesses showing very good sensitivity to any changes. The proposed biosensor exhibits high sensitivity to different polarization and incidence angles, ensuring its performance in terms of signal-to-noise ratio. The proposed biosensor was carefully compared with previously designed sensors and biosensors. The proposed biosensor exhibits amazing sensitivity, such as a quality factor of 41.1, a figure of merit (FOM) of 172466416 RIU-1, and a sensitivity (S) of 18626373 THz/RIU. The high sensitivity of the biosensor allows early detection of leukemia, demonstrating significant differences between leukemia and normal blood. Another interesting result of this article is the use of terahertz waves for imaging. Microwave imaging is performed for electric fields, magnetic fields, and power.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10494209/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10494209/</guid>
</item>
<item>
<title>LEPS: A lightweight and effective single-stage detector for pothole segmentation</title>
<description><p><span><big>LEPS: A lightweight and effective single-stage detector for pothole segmentation</big></span><br></p><p><span><small><i>Xiaoning Huang; Jintao Cheng; Qiuchi Xiang; Jun Dong; Jin Wu; Rui Fan; Xiaoyu Tang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2023.3330335</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Currently, the problem of potholes on urban roads is becoming increasingly severe. Identifying and locating road potholes promptly has become a major challenge for urban construction. Therefore, we proposed a lightweight and effective instance segmentation method called LEPS (Lightweight and Effective Pothole Segmentation Detector) for road pothole detection. To extract the image edge information and gradient information of potholes from the feature map more efficiently, we proposed a module that performs the convolutional super-position supplemented with a convolutional kernel to enhance spatial details for the backbone. We have designed a novel module applied to the Neck layer, improving the detection performance while reducing the parameters. To enable accurate segmentation of fine-grained features, we optimized the ProtoNet, which enables the segmentation head to generate high-quality masks for more accurate prediction. We have fully demonstrated the effectiveness of the method through a large number of comparative experiments. Our detector has excellent performance on two authoritative example datasets, URTIRI and COCO, and can successfully be applied to visual sensors to accurately detect and segment road potholes in real environments, accurately LEPS reached 0.892 and 0.648 in terms of Mask for AP50 and AP50:95, which is improved by 4.6% and 20.6% compared to the original model. These results demonstrate its strong competitiveness when compared to other models. Comprehensively, LEPS improves detection accuracy while maintaining lightweight, which allows the model to meet the practical application requirements of edge computing devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10315064/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10315064/</guid>
</item>
<item>
<title>UMHE: Unsupervised Multispectral Homography Estimation</title>
<description><p><span><big>UMHE: Unsupervised Multispectral Homography Estimation</big></span><br></p><p><span><small><i>Jeongmin Shin; Jiwon Kim; Seokjun Kwon; Namil Kim; Soonmin Hwang; Yukyung Choi</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383453</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Multispectral image alignment plays a crucial role in exploiting complementary information between different spectral images. Homography-based image alignment can be a practical solution considering a tradeoff between runtime and accuracy. Existing methods, however, have difficulty with multispectral images due to the additional spectral gap or require expensive human labels to train models. To solve these problems, this paper presents a comprehensive study on multispectral homography estimation in an unsupervised learning manner. We propose a curriculum data augmentation, an effective solution for models learning spectrum-agnostic representation by providing diverse input pairs. We also propose to use the phase congruency loss that explicitly calculates the reconstruction between images based on low-level structural information in the frequency domain. To encourage multispectral alignment research, we introduce a novel FLIR corresponding dataset that has manually labeled local correspondences between multispectral images. Our model achieves state-of-the-art alignment performance on the proposed FLIR correspondence dataset among supervised and unsupervised methods while running at
151 FPS
. Furthermore, our model shows good generalization ability on the M3FD dataset without finetuning.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10494213/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10494213/</guid>
</item>
<item>
<title>Calibration and Evaluation of a Low-Cost Optical Particulate Matter Sensor for Measurement of Lofted Lunar Dust</title>
<description><p><span><big>Calibration and Evaluation of a Low-Cost Optical Particulate Matter Sensor for Measurement of Lofted Lunar Dust</big></span><br></p><p><span><small><i>Abhay Vidwans; Jeffrey Gillis-Davis; Pratim Biswas</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3366436</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Several recent Earth-based investigations employed low-cost particulate matter sensors to address the lack of spatiotemporal resolution in air quality data. The lunar environment also has a particulate matter problem in the form of fine lofted and levitated dust particles. Natural and anthropogenic mobilized dust can cause a slew of difficulties for surface operations (deposition onto radiators, optical components, and mechanical devices). Despite the urgency of mitigating dust on the Moon and other airless bodies, the performance of low-cost sensors has not been critically evaluated for space applications. Upcoming long-term robotic and human exploration missions to the Moon necessitate a robust sensor that can monitor particulate matter levels and establish a spatially and temporally resolved global network. In this work, we calibrate two optical light-scattering particulate matter sensors against research-grade aerosol instruments for measuring the concentration of aerosolized lunar simulants. Sensors showed a stronger dependence on aerosol particle size distribution than particle composition. Vacuum testing showed a significant deviation in performance compared to atmospheric pressure, with a stronger dependency on lunar simulant. The predicted mass deposition, based on sensor output coupled with dust trajectory, was within an order of magnitude of the reference deposition. Our results demonstrate for the first time that low-cost particulate matter sensors can monitor dust concentrations with reasonable accuracy in a vacuum environment, with two caveats. First, precise calibrations must be performed with a dust simulant closely matching the particle size distribution of the target dust, and second, atmospheric pressure calibrations alone are insufficient.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10462021/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10462021/</guid>
</item>
<item>
<title>Reweighting Interacting Multiple-model Algorithm to Overcome Model Competition for Target Tracking in the Hybrid System</title>
<description><p><span><big>Reweighting Interacting Multiple-model Algorithm to Overcome Model Competition for Target Tracking in the Hybrid System</big></span><br></p><p><span><small><i>Guowei Li; Shurui Zhang; Yubing Han; Weixing Sheng; Thia Kirubarajan</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3369854</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The vicious competition of interacting multiple-model algorithm (IMM) is an inherent problem and would produce irreversible effects on IMM estimation results, especially combining with the radar system. In this paper, a novel reweighting IMM (RIMM) is proposed to overcome this issue. Firstly, the theoretical lower bound of model numbers in different situations is respectively provided through the analysis of IMM limitations. Furthermore, certificate the influence of model inaccuracy on the Kalman filter, which illustrates an effective method for reducing errors is increasing model numbers. Thirdly, the definition of model set density and the analysis of the true model space are given, and their connection establishes the standard of how to design the model set or add the model number. Finally, an effective method called RIMM is provided to overcome the competition caused by model increasing. The proposed RIMM holds strong adaptability for different model sets. The simulations of RIMM highlight the correctness and effectiveness of the proposed methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10462948/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10462948/</guid>
</item>
<item>
<title>A Stretchable and Wearable Ultrasonic Transducer Array for Bladder Volume Monitoring Application</title>
<description><p><span><big>A Stretchable and Wearable Ultrasonic Transducer Array for Bladder Volume Monitoring Application</big></span><br></p><p><span><small><i>Cong Pu; Ben Fu; Lehang Guo; Huixiong Xu; Chang Peng</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3382244</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Monitoring bladder volume is an essential application for patients with voiding dysfunction. Current wearable ultrasonic bladder monitors are rigid, bulky and cannot conform to human skin. This study presents a stretchable and wearable ultrasonic transducer array that can both conform to non-planar skin surfaces and continuously monitor bladder volume. The wearable transducer array mainly consists of 4 × 4 piezoelectric transducer elements, stretchable serpentine-based electrodes for electrical interconnection, electrically conductive adhesive as both matching and backing layers, and silicone elastomer for encapsulation. The transducer array has a center frequency of 3.9 MHz, -6 dB acoustic bandwidth of 57.6%, and up to 40% elastic stretchability. The volume of balloon-bladder phantom was estimated by a least square ellipsoid fitting method. With the true volume of balloon-bladder phantoms ranging from 10 mL to 405 mL, the proposed transducer array has a mean absolute percentage error of 9.4%, a mean absolute error of 11.9 mL and a coefficient of determination of 0.975, which demonstrates the capability of the proposed stretchable and wearable ultrasonic transducer array for bladder volume monitoring application.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10489841/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10489841/</guid>
</item>
<item>
<title>Intelligent fault diagnosis of bearing using multiwavelet perception kernel convolutional neural network</title>
<description><p><span><big>Intelligent fault diagnosis of bearing using multiwavelet perception kernel convolutional neural network</big></span><br></p><p><span><small><i>Yuanyuan Zhou; Hang Wang; Yongbin Liu; Xianzeng Liu; Zheng Cao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3370564</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Strong background noise characteristics of vibration signals cause issues with poor identification capability of features by fault diagnostic models. To address this issue, a method is proposed for intelligent fault diagnosis of bearing using multiwavelet perception kernel (MPK) and feature attention convolutional neural network (FA-CNN). First, four multiwavelet perception kernels are constructed to decompose the vibration signals in full-band multilevel. Second, improved multiwavelet information entropy (IMIE) of the frequency band components is calculated. The calculated component entropies of the corresponding frequency bands are integrated to construct frequency band clusters (FBC) from low to high frequencies. Third, joint approximate diagonalization eigen (JADE) is introduced to perform feature fusion for every FBC to eliminate redundant information, and fused features from low to high frequencies are obtained as original inputs. The FA-CNN bearing fault diagnosis framework is constructed for intelligent fault diagnosis of bearings. Finally, the effectiveness of the proposed method is verified by two cases. The results show that the proposed method has high fault feature recognition capability.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10458914/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10458914/</guid>
</item>
<item>
<title>Improved multipath mitigation using multiple trend-surface hemispherical map in GNSS precise point positioning</title>
<description><p><span><big>Improved multipath mitigation using multiple trend-surface hemispherical map in GNSS precise point positioning</big></span><br></p><p><span><small><i>Haijun Yuan; Zhetao Zhang; Xiufeng He; Jinwen Zeng; Hao Wang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3374458</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Global Navigation Satellite System (GNSS) Precise Point Positioning (PPP) is highly appreciated as a positioning technology; however, multipath restricts its accuracy and reliability. In this study, we proposed a Multiple Trend-surface Multipath Hemispherical Map (MT-MHM), where discrepant multipath fading frequency or orbit orientation of each satellite in each azimuth and elevation grid is considered. In a typical static scenario, consecutive 8-day observations are collected to extract multipath and conduct PPP experiments. Compared with traditional multipath hemispherical map based on trend-surface fitting, MT-MHM has better modelling performance of residuals and improves standard deviations of residuals for all satellites. Besides, the positioning errors and required convergence epochs of PPP are reduced by using MT-MHM. For the positioning accuracy of all solutions, MT-MHM exhibits 50.8, 15.2, and 27.9% improvements compared with that without multipath correction in the east, north, and up directions, respectively. In conclusion, our proposed MT-MHM exhibits better performance in terms of residual reduction, convergence time, and positioning accuracy in PPP.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472420/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472420/</guid>
</item>
<item>
<title>Design optimization and characterization of a 3D-printed tactile sensor for tissue palpation</title>
<description><p><span><big>Design optimization and characterization of a 3D-printed tactile sensor for tissue palpation</big></span><br></p><p><span><small><i>D. Lo Presti; L. Zoboli; A. Addabbo; D. Bianchi; A. Dimo; C. Massaroni; V. Altomare; A. Grasso; A. Gizzi; E. Schena</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3369337</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Manual palpation is a crucial medical procedure that relies on surface examination to detect internal tissue abnormalities, heavily reliant on healthcare professionals’ expertise and tactile sensitivity. To tackle these issues, smart palpation systems based on electrical or optical sensors have been developed to perform quantitative tactile measurements, crucial for identifying various solid tumors, including breast and prostate cancer by assessing tissue mechanical properties. In this context, fiber Bragg gratings (FBGs) are emerging as a promising technology due to their advantages (e.g., high metrological properties, multiplexing capacity, and easy packaging) making them ideal for tactile sensing. This study explores the benefits of FBG and 3D printing to develop a tactile sensor for tissue palpation. First, an optimization of the design of the sensing core of a previously developed probe was conducted through finite element analysis. The novel structure addresses the primary limitation of the previous solution, where non-uniform strain distribution on the encapsulated FBG causes compression on the grating with high risk of bending and breakage. In contrast, the modeled geometry ensures FBG elongation during tissue palpation. A 3D printing and characterization of the proposed solution was carried out to investigate the response of the enclosed FBG when pushed against different materials showing promising results in discriminating tissues according to their mechanical properties: the more rigid the indented substrate the higher the sensor output. This property will be fundamental for enhancing early tumor detection through superficial tissue palpation, advancing the efficacy of prevention measures.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10455966/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10455966/</guid>
</item>
<item>
<title>Deep Reinforcement Learning-based Joint Sequence Scheduling and Trajectory Planning in Wireless Rechargeable Sensor Networks</title>
<description><p><span><big>Deep Reinforcement Learning-based Joint Sequence Scheduling and Trajectory Planning in Wireless Rechargeable Sensor Networks</big></span><br></p><p><span><small><i>Chengpeng Jiang; Wencong Chen; Ziyang Wang; Wendong Xiao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373664</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Mobile charging has become a popular and efficient method for replenishing energy. This is done with a mobile charger (MC) and wireless energy transfer technology (WET), which helps to alleviate the issue of energy constraints in wireless rechargeable sensor networks (WRSNs). Notably, designing mobile charging scheduling schemes is essential for improving charging performance. Most current studies assume that the networks are obstacle-free. Unlike the existing studies, this paper focuses on joint sequence scheduling and trajectory planning problems (JSSTP), which assumes that the network has multiple static obstacles. To address this issue, we propose a novel deep reinforcement learning-based JSSTP (DRL-JSSTP) that enables the MC to avoid obstacles and reach the charging target to charge the sensors fully. This approach maximizes energy usage efficiency and sensor survival rate while satisfying MC energy capacity constraints. DRL-JSSTP includes a charging target selector and a trajectory planner, which determine the index of the next charging target and plan the movement trajectory to avoid obstacles, respectively. We adopt 1-D convolutional neural networks to extract feature information about the environment state and gated recurrent units to predict the charging decisions. Simulation results demonstrate that DRL-JSSTP outperforms existing approaches, achieving higher energy usage efficiency and sensor survival rate.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466524/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466524/</guid>
</item>
<item>
<title>A Simultaneous Wirel |
|
Successfully generated as following: http://localhost:1200/ieee/journal/8782710 - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Open Journal of Signal Processing</title>
<link>https://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=8782710</link>
<atom:link href="http://localhost:1200/ieee/journal/8782710" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Open Journal of Signal Processing - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Wed, 10 Apr 2024 15:26:04 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Synthbuster: Towards Detection of Diffusion Model Generated Images</title>
<description><p><span><big>Synthbuster: Towards Detection of Diffusion Model Generated Images</big></span><br></p><p><span><small><i>Quentin Bammey</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337714</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Synthetically-generated images are getting increasingly popular. Diffusion models have advanced to the stage where even non-experts can generate photo-realistic images from a simple text prompt. They expand creative horizons but also open a Pandora's box of potential disinformation risks. In this context, the present corpus of synthetic image detection techniques, primarily focusing on older generative models like Generative Adversarial Networks, finds itself ill-equipped to deal with this emerging trend. Recognizing this challenge, we introduce a method specifically designed to detect synthetic images produced by diffusion models. Our approach capitalizes on the inherent frequency artefacts left behind during the diffusion process. Spectral analysis is used to highlight the artefacts in the Fourier transform of a residual image, which are used to distinguish real from fake images. The proposed method can detect diffusion-model-generated images even under mild
jpeg
compression, and generalizes relatively well to unknown models. By pioneering this novel approach, we aim to fortify forensic methodologies and ignite further research into the detection of AI-generated images.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334046/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334046/</guid>
</item>
<item>
<title>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</title>
<description><p><span><big>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</big></span><br></p><p><span><small><i>Yanbin Zou; Jingna Fan; Zekai Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we consider using time-of-arrival (TOA) measurements from a single moving receiver to locate a moving target at constant velocity that emits a periodic signal with unknown signal period. First, we give the TOA measurement model and deduce the Cram
$\acute{\text{e}}$
r-Rao lower bounds (CRLB). Then, we formulate a nonlinear least squares (NLS) problem to estimate the unknown parameters. We use semidefinite programming (SDP) techniques to relax the nonconvex NLS problem. However, it is shown that the SDP localization algorithm cannot provide a high-quality solution. Subsequently, we develop a fixed point iteration (FPI) method to improve the performance of the SDP algorithm. In addition, we also consider the presence of receiver position errors and develop the corresponding localization algorithm. Numerical simulations are conducted to demonstrate the localization performance of the proposed algorithms by comparing them with the CRLB.
Index Term
-Fixed point iteration (FPI), semidefinite programming (SDP), single moving receiver, target localization, time-of-arrival (TOA).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336384/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336384/</guid>
</item>
<item>
<title>The Neural-SRP Method for Universal Robust Multi-Source Tracking</title>
<description><p><span><big>The Neural-SRP Method for Universal Robust Multi-Source Tracking</big></span><br></p><p><span><small><i>Eric Grinstein; Christopher M. Hicks; Toon van Waterschoot; Mike Brookes; Patrick A. Naylor</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340057</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Neural networks have achieved state-of-the-art performance on the task of acoustic Direction-of-Arrival (DOA) estimation using microphone arrays. Neural models can be classified as end-to-end or hybrid, each class showing advantages and disadvantages. This work introduces Neural-SRP, an end-to-end neural network architecture for DOA estimation inspired by the classical Steered Response Power (SRP) method, which overcomes limitations of current neural models. We evaluate the architecture on multiple scenarios, namely, multi-source DOA tracking and single-source DOA tracking under the presence of directional and diffuse noise. The experiments demonstrate that our proposed method compares favourably in terms of computational and localization performance with established neural methods on various recorded and simulated benchmark datasets.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345765/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345765/</guid>
</item>
<item>
<title>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</title>
<description><p><span><big>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</big></span><br></p><p><span><small><i>Alexander Venus; Erik Leitinger; Stefan Tertinek; Klaus Witrisal</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper presents a neural-enhanced probabilistic model and corresponding factor graph-based sum-product algorithm for robust localization and tracking in multipath-prone environments. The introduced hybrid probabilistic model consists of physics-based and data-driven measurement models capturing the information contained in both, the line-of-sight (LOS) component as well as in multipath components (NLOS components). The physics-based and data-driven models are embedded in a joint Bayesian framework allowing to derive from first principles a factor graph-based algorithm that fuses the information of these models. The proposed algorithm uses radio signal measurements from multiple base stations to robustly estimate the mobile agent's position together with all model parameters. It provides high localization accuracy by exploiting the position-related information of the LOS component via the physics-based model and robustness by exploiting the geometric imprint of multipath components independent of the propagation channel via the data-driven model. In a challenging numerical experiment involving obstructed LOS situations to all anchors, we show that the proposed sequential algorithm significantly outperforms state-of-the-art methods and attains the posterior Cramér-Rao lower bound even with training data limited to local regions.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336409/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336409/</guid>
</item>
<item>
<title>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</title>
<description><p><span><big>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</big></span><br></p><p><span><small><i>AMIT EFRAIM; Joseph M. Francos</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In solving a model fitting problem, the existence of outliers in the set of measurements can have a devastating effect on the solution accuracy. Traditionally, in order to overcome this problem, robust point cloud registration algorithms are composed of transformation hypothesis generation, followed by hypothesis evaluation aimed at selecting the best hypothesized estimate. Hypotheses evaluation is commonly performed using the sample consensus criterion. However, since this criterion accounts only for the consensus size, it fails when the maximal sample consensus is incorrect. We propose a new hypothesis evaluation approach, generalizing the sample consensus approach, where instead of the sample consensus, the transformation that maximizes the point clouds feature correlation is selected as the best hypothesis. The feature vector at each point contains information such as on local geometry and semantic context. Utilizing this information in the hypotheses evaluation and selection procedure allows for a correct decision even when the hypothesis yielding the maximal sample consensus is false. Consequently, the probability of selecting the correct model increases. We show both mathematically and empirically that substituting the sample consensus criterion with the proposed point cloud feature correlation hypothesis test (PC-FCHT) lowers the probability of large registration errors, compared to using the special case of sample consensus. The proposed PC-FCHT is applicable to any algorithm that follows the hypothesis generation and evaluation scheme, potentially improving the success rates of a wide family of point cloud registration algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345750/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345750/</guid>
</item>
<item>
<title>Joint PAPR and OBP Reduction for NC-OFDM Systems</title>
<description><p><span><big>Joint PAPR and OBP Reduction for NC-OFDM Systems</big></span><br></p><p><span><small><i>Hsuan-Fu Wang; Fang-Biau Ueng; Bo-Heng Yeh</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3329757</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The spectrum resource is always a critical issue for wireless communications since it directly impacts the data rate and capacity. However, the problem of spectrum resource scarcity always exists. Moreover, spectrum resource scarcity becomes more severe as new communication technologies and wireless applications sprout. Noncontiguous orthogonal frequency division multiplexing (NC-OFDM) is a multicarrier method for bandwidth utilization. Unfortunately, this system has two fatal defects: high peak-to-average power ratio (PAPR) and considerable out-of-band power (OBP), which are detrimental to the system's performance. To solve these two problems, we propose a convex optimization-based method for joint PAPR and OBP reduction in NC-OFDM Systems. The strategy is to permit the secondary user to utilize the unoccupied spectrum of the primary user with dynamic spectrum sharing (DSS) based on a cognitive radio network (CRN). To this end, a flexible system operating over noncontiguous bands and DSS scenarios is necessary. The simulation results have shown that our method could effectively improve the overall performance and outperform other schemes, i.e., projections onto convex sets (POCS) and alternating projections onto convex and non-convex sets (APOCNCS), without harming the transmission of the primary system. The collaboration between secondary and primary systems is viable with the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10305259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10305259/</guid>
</item>
<item>
<title>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</title>
<description><p><span><big>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</big></span><br></p><p><span><small><i>Gal Itzhak; Israel Cohen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339433</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article presents a Kronecker-product (KP) beamforming approach incorporating sparse concentric circular arrays (SCCAs). The locations of the microphones on the SCCA are optimized concerning the broadband array directivity over a wide range of direction-of-arrival (DOA) deviations of a desired signal. A maximum directivity factor (MDF) sub-beamformer is derived accordingly with the optimal locations. Then, we propose two global beamformers obtained as a Kronecker product of a uniform linear array (ULA) and the SCCA sub-beamformer. The global beamformers differ by the type of the ULA, which is designed either as an MDF sub-beamformer along the
$\mathsf {x}$
-axis or as a maximum white noise gain sub-beamformer along the
$\mathsf {y}$
-axis. We analyze the performance of the proposed beamformers in terms of the directivity factor, the white noise gain, and their spatial beampatterns. Compared to traditional beamformers, the proposed beamformers exhibit considerably larger tolerance to DOA deviations concerning both the azimuth and elevation angles. Experimental results with speech signals in noisy and reverberant environments demonstrate that the proposed approach outperforms traditional beamformers regarding the perceptual evaluation of speech quality (PESQ) and short-time objective intelligibility (STOI) scores when the desired speech signals deviate from the nominal DOA.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342869/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342869/</guid>
</item>
<item>
<title>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</title>
<description><p><span><big>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</big></span><br></p><p><span><small><i>Karn N. Watcharasupat; Chih-Wei Wu; Yiwei Ding; Iroro Orife; Aaron J. Hipple; Phillip A. Williams; Scott Kramer; Alexander Lerch; William Wolcott</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339428</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Cinematic audio source separation is a relatively new subtask of audio source separation, with the aim of extracting the dialogue, music, and effects stems from their mixture. In this work, we developed a model generalizing the Bandsplit RNN for any complete or overcomplete partitions of the frequency axis. Psychoacoustically motivated frequency scales were used to inform the band definitions which are now defined with redundancy for more reliable feature extraction. A loss function motivated by the signal-to-noise ratio and the sparsity-promoting property of the 1-norm was proposed. We additionally exploit the information-sharing property of a common-encoder setup to reduce computational complexity during both training and inference, improve separation performance for hard-to-generalize classes of sounds, and allow flexibility during inference time with detachable decoders. Our best model sets the state of the art on the Divide and Remaster dataset with performance above the ideal ratio mask for the dialogue stem.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342812/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342812/</guid>
</item>
<item>
<title>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</title>
<description><p><span><big>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</big></span><br></p><p><span><small><i>Miguel Ferrer; María de Diego; Alberto Gonzalez</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340106</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The LMS algorithm is widely employed in adaptive systems due to its robustness, simplicity, and reasonable performance. However, it is well known that this algorithm suffers from a slow convergence speed when dealing with colored reference signals. Numerous variants and alternative algorithms have been proposed to address this issue, though all of them entail an increase in computational cost. Among the proposed alternatives, the affine projection algorithm stands out. This algorithm has the peculiarity of starting from
$N$
data vectors of the reference signal. It transforms these vectors into as many data vectors suitably normalized in energy and mutually orthogonal. In this work, we propose a version of the LMS algorithm that, similar to the affine projection algorithm, starts from
$N$
data vectors of the reference signal but corrects them by using only a scalar factor that functions as a convergence step. Our goal is to align the behavior of this algorithm with the behavior of the affine projection algorithm without significantly increasing the computational cost of the LMS.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345730/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345730/</guid>
</item>
<item>
<title>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</title>
<description><p><span><big>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</big></span><br></p><p><span><small><i>Zongyao Li; Ren Togo; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340616</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article proposes a method for transferring knowledge of semantic segmentation from a labeled source domain to an unlabeled target domain without using the source-domain data. Such a problem is called source-data-free domain adaptation, in which a pre-trained source-domain model and the unlabeled target-domain data are used to transfer the label knowledge across the domains. Like most previous methods, our method uses pseudo labels for distilling and transferring the source-domain knowledge. On the basis of the pseudo-label learning, our method improves the domain adaptation performance in two innovative ways: 1) reducing the domain differences by source-data-free style transfer and 2) exploring the style diversity within the target domain by style modification. To this end, we introduce two additional modules: 1) an inter-domain style transfer module which aligns the feature statistics of the source and target domains before producing the pseudo labels thereby improving the pseudo labels' accuracy, and 2) an intra-domain style modification module which modifies the image styles within the target domain for learning intra-domain style-invariant features. Our method with the two modules outperforms previous source-data-free domain adaptation methods in two commonly used benchmarks. Moreover, our method is well compatible with the previous methods for further improvement.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10356748/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10356748/</guid>
</item>
<item>
<title>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</title>
<description><p><span><big>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</big></span><br></p><p><span><small><i>Tobias Kabzinski; Peter Jax</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337721</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Quasi-continuous system identification of time-variant linear acoustic systems can be applied in various audio signal processing applications when numerous acoustic transfer functions must be measured. A prominent application is measuring head-related transfer functions. We treat the underlying multiple-input-multiple-output (MIMO) system identification problem in a state-space model as a joint estimation problem for states, representing impulse responses, and state-space model parameters using the expectation maximization (EM) algorithm. We address limitations of prior work by imposing different model structures, especially for dependencies within a (transformed) state vector. This results in block diagonal matrix structures, for which we derive M-step update rules. Making assumptions about this model structure and choosing a block size for a given application define the computational complexity. In examples, we found that applying this framework yields improvements of up to 10 dB in relative system distance in comparison to a conventional method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334061/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334061/</guid>
</item>
<item>
<title>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</title>
<description><p><span><big>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</big></span><br></p><p><span><small><i>Damir Rakhimov; Martin Haardt</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337729</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this paper, we present an analytical performance assessment of 2-D Tensor ESPRIT in terms of physical parameters. We show that the error in the
$r$
-mode depends only on two components, irrespective of the dimensionality of the problem. We obtain analytical expressions in closed form for the mean squared error (MSE) in each dimension as a function of the signal-to-noise (SNR) ratio, the array steering matrices, the number of antennas, the number of snapshots, the selection matrices, and the signal correlation. The derived expressions allow a better understanding of the difference in performance between the tensor and the matrix versions of the ESPRIT algorithm. The simulation results confirm the coincidence between the presented analytical expression and the curves obtained via Monte Carlo trials. We analyze the behavior of each of the two error components in different scenarios.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334446/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334446/</guid>
</item>
<item>
<title>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</title>
<description><p><span><big>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</big></span><br></p><p><span><small><i>Yaman Kındap; Simon Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343341</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this work we study linear vector stochastic differential equation (SDE) models driven by the generalised hyperbolic (GH) Lévy process for inference in continuous-time non-Gaussian filtering problems. The GH family of stochastic processes offers a flexible framework for modelling of non-Gaussian, heavy-tailed characteristics and includes the normal inverse-Gaussian, variance-gamma and Student-t processes as special cases. We present continuous-time simulation methods for the solution of vector SDE models driven by GH processes and novel inference methodologies using a variant of sequential Markov chain Monte Carlo (MCMC). As an example a particular formulation of Langevin dynamics is studied within this framework. The model is applied to both a synthetically generated data set and a real-world financial series to demonstrate its capabilities.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360268/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360268/</guid>
</item>
<item>
<title>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</title>
<description><p><span><big>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</big></span><br></p><p><span><small><i>James M. Cozens; Simon J. Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344048</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper proposes a probabilistic approach for extracting time-varying and irregular time signature information from polyphonic audio extracts, subsequently providing beat and bar line positions given inferred time signature divisions. This is achieved via dynamically evaluating the beat tempo as a function of time through finding an optimal compromise in beat and bar alignment in the time and tempo domains. Time signature divisions are determined based on a new representation, termed the Metrogram, that presents time-varying information regarding rhythmic and metric periodicities in the Tempogram. Our methodology is characterised by its ability to provide a distribution over metric interpretations, offering insights into the diverse ways music can be rhythmically perceived. Results indicate high-level accuracy for a variety of polyphonic extracts containing irregular, complex, irrational, and time-varying time signatures. Accuracy rivalling state-of-the-art methodologies is also reported in a beat tracking task performed on the standard Ballroom Dataset. The paper offers insights into the field of dynamic time signature recognition and beat tracking, offering a valuable and versatile resource for the analysis, composition, and performance of music.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363392/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363392/</guid>
</item>
<item>
<title>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</title>
<description><p><span><big>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</big></span><br></p><p><span><small><i>Yuto Watanabe; Ren Togo; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343335</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Although text-guided image manipulation approaches have demonstrated highly accurate performance for editing the appearance of images in a virtual or simple scenario, their real-world applications face significant challenges. The primary cause of these challenges is the misalignment in the distribution of training and real-world data, which leads to unstable text-guided image manipulation. In this work, we propose a novel framework called TolerantGAN and tackle the new task of real-world text-guided image manipulation independent of the training data. To achieve this, we introduce two key concepts of a border smoothly connection module (BSCM) and a manipulation direction-based attention module (MDAM). BSCM smoothens the misalignment in the distribution of training and real-world data. MDAM extracts only regions highly relevant for image manipulation and assists in reconstructing unobserved objects in the training data. For in-the-wild input images of various classes, TolerantGAN robustly outperforms the state-of-the-art methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360283/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360283/</guid>
</item>
<item>
<title>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</title>
<description><p><span><big>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</big></span><br></p><p><span><small><i>Anastasia Avdeeva; Aleksei Gusev; Tseren Andzhukaev; Artem Ivanov</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343342</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Whispered speech is a quiet voice without vocalization. One of the common cases of using whispered speech is a technique that can help overcome stuttering. But whispered speech can be uncomfortable and difficult to understand in everyday communication. To address these problems, we propose a method of low-delayed whisper-to-speech voice conversion, which can be useful in real life communication of people with disordered speech. As part of our research, we study the impact of streaming Automatic Speech Recognition models on the quality of voice conversion, comparing different streaming models and methods for model adaptation to streaming settings, and showing the importance of using such models in cases of low-delayed voice conversion.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360259/</guid>
</item>
<item>
<title>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</title>
<description><p><span><big>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</big></span><br></p><p><span><small><i>Denis C. Ilie-Ablachim; Andra Băltoiu; Bogdan Dumitrescu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344313</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We propose sparse representation and dictionary learning algorithms for dictionaries whose atoms are characterized by Gaussian probability distributions around some central atoms. This extension of the space covered by the atoms permits a better characterization of localized features of the signals. The representations are computed by trading-off representation error and the probability of the actual atoms used in the representation. We propose two types of algorithms: a greedy one, similar in spirit with Orthogonal Matching Pursuit, and one based on L1-regularization. We apply our new algorithms to unsupervised anomaly detection, where the representation error is used as anomaly score. The flexibility of our approach appears to improve more the representations of the many normal signals than those of the few outliers, at least for an anomaly type called dependency, thus improving the detection quality. Comparison with both standard dictionary learning algorithms and established anomaly detection methods is favorable.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365333/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365333/</guid>
</item>
<item>
<title>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</title>
<description><p><span><big>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</big></span><br></p><p><span><small><i>Yuya Moroto; Yingrui Ye; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344079</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>There are various sentiment theories for categorizing human sentiments into several discrete sentiment categories, which means that the theory used for training sentiment prediction methods does not always match that used in the test phase. As a solution to this problem, zero-shot visual sentiment prediction methods have been proposed to predict unseen sentiments for which no images are available in the training phase. However, the training of these previous zero-shot methods relies on a single sentiment theory, which limits their ability to handle sentiments from other theories. Thus, this article proposes a more robust zero-shot visual sentiment prediction method that can handle cross-domain sentiments defined in different sentiment theories. Specifically, by focusing on the fact that sentiments are abstract concepts common to humans regardless of whether their theories are different, we incorporate knowledge distillation into our method to construct a teacher–student model that can train the implicit relationships between sentiments defined in different sentiment theories. Furthermore, to enhance sentiment discrimination capability and strengthen the implicit relationships between sentiments, we introduce a novel sentiment loss between the teacher and student models. In this way, our model becomes robust to unseen sentiments by exploiting the implicit relationships between sentiments. The contributions of this article are the introduction of knowledge distillation and a novel sentiment loss between the teacher and student models for zero-shot visual sentiment prediction, and improved performance of zero-shot visual sentiment prediction. Experiments on several open datasets demonstrate the effectiveness of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363382/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363382/</guid>
</item>
<item>
<title>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</title>
<description><p><span><big>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</big></span><br></p><p><span><small><i>Chengen Liu; Geert Leus; Elvin Isufi</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339376</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The edge flow reconstruction task consists of retreiving edge flow signals from corrupted or incomplete measurements. This is typically solved by a regularized optimization problem on higher-order networks such as simplicial complexes and the corresponding regularizers are chosen based on prior knowledge. Tailoring this prior to the setting of interest can be challenging or it may not even be possible. Thus, we consider to learn this prior knowledge via a model-based deep learning approach. We propose a new regularized optimization problem for the simplicial edge flow reconstruction task, the simplicial ElasticNet, which combines the advantages of the
$\ell _{1}$
and
$\ell _{2}$
norms. We solve the simplicial ElasticNet problem via the multi-block alternating direction method of multipliers (ADMM) algorithm and provide conditions on its convergence. By unrolling the ADMM iterative steps, we develop a model-based neural network with a low requirement on the number of training data. This unrolling network replaces the fixed parameters in the iterative algorithm by learnable weights, thus exploiting the neural network's learning capability while preserving the iterative algorithm's interpretability. We enhance this unrolling network via simplicial convolutional filters to aggregate information from the edge flow neighbors, ultimately, improving the network learning expressivity. Extensive experiments on real-world and synthetic datasets validate the proposed approaches and show considerable improvements over both baselines and traditional non-model-based neural networks.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342735/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342735/</guid>
</item>
<item>
<title>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</title>
<description><p><span><big>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</big></span><br></p><p><span><small><i>Jim Beckers; Bart Van Erp; Ziyue Zhao; Kirill Kondrashov; Bert De Vries</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337718</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Bayesian model reduction provides an efficient approach for comparing the performance of all nested sub-models of a model, without re-evaluating any of these sub-models. Until now, Bayesian model reduction has been applied mainly in the computational neuroscience community on simple models. In this paper, we formulate and apply Bayesian model reduction to perform principled pruning of Bayesian neural networks, based on variational free energy minimization. Direct application of Bayesian model reduction, however, gives rise to approximation errors. Therefore, a novel iterative pruning algorithm is presented to alleviate the problems arising with naive Bayesian model reduction, as supported experimentally on the publicly available UCI datasets for different inference algorithms. This novel parameter pruning scheme solves the shortcomings of current state-of-the-art pruning methods that are used by the signal processing community. The proposed approach has a clear stopping criterion and minimizes the same objective that is used during training. Next to these benefits, our experiments indicate better model performance in comparison to state-of-the-art pruning schemes.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334001/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334001/</guid>
</item>
<item>
<title>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</title>
<description><p><span><big>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</big></span><br></p><p><span><small><i>Anh Minh Truong; Wilfried Philips; Peter Veelaert</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340064</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Many image-based rendering (IBR) methods rely on depth estimates obtained from structured light or time-of-flight depth sensors to synthesize novel views from sparse camera networks. However, these estimates often contain missing or noisy regions, resulting in an incorrect mapping between source and target views. This situation makes the fusion process more challenging, as the visual information is misaligned, inconsistent, or missing. In this work, we first implement a lightweight network based on the transformer, which is well-known for its capability to model long-range relationships within input data, to extract spatial features from color images. These features are then used to enhance the quality of completed depth maps. Furthermore, we combine a sequential deep neural network with a spatial attention mechanism to effectively fuse the projected features from multiple source viewpoints. This approach enables us to integrate information from an arbitrary number of source viewpoints as well as improve accuracy in synthesized views. Experimental results on challenging datasets demonstrate that our method achieves superior synthesized image quality compared to state-of-the-art (SOTA) methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345792/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345792/</guid>
</item>
<item>
<title>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</title>
<description><p><span><big>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</big></span><br></p><p><span><small><i>Reza Mirzaeifard; Naveen K. D. Venkategowda; Vinay Chakravarthi Gogineni; Stefan Werner</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344395</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper investigates quantile regression in the presence of non-convex and non-smooth sparse penalties, such as the minimax concave penalty (MCP) and smoothly clipped absolute deviation (SCAD). The non-smooth and non-convex nature of these problems often leads to convergence difficulties for many algorithms. While iterative techniques such as coordinate descent and local linear approximation can facilitate convergence, the process is often slow. This sluggish pace is primarily due to the need to run these approximation techniques until full convergence at each step, a requirement we term as a
secondary convergence iteration
. To accelerate the convergence speed, we employ the alternating direction method of multipliers (ADMM) and introduce a novel single-loop smoothing ADMM algorithm with an increasing penalty parameter, named SIAD, specifically tailored for sparse-penalized quantile regression. We first delve into the convergence properties of the proposed SIAD algorithm and establish the necessary conditions for convergence. Theoretically, we confirm a convergence rate of
$o({k^{-\frac{1}{4}}})$
for the sub-gradient bound of the augmented Lagrangian, where
$k$
denotes the number of iterations. Subsequently, we provide numerical results to showcase the effectiveness of the SIAD algorithm. Our findings highlight that the SIAD method outperforms existing approaches, providing a faster and more stable solution for sparse-penalized quantile regression.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365338/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365338/</guid>
</item>
<item>
<title>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</title>
<description><p><span><big>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</big></span><br></p><p><span><small><i>Oliver Lang; Christian Hofbauer; Reinhard Feger; Mario Huemer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343308</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>A promising waveform candidate for future joint sensing and communication systems is orthogonal frequency-division multiplexing (OFDM). For such systems, supporting multiple transmit antennas requires multiplexing methods for the generation of orthogonal transmit signals, where equidistant subcarrier interleaving (ESI) is the most popular multiplexing method. In this work, we analyze a multiplexing method called Doppler-division multiplexing (DDM). This method applies a phase shift from OFDM symbol to OFDM symbol to separate signals transmitted by different Tx antennas along the velocity axis of the range-Doppler map. The main focus of this work lies on the implications of DDM on the communication task. It will be shown that for DDM, the channels observed in the communication receiver are heavily time-varying, preventing any meaningful transmission of data when not taken into account. In this work, a communication system designed to combat these time-varying channels is proposed, which includes methods for data estimation, synchronization, and channel estimation. Bit error ratio (BER) simulations demonstrate the superiority of this communications system compared to ESI-based systems.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360223/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360223/</guid>
</item>
<item>
<title>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</title>
<description><p><span><big>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</big></span><br></p><p><span><small><i>Nils L. Westhausen; Bernd T. Meyer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343320</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we introduce a causal low-latency low-complexity approach for binaural multichannel blind speaker separation in noisy reverberant conditions. The model, referred to as Group Communication Binaural Filter and Sum Network (GCBFSnet) predicts complex filters for filter-and-sum beamforming in the time-frequency domain. We apply Group Communication (GC), i.e., latent model variables are split into groups and processed with a shared sequence model with the aim of reducing the complexity of a simple model only containing one convolutional and one recurrent module. With GC we are able to reduce the size of the model by up to 83% and the complexity up to 73% compared to the model without GC, while mostly retaining performance. Even for the smallest model configuration, GCBFSnet matches the performance of a low-complexity TasNet baseline in most metrics despite the larger size and higher number of required operations of the baseline.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360275/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360275/</guid>
</item>
<item>
<title>Reverse Ordering Techniques for Attention-Based Channel Prediction</title>
<description><p><span><big>Reverse Ordering Techniques for Attention-Based Channel Prediction</big></span><br></p><p><span><small><i>Valentina Rizzello; Benedikt Böck; Michael Joham; Wolfgang Utschick</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344024</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Channel state information (CSI) is crucial for enhancing the performance of wireless systems by allowing to adjust the transmission strategies based on the current channel conditions. However, obtaining precise CSI is difficult because of the fast-changing channel conditions caused by multi-path fading. An inaccurate CSI hinders the performance of various adaptive wireless systems, highlighting the need for channel prediction techniques to effectively mitigate the drawbacks of outdated CSI. Conventional methods typically depend on assumptions regarding user velocity or require knowledge of the Doppler frequency. In contrast to existing approaches, we aim for a more robust and practical solution by training neural networks without making any assumptions about user velocity, relying solely on noisy channel observations during training. Specifically, we adapt both the sequence-to-sequence with attention (Seq2Seq-attn) and transformer models for channel prediction. Additionally, a new technique called reverse positional encoding is introduced in the transformer model to improve the robustness of the model against varying sequence lengths. Similarly, the encoder outputs of the Seq2Seq-attn model are reversed prior to the application of attention mechanisms. By means of simulations, we show that these proposed techniques enable the models to effectively capture relationships within sequences of channel snapshots without increasing the complexity. Importantly, this capability remains robust across varying sequence lengths, representing a substantial improvement over existing methodologies.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363354/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363354/</guid>
</item>
<item>
<title>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</title>
<description><p><span><big>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</big></span><br></p><p><span><small><i>Sarina Meyer; Xiaoxiao Miao; Ngoc Thang Vu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344375</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Speaker anonymization is the task of modifying a speech recording such that the original speaker cannot be identified anymore. Since the first Voice Privacy Challenge in 2020, along with the release of a framework, the popularity of this research topic is continually increasing. However, the comparison and combination of different anonymization approaches remains challenging due to the complexity of evaluation and the absence of user-friendly research frameworks. We therefore propose an efficient speaker anonymization and evaluation framework based on a modular and easily extendable structure, almost fully in Python. The framework facilitates the orchestration of several anonymization approaches in parallel and allows for interfacing between different techniques. Furthermore, we propose modifications to common evaluation methods which improves the quality of the evaluation and reduces their computation time by 65 to 95%, depending on the metric. Our code is fully open source.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365329/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365329/</guid>
</item>
<item>
<title>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</title>
<description><p><span><big>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</big></span><br></p><p><span><small><i>Alessandro Ilic Mezza; Matteo Amerena; Alberto Bernardini; Augusto Sarti</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343318</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Real-time audio communications over IP have become essential to our daily lives. Packet-switched networks, however, are inherently prone to jitter and data losses, thus creating a strong need for effective packet loss concealment (PLC) techniques. Though solutions based on deep learning have made significant progress in that direction as far as speech is concerned, extending the use of such methods to applications of Networked Music Performance (NMP) presents significant challenges, including high fidelity requirements, higher sampling rates, and stringent temporal constraints associated to the simultaneous interaction between remote musicians. In this article, we present PARCnet, a hybrid PLC method that utilizes a feed-forward neural network to estimate the time-domain residual signal of a parallel linear autoregressive model. Objective metrics and a listening test show that PARCnet provides state-of-the-art results while enabling real-time operation on CPU.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360264/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360264/</guid>
</item>
<item>
<title>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</title>
<description><p><span><big>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</big></span><br></p><p><span><small><i>Katerina Zmolikova; Michael Syskind Pedersen; Jesper Jensen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Supervised learning-based speech enhancement methods often work remarkably well in acoustic situations represented in the training corpus but generalize poorly to out-of-domain situations, i.e. situations not seen during training. This stands in the way of further improvement of these methods in realistic scenarios, as collecting paired noisy-clean recordings in the target application domain is typically not feasible. Recording noisy-only in-domain data is, though, much more practical. In this article, we tackle the problem of unsupervised domain adaptation in speech enhancement. Specifically, we propose a way to use in-domain noisy-only data in the training of a neural network to improve upon a model trained solely on out-of-domain paired data. For this, we make use of masked spectrogram prediction, a technique from self-supervised learning that aims to interpolate masked regions of a spectrogram. We hypothesize that masked spectrogram prediction encourages learning of features that represent well both speech and noise components of the noisy signals. These features can then be used to train a more robust speech enhancement system. We evaluate the proposed method on the VoiceBank-DEMAND and LibriFSD50k databases, with WSJ0-CHiME3 serving as the out-of-domain database. We show that the proposed method outperforms both the out-of-domain system and the baseline approaches, i.e. RemixIT and noisy-target training, and also combines well with the previously proposed RemixIT method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360251/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360251/</guid>
</item>
<item>
<title>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</title>
<description><p><span><big>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</big></span><br></p><p><span><small><i>Alexander Karataev; Christian Forsch; Laura Cottatellucci</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3348343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We consider a cell-free massive multiple-input multiple-output (CF-MaMIMO) communication system in the uplink transmission and propose a novel algorithm for blind or semi-blind joint channel estimation and data detection (JCD). We formulate the problem in the framework of bilinear inference and develop a solution based on the expectation propagation (EP) method for both channel estimation and data detection. We propose a new approximation of the joint a posteriori distribution of the channel and data whose representation as a factor graph enables the application of the EP approach using the message-passing technique, local low-complexity computations at the nodes, and an effective modeling of channel-data interplay. The derived algorithm, called bilinear-EP JCD, allows for a distributed implementation among access points (APs) and the central processing unit (CPU) and has polynomial complexity. Our simulation results show that it outperforms other EP-based state-of-the-art polynomial time algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10378663/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10378663/</guid>
</item>
<item>
<title>Adversarial Representation Learning for Robust Privacy Preservation in Audio</title>
<description><p><span><big>Adversarial Representation Learning for Robust Privacy Preservation in Audio</big></span><br></p><p><span><small><i>Shayan Gharib; Minh Tran; Diep Luong; Konstantinos Drossos; Tuomas Virtanen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Sound event detection systems are widely used in various applications such as surveillance and environmental monitoring where data is automatically collected, processed, and sent to a cloud for sound recognition. However, this process may inadvertently reveal sensitive information about users or their surroundings, hence raising privacy concerns. In this study, we propose a novel adversarial training method for learning representations of audio recordings that effectively prevents the detection of speech activity from the latent features of the recordings. The proposed method trains a model to generate invariant latent representations of speech-containing audio recordings that cannot be distinguished from non-speech recordings by a speech classifier. The novelty of our work is in the optimization algorithm, where the speech classifier's weights are regularly replaced with the weights of classifiers trained in a supervised manner. This increases the discrimination power of the speech classifier constantly during the adversarial training, motivating the model to generate latent representations in which speech is not distinguishable, even using new speech classifiers trained outside the adversarial training loop. The proposed method is evaluated against a baseline approach with no privacy measures and a prior adversarial training method, demonstrating a significant reduction in privacy violations compared to the baseline approach. Additionally, we show that the prior adversarial method is practically ineffective for this purpose.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10379095/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10379095/</guid>
</item>
<item>
<title>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</title>
<description><p><span><big>Signal Processing in the Retina: Interpretable Graph Classifier to Predict Ganglion Cell Responses</big></span><br></p><p><span><small><i>Yasaman Parhizkar; Gene Cheung; Andrew W. Eckford</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>It is a popular hypothesis in neuroscience that ganglion cells in the retina are activated by selectively detecting visual features in an observed scene. While ganglion cell firings can be predicted via data-trained deep neural nets, the n... |
http://localhost:1200/ieee/journal/4427201/earlyaccess - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Sensors Journal</title>
<link>https://ieeexplore.ieee.org/xpl/tocresult.jsp?isnumber=4427201</link>
<atom:link href="http://localhost:1200/ieee/journal/4427201/earlyaccess" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Sensors Journal - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Wed, 10 Apr 2024 15:26:09 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Torsion Sensor Based on Helical Long-Period Grating Inscribed in the Etched Double Cladding Fiber</title>
<description><p><span><big>Torsion Sensor Based on Helical Long-Period Grating Inscribed in the Etched Double Cladding Fiber</big></span><br></p><p><span><small><i>Yanping He; Yuehui Ma; Chen Jiang; Yunqi Liu</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373623</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>We proposed a high-sensitivity torsion sensor based on the helical long-period grating (HLPG) inscribed in the cladding-etched double cladding fiber (DCF). The torsion sensitivity has been greatly improved by reducing the cladding diameter of the DCF, from 0.342 nm/(rad/m) to 1.162 nm/(rad/m), which is one order of magnitude higher than that of the conventional long-period fiber grating. The cladding mode coupling characteristics during the etching process are also investigated experimentally. The DCF-HLPG has a low strain and temperature sensitivity, and the maximum strain sensitivity is only about 0.0022 nm/με. The proposed DCF-HLPG has a potential application as a high-sensitivity torsion sensor with low strain and temperature crosstalk.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466518/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466518/</guid>
</item>
<item>
<title>Deep-Learning-Based Prediction Algorithm for Fuel Cell Electric Vehicle Energy with Shift Mixup</title>
<description><p><span><big>Deep-Learning-Based Prediction Algorithm for Fuel Cell Electric Vehicle Energy with Shift Mixup</big></span><br></p><p><span><small><i>Tae-Ho Kim; Jae-Heung Cho; Young-Kwang Kim; Joon-Hyuk Chang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373078</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The automobile industry is switching from fossil-fuel-based energy sources to green energy sources. In particular, because hydrogen has a high energy density and efficiency compared to other energy sources, it is a major research field for commercial vehicles that require large amounts of energy and travel long distances on average, such as buses and trucks. In fuel cell electric vehicles (FCEVs), maintaining an adequate energy production intensity is necessary to improve energy production efficiency and prevent falling into a state of inoperability owing to a lack of energy. In this case, energy prediction becomes a significant factor. In this study, a deep-learning-based prediction method for FCEV powertrain energy is proposed. The proposed method uses only internal data, which can be obtained from the vehicle, and does not require external information regarding future routes. Additionally, we designed a model considering the vehicle data time-series characteristics, and proposed a shift mixup, which is a data augmentation method that does not compromise the vehicle’s dynamic characteristics, to address the data shortage problem. Furthermore, a pretext task-learning method that can improve model performance without external data during inference is introduced. This method includes pretext tasks designed specifically for the vehicle domain. Finally, a distance-based loss mask and contrastive learning that the representation can learn semantic information are proposed. Experimental results for the actual driving dataset show that our method improved by 28.05% and by 30.04% compared to the exponential moving average in terms of root mean squared error and mean absolute error, respectively. We demonstrate with energy management strategy (EMS) that an effective energy prediction algorithm helps sustain an optimal state of charge (SOC).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466522/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466522/</guid>
</item>
<item>
<title>Effect of substrate temperature on Cu2-xO growth for enhanced photodetector performance in visible region</title>
<description><p><span><big>Effect of substrate temperature on Cu2-xO growth for enhanced photodetector performance in visible region</big></span><br></p><p><span><small><i>Karthickraja Ramakrishnan; Y. Ashok Kumar Reddy; B. Ajitha</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373768</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The present study investigates the effect of substrate temperature (30-300°C) on the sputter-deposited Cu
2-x
O films for improved photodetector performance in the visible region. Cubic structured Cu2O phase is observed for all deposited films through structural analysis. Increasing substrate temperature causes the formation of nanospheres with higher grain size at 200°C confirmed by the morphological images. Moreover, all the deposited thin films show a strong absorption band in the visible region (526-550 nm). Among deposited films, a lower bandgap (2.02 eV) is observed for the film deposited at 200°C due to the higher surface area-to-volume ratio and increased grain size. The better crystalline nature, more copper vacancy, and formation of nanospheres with a larger grain size of the film deposited at 200°C result in higher mobility and conductivity. Further, the photodetector performance of the Au/Cu
2-x
O/Au test-devices is studied under the visible light (λ=530 nm) irradiation. Among samples, the sample grown at 200°C showed higher photocurrent (36.10 μA), I
on
/I
off
ratio (183.50), photoresponsivity (1.127 A/W), specific detectivity (22.40×10
11
Jones), and faster response times (τ
r
=92 ms and τ
f
=84 ms) even at lower incident optical power density (0.1283 mW/cm
2
). These improved experimental results signify the potential of the fabricated test-device at 200°C for the optoelectronics devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10471292/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10471292/</guid>
</item>
<item>
<title>A Cavity-Type Pressure Sensor Array with High Anti-Disturbance Performance Inspired by Fish Lateral Line Canal</title>
<description><p><span><big>A Cavity-Type Pressure Sensor Array with High Anti-Disturbance Performance Inspired by Fish Lateral Line Canal</big></span><br></p><p><span><small><i>Qian Yang; Qiao Hu; Liuhao Shan; Guangyu Jiang; Yuanji Yao; Long Tang; Zicai Zhu; Alvo Aabloo</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3381162</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Effective sensing is challenging in real underwater environment swarming with various disturbances. However, most research on underwater sensors has been conducted in still water, which may not provide sufficient insights for environments with disturbances. In addressing this issue, our paper not only proposes the development of anti-disturbance sensor units but also delves into evaluating the localization capabilities of artificial lateral line arrays in disturbed water environments. Concretely, inspired by the lateral line canal system of fish, we proposed a cavity-type pressure sensing device (denoted as a C-sensor) with notable anti-disturbance performance. On one hand, a 3D-printed resin cavity is filled with silicone oil, with flexible latex film packaging the upper and lower openings to achieve anti-disturbance capabilities akin to those of fish lateral line canals. On the other hand, employing IPMC materials based on ion sensing principles, the sensor generates electrical signals similar to sensing cilia in fish lateral line systems. The results demonstrated that the cavityless sensor (denoted as N-sensor) showed self-oscillation behavior, leading to significant errors in weak target signal detection and even sensor failure in extreme cases. However, the C-sensor exhibits remarkable sensing performance in still water and water disturbances, thereby confirming the effectiveness of its cavity-type structure in mitigating the impact of water disturbances. Furthermore, we developed two distinct artificial lateral line arrays based on the above sensors (C-array and N-array). The experimental results showed a marked 78% reduction in mean localization error for C-array compared to N-array. Additonally, the C-array also exhibited good ability in multiple dipoles detection. Such results indicate the superior applicability of the C-array within real complex underwater environments, promoting the practical application of IPMC sensors.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10486206/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10486206/</guid>
</item>
<item>
<title>Multi-sensor multiple extended objects tracking based on the message passing</title>
<description><p><span><big>Multi-sensor multiple extended objects tracking based on the message passing</big></span><br></p><p><span><small><i>Yuansheng Li; Tao Shen; Lin Gao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3384560</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Multiple extended objects (EOs) tracking has attracted a lot of attention due to the fast development of high-resolution sensors. A remarkable feature of EO, compared to the traditional point target, is that an EO normally produces more than one measurement, resulting in challenges for finding the associations among objects and measurements. In this paper, we are interested in tracking an unknown number of EOs with multiple sensors by resorting to the message passing method. The existence probability and belief of each EO are explicitly estimated, where the belief is modeled by a mixture gamma Gaussian inverse Wishart (GGIW) distribution, so as to jointly estimate the measurement rate (MR), centroid state and extension. The marginal posterior of EOs is approximately by belief, which is obtained by running the loopy sum-product algorithm (LSPA) on a suitably devised factor graph. As a result, the computational load of the proposed algorithm increases linearly with respect to the number of targets, thus admitting the scalability. Simulation experiments are carried out to verify the performance of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10495766/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10495766/</guid>
</item>
<item>
<title>Dual-Resonance Split Ring Resonator Metasurface for Terahertz Biosensing</title>
<description><p><span><big>Dual-Resonance Split Ring Resonator Metasurface for Terahertz Biosensing</big></span><br></p><p><span><small><i>Arslan Asim; Michael Cada; Yuan Ma; Alan Fine</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3376290</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this paper, the ‘terahertz gap’ has been addressed by designing a novel THz metasurface for potential use in biosensing applications. The metasurface sensor employs Surface Plasmon Resonance (SPR). It operates in the 0 - 1 THz band. Two sharp reflection dips are provided by the sensor, which serve as indicators of analyte refractive index variations. Geometrical as well as compositional parameters of the biosensor design have been studied to optimize the performance in the targeted frequency band. The sensor design shows compatibility with different metals. The performance of the metasurface with gold, copper and aluminum has been investigated. The metasurface geometry is decently resilient to fabrication tolerances. The sensor maintains its resonance conditions when the angle of incidence is changed with minor deviations in the spectral response, but the polarization state of the incident terahertz beam clearly disturbs the absorption peak. Therefore, the sensing performance is restricted to a maximum allowable incidence angle of 20 degrees and circularly polarized terahertz beams. The resonance conditions for the metasurface appear around 0.4 and 0.7 THz. Both resonances have been investigated with respect to changes in the analyte refractive index. The chosen refractive index range is 1 to 1.5. The sensor response is calibrated by plotting the resonance frequency versus the refractive index. Least squares regression technique has been used to extract a data model for sensor response. Comparison of the proposed design with contemporary works has been incorporated into the paper. The sensor provides sensitivities of 0.1614 THz/ RIU and 0.23 THz/ RIU. The electromagnetic simulations have been carried out through the Finite Element Method (FEM).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10474307/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10474307/</guid>
</item>
<item>
<title>An Intestine based Bio-Compatible Humidity Sensor for Environmental and Medical Measurements</title>
<description><p><span><big>An Intestine based Bio-Compatible Humidity Sensor for Environmental and Medical Measurements</big></span><br></p><p><span><small><i>Emrehan Yavsan; Muhammet Rojhat Kara; Mehmet Akif Erismis</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3367032</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The bio-compatible devices suitable for recycling and bio-degrading can be achieved with organic materials in nature. In this work, a bio-compatible capacitive humidity sensor is presented for reducing the amount of electronic waste and contributing to the sustainability of natural resources and the future. The sensor consists of 3 layers. The first layer is the processed intestine layer of cattle. Bio-compatibility is achieved with this layer. In addition to being a highly absorbing tissue, the intestine has been used for centuries for long-term preservation of meat based food. Correspondingly, the developed sensor is found to be more durable and long-lasting than other natural-material based humidity sensors in the literature. The other layers of the sensor are interdigitated copper electrodes and a 0.2 mm thick thin film strip. Thin film strip increases mechanical strength as well as flexibility. The developed sensor prototype was subjected to various tests in the humidity range of 20%-90%. In these tests, the hysteresis characteristic of the sensor, its response-recovery time, and its long-term stability and short-term step responses were examined. Moreover, as a possible application in medicine, the sensor can be used to detect breathing cycles. The sensor’s response and recovery times were measured as 8.72’ and 4.47’, respectively, possibly attributed to the stabilization of our test setup, while the sensor successfully detected deep, normal and fast breathing. Despite being kept in an uncontrolled environment, the sensor continued to operate consistently for breath measurements after 56 weeks, which is more than a year.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10445282/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10445282/</guid>
</item>
<item>
<title>Deep neural network assisted terahertz metasurface sensors for the detection of lung cancer biomarkers</title>
<description><p><span><big>Deep neural network assisted terahertz metasurface sensors for the detection of lung cancer biomarkers</big></span><br></p><p><span><small><i>Jie Chen; Fangrong Hu; Xiaoya Ma; Mo Yang; Shangjun Lin; An Su</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3384578</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>MicroRNA (miRNA) is closely related to various cancers, and the change in its expression level is closely related to the development and death of tumor cells. Here, we design and manufacture a terahertz (THz) metasurface sensor to realize the concentration detection and category identification of the miRNAs related to lung cancer. We established two spectral classification datasets, one contains 9 concentrations of miRNAs, and the other contains 3 categories of lung cancer-related miRNAs. And then, we used a deep neural network (DNN) algorithm to classify these spectral datasets and obtained a LOD (Limit-of-Detection) of 100 aM for miRNAs. Moreover, this method can identify the categories of miRNAs. Compared with the other five machine learning (ML) algorithms, the proposed neural network framework achieves the best classification results. This work provides a new way for the detection and identification of trace nucleic acid and biomarkers of many cancers.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10495770/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10495770/</guid>
</item>
<item>
<title>A Linearized Temperature Measurement System Based on the B-mode of SC-cut Quartz Crystal</title>
<description><p><span><big>A Linearized Temperature Measurement System Based on the B-mode of SC-cut Quartz Crystal</big></span><br></p><p><span><small><i>Zhiqi Li; Haipeng Zhai; Miao Miao; Wei Zhou</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3374763</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This article demonstrates a linearized temperature measurement system based on the B-mode of SC-cut quartz crystal, which has high precision and good linearity. The proposed system uses a digital linear phase comparison measurement method by selecting phase detection region, which achieves high resolution, and fast response for temperature measurement. Extensive error analysis and performance evaluation of the system were conducted in this study. The developed linearized temperature measurement system demonstrated excellent performance, including: wide temperature measuring range (-10°C to 80°C); providing sufficient resolution and extremely fast response time with a resolution of 0.01°C at 10 μs, a resolution of 5E-5°C at 10 ms, and a resolution of 1E-7°C at 1 s when the temperature keeps stable. It achieves extremely low temperature measurement relative error (R.E.&lt; 0.24%) and nonlinearity (N.L.&lt; 0.18%). The measurement results show that the maximum estimate of the Allan variance is 2.3E-4°C. The proposed system has wide prospects for intelligent applications and can play a crucial role in scenarios that require high-precision and fast-response temperature measurements.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472879/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472879/</guid>
</item>
<item>
<title>Mesoporous silica modified by poly(ionic liquid)s for low humidity sensing</title>
<description><p><span><big>Mesoporous silica modified by poly(ionic liquid)s for low humidity sensing</big></span><br></p><p><span><small><i>Zhiyan Ma; Yaping Song; Hongran Zhao; Sen Liu; Xi Yang; Teng Fei; Tong Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373030</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this study, imidazole-based poly(ionic liquid)s are introduced into acid-treated mesoporous silica by solid grinding and subsequent thermal polymerization methods. The purpose of treating the mesoporous silica with hydrochloric acid is to increase the number of hydroxyl groups on the surface to fix more poly(ionic liquid)s. Due to the synergism of enhanced hydrophilicity and the mesoporous structure, the synthesized composite materials exhibited excellent capacities to adsorb and transport water vapor molecules and are suitable for low humidity sensing applications. In the 7-33% relative humidity (RH) range, the response of the as-prepared sensor reached 806% with a short response/recovery time of 8/13 s.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466490/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466490/</guid>
</item>
<item>
<title>PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards</title>
<description><p><span><big>PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards</big></span><br></p><p><span><small><i>Elias Arbash; Margret Fuchs; Behnood Rasti; Sandra Lorenz; Pedram Ghamisi; Richard Gloaguen</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3380826</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Addressing the critical theme of recycling electronic waste (E-waste), this contribution is dedicated to developing advanced automated data processing pipelines as a basis for decision-making and process control. Aligning with the broader goals of the circular economy and the United Nations (UN) Sustainable Development Goals (SDG), our work leverages non-invasive analysis methods utilizing RGB and hyperspectral imaging data to provide both quantitative and qualitative insights into the E-waste stream composition for optimizing recycling efficiency. In this paper, we introduce ’PCB-Vision’; a pioneering RGB-hyperspectral printed circuit board (PCB) benchmark dataset, comprising 53 RGB images of high spatial resolution paired with their corresponding high spectral resolution hyperspectral data cubes in the visible and near-infrared (VNIR) range. Grounded in open science principles, our dataset provides a comprehensive resource for researchers through high-quality ground truths, focusing on three primary PCB components: integrated circuits (IC), capacitors, and connectors. We provide extensive statistical investigations on the proposed dataset together with the performance of several state-of-the-art (SOTA) models, including U-Net, Attention UNet, Residual U-Net, LinkNet, and DeepLabv3+. By openly sharing this multiscene benchmark dataset along with the baseline codes, we hope to foster transparent, traceable, and comparable developments of advanced data processing across various scientific communities, including, but not limited to, computer vision and remote sensing. Emphasizing our commitment to supporting a collaborative and inclusive scientific community, all materials, including code, data, ground truth, and masks, will be accessible at https://github.com/hifexplo/PCBVision.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10485259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10485259/</guid>
</item>
<item>
<title>DEEP-CARDIO: Recommendation System for Cardiovascular Disease Prediction using IOT Network</title>
<description><p><span><big>DEEP-CARDIO: Recommendation System for Cardiovascular Disease Prediction using IOT Network</big></span><br></p><p><span><small><i>A. Yashudas; Dinesh Gupta; G. C. Prashant; Amit Dua; Dokhyl AlQahtani; A. Siva Krishna Reddy</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373429</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Internet of Things (IoT) based remote healthcare applications provide fast and preventative medical services to the patients at risk. However, predicting heart disease is a complex task and diagnosis results are rarely accurate. To address this issue, a novel Recommendation System for Cardiovascular Disease Prediction Using IoT Network (DEEP-CARDIO) has been proposed for providing prior diagnosis, treatment, and dietary recommendations for cardiac diseases. Initially, the physiological data are collected from the patient’s remotely by using the four bio sensors such as ECG sensor, Pressure sensor, Pulse sensor and Glucose sensor. An Arduino controller receives the collected data from the IoT sensors to predict and diagnose the disease. A cardiovascular disease prediction model is implemented by using BiGRU (Bidirectional-Gated Recurrent Unit) attention model which diagnose the cardiovascular disease and classify into five available cardiovascular classes. The recommendation system provides physical and dietary recommendations to cardiac patients based on the classified data, via user mobile application. The performance of the DEEP-CARDIO is validated by Cloud Simulator (CloudSim) using the real-time Framingham’s and Statlog heart disease dataset. The proposed DEEP CARDIO method achieves an overall accuracy of 99.90% whereas, the MABC-SVM, HCBDA and MLbPM method achieves 86.91%, 88.65% and 93.63% respectively.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472883/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472883/</guid>
</item>
<item>
<title>AuNP-decorated textile as chemo resistive sensor for acetone detection</title>
<description><p><span><big>AuNP-decorated textile as chemo resistive sensor for acetone detection</big></span><br></p><p><span><small><i>S. Casalinuovo; D. Caschera; S. Quaranta; D. Caputo</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2023.3348693</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This paper presents the development of chemo resistive sensors for the detection of volatile organic compounds (VOCs). The proposed sensor is based on citrate-functionalized gold nanoparticles (AuNPs) serving as a sensitive layer deposited on cotton fabric. Impedance variations due to VOC/substrate interaction are used as a detection principle. Specifically, this work focuses on acetone detection after exposing the AuNP-decorated cotton to a CH3COCH3 aqueous solution. Such an interaction resulted in a reduction of the total impedance (i.e., magnitude) of the system. This behavior can be ascribed to Van der Waals forces existing between the C=O group and the citrate moieties adsorbed on the gold nanoparticles, which favor charge injection to the substrate. Response to water was also tested for comparison, assuring that the solvent interacts with the sensitive layer by a different adsorption mechanism, not influencing the overall results. Sensor selectivity was also verified by considering ethanol (representative of alcohol group). Indeed, impedance curves reflect the different type of chemical interaction between the analyte and the substrate. In addition, sensor limit of detection for acetone was found to be 1% v/v, in the considered frequency range. Furthermore, sensor performance in terms of reusability was evaluated, showing that the Au-cotton ability in VOCs detection could be restored after about 90 min with a percentage up to 97 % in the frequency of 1Hz. These results can be considered the starting point for the development of portable, sensitive and user-friendly devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10384362/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10384362/</guid>
</item>
<item>
<title>L-cysteine modified gold nanoparticles for copper (II) ion detection using etched Fiber Bragg Grating sensor</title>
<description><p><span><big>L-cysteine modified gold nanoparticles for copper (II) ion detection using etched Fiber Bragg Grating sensor</big></span><br></p><p><span><small><i>S Srivatzen; B S Kavitha; Asha Prasad; S Asokan</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3374514</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>A highly sensitive and selective optical detection of copper (II) (Cu
2+
) concentration in water has been proposed using L-cysteine functionalized gold nanoparticles (AuNPs) using an etched Fiber Bragg grating sensor (eFBG). The eFBG sensor produces a Bragg wavelength shift (ΔλB) in correspondence to the added Cu
2+
concentration. The amino acid L-Cysteine facilitates the formation of core-satellite nanoassemblies of gold in the presence of Cu
2+
. A linear wavelength shift is observed in accordance with the mediating Cu
2+
concentration. The sensor’s detection limit is found to be 1picomolar (pM) which is well within the world health organization (WHO) acceptable limit for safe drinking water and the dynamic range of the sensor lies between 1pM to 100 μM. The sensor is highly sensitive and selective to Cu
2+
with sensitivity of 230 pm 10
-1
M. The sensor has proven its capability as an ideal probe for onsite, real-time measurement of Cu
2+
in drinking water by giving a recovery percentage of &gt;90% in real water sample additional to good reproducibility (standard error&lt; 20 pm), high sensitivity, portability and specificity.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472453/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472453/</guid>
</item>
<item>
<title>AcHand: Detecting Respiratory Rate of operator via Smartphone Microphone</title>
<description><p><span><big>AcHand: Detecting Respiratory Rate of operator via Smartphone Microphone</big></span><br></p><p><span><small><i>KounKou Vincent; Lin Wang; Nan Jing; Wenyuan Liu</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3372255</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this paper, we dedicate a handgrip-free technique, called ”AcHand”, to continuously detect respiratory rates under acoustic signals sensing via a smartphone. Although the particularity of the AcHand is to follow the subtle movements of the heart during its displacement, we acknowledge specific challenges that can make the signal vulnerable to handgripfree scenarios such as 1) stochastic hand activities: fluctuations in movement can alter the trajectory of the acoustic signal to the microphones and cause random changes in the breathing curve, resulting in complex and unpredictable dynamics, and 2) clicking on the phone screen can disrupt the breathing by generating a large signal variation to create undesirable peaks-to-noisy the signal. However, by analyzing the recorded signal reflection evoked by a chirp sound stimulus, we propose a linear transformation of the Cardioreflector-Manoreflector to analyze the respiratory patterns variation and distinguish weak breathing signals under the handgrip-free. We compare the present signal with the ground truth recorded by ECG to approve the efficiency of the proposed method. Our experiments show AcHand achieves a MAE of 0.341 bpm, which is a significant improvement over the state-of-the-art devices, and enhancing detection capability across handgrip-free scenarios.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10462909/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10462909/</guid>
</item>
<item>
<title>Optimization Method for Node Deployment of Closed-Barrier Coverage in Hybrid Directional Sensor Networks</title>
<description><p><span><big>Optimization Method for Node Deployment of Closed-Barrier Coverage in Hybrid Directional Sensor Networks</big></span><br></p><p><span><small><i>Peng Wang; Yonghua Xiong; Jinhua She; Anjun Yu</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3378998</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Closed-barrier coverage is a new coverage problem that requires sensors on the barrier to form a closed loop. Due to the features of end-to-end connection, constructing closed barriers is a challenging task. In a Hybrid Directional Sensor Network, sensors can both rotate and move, making it more difficult to schedule sensors to construct closed barriers. To address this challenge, we propose an optimization method for node deployment of closed-barrier coverage (CCOM). First, selecting multiple initiators transforms the closed-barrier coverage problem into a simple linear-barrier construction problem. Then, a weighted barrier graph is used to search the optimal barrier path. Finally, design a rotation strategy and moving plan for the sensors, fully utilizing their rotation and moving capabilities to achieve closed-barrier coverage with the minimum number of sensors and energy consumption. The simulation experiment results show that compared with other advanced methods, our proposed method has better performance in terms of barrier construction success rate, minimum required number of mobile sensors, moving distance, and network lifetime.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10478814/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10478814/</guid>
</item>
<item>
<title>Enhanced Multicast Protocol for Low-power and Lossy IoT Networks</title>
<description><p><span><big>Enhanced Multicast Protocol for Low-power and Lossy IoT Networks</big></span><br></p><p><span><small><i>Issam Eddine Lakhlef; Badis Djamaa; Mustapha Reda Senouci; Abbas Bradai</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3375797</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Communication protocols in the Internet of Things (IoT) should take into account the resource-constrained nature of Low-power and Lossy Networks (LLNs). IP multicast protocols allow a packet to be routed from one source to multiple destinations in a single transmission. Hence, resources such as bandwidth, energy, and time are saved for a multitude of LLN applications, ranging from over-the-air programming and information sharing to device configuration and resource discovery. In this context, several multicast routing protocols have recently been proposed for LLNs, including the Multicast Protocol for LLNs (MPL). MPL has proven to be very reliable in many scenarios. However, the great resource consumption, especially in terms of energy and bandwidth, remains the main drawback of this protocol. In this paper, after a detailed overview, we provide an in-depth analysis of the MPL protocol to highlight its functional weaknesses. Then, we propose several improvements that touch on different areas to address such limitations. Extensive realistic simulations and experiments were performed to study the performance of the proposed improvements to MPL. The results obtained show that our proposals outperform the MPL protocol in terms of resource consumption (memory, bandwidth, and energy) while improving its performance in terms of end-to-end delay and maintaining the same reliability of data packet delivery.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10473686/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10473686/</guid>
</item>
<item>
<title>A 7.25μK Ultra-High Resolution MEMS Resonant Thermometer</title>
<description><p><span><big>A 7.25μK Ultra-High Resolution MEMS Resonant Thermometer</big></span><br></p><p><span><small><i>Zheng Wang; Liangbo Ma; Xiaorui Bie; Xingyin Xiong; Zhaoyang Zhai; Wuhao Yang; Yongjian Lu; Xudong Zou</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3370956</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This paper proposes an ultra-high-resolution MEMS resonant thermometer that exploits differences in Young’s modulus and coefficient of thermal expansion (CTE) between structural layers to increase thermal stress due to temperature variation. The increased thermal stress eventually acts on the resonator to produce a large frequency shift, and the temperature coefficient of frequency (TCF) can be increased to 17.4 times the original, which is consistent with theoretical analysis and finite element method simulations. The MEMS resonator chip is manufactured by the standard Silicon-on-insulator process and stacked on a ceramic chip carrier through multiple layers of materials. A self-sustaining oscillator, mainly composed of a packaged MEMS resonator chip and a low-noise application-specific integrated circuit (ASIC), is built to track the resonant frequency shift of the resonator. This MEMS resonant thermometer prototype demonstrates a high temperature coefficient of frequency (TCF) of -866.84ppm/K from -50°C to 40°C and exhibits good linearity. An ultra-high resolution of 7.25μK is achieved in the closed-loop experimental test. This is the best result achieved for a MEMS thermometer employing the resonant sensing paradigm to date.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10460450/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10460450/</guid>
</item>
<item>
<title>Design of MTM-based Multi-band Micro-Biosensor in Terahertz region as perfect absorber for Early-Stage Leukemia Diagnosis with sensitivity 18626373 TH...</title>
<description><p><span><big>Design of MTM-based Multi-band Micro-Biosensor in Terahertz region as perfect absorber for Early-Stage Leukemia Diagnosis with sensitivity 18626373 THz/RIU</big></span><br></p><p><span><small><i>Musa N. Hamza; Mohammad Tariqul Islam</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383522</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this article, a novel highly sensitive biosensor based on perfect metamaterial absorbers is presented. A detailed study is presented for several different models, different types of substrate materials, different types of resonant materials and substrate thicknesses showing very good sensitivity to any changes. The proposed biosensor exhibits high sensitivity to different polarization and incidence angles, ensuring its performance in terms of signal-to-noise ratio. The proposed biosensor was carefully compared with previously designed sensors and biosensors. The proposed biosensor exhibits amazing sensitivity, such as a quality factor of 41.1, a figure of merit (FOM) of 172466416 RIU-1, and a sensitivity (S) of 18626373 THz/RIU. The high sensitivity of the biosensor allows early detection of leukemia, demonstrating significant differences between leukemia and normal blood. Another interesting result of this article is the use of terahertz waves for imaging. Microwave imaging is performed for electric fields, magnetic fields, and power.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10494209/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10494209/</guid>
</item>
<item>
<title>LEPS: A lightweight and effective single-stage detector for pothole segmentation</title>
<description><p><span><big>LEPS: A lightweight and effective single-stage detector for pothole segmentation</big></span><br></p><p><span><small><i>Xiaoning Huang; Jintao Cheng; Qiuchi Xiang; Jun Dong; Jin Wu; Rui Fan; Xiaoyu Tang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2023.3330335</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Currently, the problem of potholes on urban roads is becoming increasingly severe. Identifying and locating road potholes promptly has become a major challenge for urban construction. Therefore, we proposed a lightweight and effective instance segmentation method called LEPS (Lightweight and Effective Pothole Segmentation Detector) for road pothole detection. To extract the image edge information and gradient information of potholes from the feature map more efficiently, we proposed a module that performs the convolutional super-position supplemented with a convolutional kernel to enhance spatial details for the backbone. We have designed a novel module applied to the Neck layer, improving the detection performance while reducing the parameters. To enable accurate segmentation of fine-grained features, we optimized the ProtoNet, which enables the segmentation head to generate high-quality masks for more accurate prediction. We have fully demonstrated the effectiveness of the method through a large number of comparative experiments. Our detector has excellent performance on two authoritative example datasets, URTIRI and COCO, and can successfully be applied to visual sensors to accurately detect and segment road potholes in real environments, accurately LEPS reached 0.892 and 0.648 in terms of Mask for AP50 and AP50:95, which is improved by 4.6% and 20.6% compared to the original model. These results demonstrate its strong competitiveness when compared to other models. Comprehensively, LEPS improves detection accuracy while maintaining lightweight, which allows the model to meet the practical application requirements of edge computing devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10315064/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10315064/</guid>
</item>
<item>
<title>UMHE: Unsupervised Multispectral Homography Estimation</title>
<description><p><span><big>UMHE: Unsupervised Multispectral Homography Estimation</big></span><br></p><p><span><small><i>Jeongmin Shin; Jiwon Kim; Seokjun Kwon; Namil Kim; Soonmin Hwang; Yukyung Choi</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383453</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Multispectral image alignment plays a crucial role in exploiting complementary information between different spectral images. Homography-based image alignment can be a practical solution considering a tradeoff between runtime and accuracy. Existing methods, however, have difficulty with multispectral images due to the additional spectral gap or require expensive human labels to train models. To solve these problems, this paper presents a comprehensive study on multispectral homography estimation in an unsupervised learning manner. We propose a curriculum data augmentation, an effective solution for models learning spectrum-agnostic representation by providing diverse input pairs. We also propose to use the phase congruency loss that explicitly calculates the reconstruction between images based on low-level structural information in the frequency domain. To encourage multispectral alignment research, we introduce a novel FLIR corresponding dataset that has manually labeled local correspondences between multispectral images. Our model achieves state-of-the-art alignment performance on the proposed FLIR correspondence dataset among supervised and unsupervised methods while running at
151 FPS
. Furthermore, our model shows good generalization ability on the M3FD dataset without finetuning.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10494213/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10494213/</guid>
</item>
<item>
<title>Calibration and Evaluation of a Low-Cost Optical Particulate Matter Sensor for Measurement of Lofted Lunar Dust</title>
<description><p><span><big>Calibration and Evaluation of a Low-Cost Optical Particulate Matter Sensor for Measurement of Lofted Lunar Dust</big></span><br></p><p><span><small><i>Abhay Vidwans; Jeffrey Gillis-Davis; Pratim Biswas</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3366436</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Several recent Earth-based investigations employed low-cost particulate matter sensors to address the lack of spatiotemporal resolution in air quality data. The lunar environment also has a particulate matter problem in the form of fine lofted and levitated dust particles. Natural and anthropogenic mobilized dust can cause a slew of difficulties for surface operations (deposition onto radiators, optical components, and mechanical devices). Despite the urgency of mitigating dust on the Moon and other airless bodies, the performance of low-cost sensors has not been critically evaluated for space applications. Upcoming long-term robotic and human exploration missions to the Moon necessitate a robust sensor that can monitor particulate matter levels and establish a spatially and temporally resolved global network. In this work, we calibrate two optical light-scattering particulate matter sensors against research-grade aerosol instruments for measuring the concentration of aerosolized lunar simulants. Sensors showed a stronger dependence on aerosol particle size distribution than particle composition. Vacuum testing showed a significant deviation in performance compared to atmospheric pressure, with a stronger dependency on lunar simulant. The predicted mass deposition, based on sensor output coupled with dust trajectory, was within an order of magnitude of the reference deposition. Our results demonstrate for the first time that low-cost particulate matter sensors can monitor dust concentrations with reasonable accuracy in a vacuum environment, with two caveats. First, precise calibrations must be performed with a dust simulant closely matching the particle size distribution of the target dust, and second, atmospheric pressure calibrations alone are insufficient.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10462021/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10462021/</guid>
</item>
<item>
<title>Reweighting Interacting Multiple-model Algorithm to Overcome Model Competition for Target Tracking in the Hybrid System</title>
<description><p><span><big>Reweighting Interacting Multiple-model Algorithm to Overcome Model Competition for Target Tracking in the Hybrid System</big></span><br></p><p><span><small><i>Guowei Li; Shurui Zhang; Yubing Han; Weixing Sheng; Thia Kirubarajan</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3369854</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The vicious competition of interacting multiple-model algorithm (IMM) is an inherent problem and would produce irreversible effects on IMM estimation results, especially combining with the radar system. In this paper, a novel reweighting IMM (RIMM) is proposed to overcome this issue. Firstly, the theoretical lower bound of model numbers in different situations is respectively provided through the analysis of IMM limitations. Furthermore, certificate the influence of model inaccuracy on the Kalman filter, which illustrates an effective method for reducing errors is increasing model numbers. Thirdly, the definition of model set density and the analysis of the true model space are given, and their connection establishes the standard of how to design the model set or add the model number. Finally, an effective method called RIMM is provided to overcome the competition caused by model increasing. The proposed RIMM holds strong adaptability for different model sets. The simulations of RIMM highlight the correctness and effectiveness of the proposed methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10462948/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10462948/</guid>
</item>
<item>
<title>A Stretchable and Wearable Ultrasonic Transducer Array for Bladder Volume Monitoring Application</title>
<description><p><span><big>A Stretchable and Wearable Ultrasonic Transducer Array for Bladder Volume Monitoring Application</big></span><br></p><p><span><small><i>Cong Pu; Ben Fu; Lehang Guo; Huixiong Xu; Chang Peng</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3382244</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Monitoring bladder volume is an essential application for patients with voiding dysfunction. Current wearable ultrasonic bladder monitors are rigid, bulky and cannot conform to human skin. This study presents a stretchable and wearable ultrasonic transducer array that can both conform to non-planar skin surfaces and continuously monitor bladder volume. The wearable transducer array mainly consists of 4 × 4 piezoelectric transducer elements, stretchable serpentine-based electrodes for electrical interconnection, electrically conductive adhesive as both matching and backing layers, and silicone elastomer for encapsulation. The transducer array has a center frequency of 3.9 MHz, -6 dB acoustic bandwidth of 57.6%, and up to 40% elastic stretchability. The volume of balloon-bladder phantom was estimated by a least square ellipsoid fitting method. With the true volume of balloon-bladder phantoms ranging from 10 mL to 405 mL, the proposed transducer array has a mean absolute percentage error of 9.4%, a mean absolute error of 11.9 mL and a coefficient of determination of 0.975, which demonstrates the capability of the proposed stretchable and wearable ultrasonic transducer array for bladder volume monitoring application.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10489841/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10489841/</guid>
</item>
<item>
<title>Intelligent fault diagnosis of bearing using multiwavelet perception kernel convolutional neural network</title>
<description><p><span><big>Intelligent fault diagnosis of bearing using multiwavelet perception kernel convolutional neural network</big></span><br></p><p><span><small><i>Yuanyuan Zhou; Hang Wang; Yongbin Liu; Xianzeng Liu; Zheng Cao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3370564</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Strong background noise characteristics of vibration signals cause issues with poor identification capability of features by fault diagnostic models. To address this issue, a method is proposed for intelligent fault diagnosis of bearing using multiwavelet perception kernel (MPK) and feature attention convolutional neural network (FA-CNN). First, four multiwavelet perception kernels are constructed to decompose the vibration signals in full-band multilevel. Second, improved multiwavelet information entropy (IMIE) of the frequency band components is calculated. The calculated component entropies of the corresponding frequency bands are integrated to construct frequency band clusters (FBC) from low to high frequencies. Third, joint approximate diagonalization eigen (JADE) is introduced to perform feature fusion for every FBC to eliminate redundant information, and fused features from low to high frequencies are obtained as original inputs. The FA-CNN bearing fault diagnosis framework is constructed for intelligent fault diagnosis of bearings. Finally, the effectiveness of the proposed method is verified by two cases. The results show that the proposed method has high fault feature recognition capability.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10458914/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10458914/</guid>
</item>
<item>
<title>Improved multipath mitigation using multiple trend-surface hemispherical map in GNSS precise point positioning</title>
<description><p><span><big>Improved multipath mitigation using multiple trend-surface hemispherical map in GNSS precise point positioning</big></span><br></p><p><span><small><i>Haijun Yuan; Zhetao Zhang; Xiufeng He; Jinwen Zeng; Hao Wang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3374458</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Global Navigation Satellite System (GNSS) Precise Point Positioning (PPP) is highly appreciated as a positioning technology; however, multipath restricts its accuracy and reliability. In this study, we proposed a Multiple Trend-surface Multipath Hemispherical Map (MT-MHM), where discrepant multipath fading frequency or orbit orientation of each satellite in each azimuth and elevation grid is considered. In a typical static scenario, consecutive 8-day observations are collected to extract multipath and conduct PPP experiments. Compared with traditional multipath hemispherical map based on trend-surface fitting, MT-MHM has better modelling performance of residuals and improves standard deviations of residuals for all satellites. Besides, the positioning errors and required convergence epochs of PPP are reduced by using MT-MHM. For the positioning accuracy of all solutions, MT-MHM exhibits 50.8, 15.2, and 27.9% improvements compared with that without multipath correction in the east, north, and up directions, respectively. In conclusion, our proposed MT-MHM exhibits better performance in terms of residual reduction, convergence time, and positioning accuracy in PPP.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10472420/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10472420/</guid>
</item>
<item>
<title>Design optimization and characterization of a 3D-printed tactile sensor for tissue palpation</title>
<description><p><span><big>Design optimization and characterization of a 3D-printed tactile sensor for tissue palpation</big></span><br></p><p><span><small><i>D. Lo Presti; L. Zoboli; A. Addabbo; D. Bianchi; A. Dimo; C. Massaroni; V. Altomare; A. Grasso; A. Gizzi; E. Schena</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3369337</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Manual palpation is a crucial medical procedure that relies on surface examination to detect internal tissue abnormalities, heavily reliant on healthcare professionals’ expertise and tactile sensitivity. To tackle these issues, smart palpation systems based on electrical or optical sensors have been developed to perform quantitative tactile measurements, crucial for identifying various solid tumors, including breast and prostate cancer by assessing tissue mechanical properties. In this context, fiber Bragg gratings (FBGs) are emerging as a promising technology due to their advantages (e.g., high metrological properties, multiplexing capacity, and easy packaging) making them ideal for tactile sensing. This study explores the benefits of FBG and 3D printing to develop a tactile sensor for tissue palpation. First, an optimization of the design of the sensing core of a previously developed probe was conducted through finite element analysis. The novel structure addresses the primary limitation of the previous solution, where non-uniform strain distribution on the encapsulated FBG causes compression on the grating with high risk of bending and breakage. In contrast, the modeled geometry ensures FBG elongation during tissue palpation. A 3D printing and characterization of the proposed solution was carried out to investigate the response of the enclosed FBG when pushed against different materials showing promising results in discriminating tissues according to their mechanical properties: the more rigid the indented substrate the higher the sensor output. This property will be fundamental for enhancing early tumor detection through superficial tissue palpation, advancing the efficacy of prevention measures.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10455966/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10455966/</guid>
</item>
<item>
<title>Deep Reinforcement Learning-based Joint Sequence Scheduling and Trajectory Planning in Wireless Rechargeable Sensor Networks</title>
<description><p><span><big>Deep Reinforcement Learning-based Joint Sequence Scheduling and Trajectory Planning in Wireless Rechargeable Sensor Networks</big></span><br></p><p><span><small><i>Chengpeng Jiang; Wencong Chen; Ziyang Wang; Wendong Xiao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3373664</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Mobile charging has become a popular and efficient method for replenishing energy. This is done with a mobile charger (MC) and wireless energy transfer technology (WET), which helps to alleviate the issue of energy constraints in wireless rechargeable sensor networks (WRSNs). Notably, designing mobile charging scheduling schemes is essential for improving charging performance. Most current studies assume that the networks are obstacle-free. Unlike the existing studies, this paper focuses on joint sequence scheduling and trajectory planning problems (JSSTP), which assumes that the network has multiple static obstacles. To address this issue, we propose a novel deep reinforcement learning-based JSSTP (DRL-JSSTP) that enables the MC to avoid obstacles and reach the charging target to charge the sensors fully. This approach maximizes energy usage efficiency and sensor survival rate while satisfying MC energy capacity constraints. DRL-JSSTP includes a charging target selector and a trajectory planner, which determine the index of the next charging target and plan the movement trajectory to avoid obstacles, respectively. We adopt 1-D convolutional neural networks to extract feature information about the environment state and gated recurrent units to predict the charging decisions. Simulation results demonstrate that DRL-JSSTP outperforms existing approaches, achieving higher energy usage efficiency and sensor survival rate.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10466524/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10466524/</guid>
</item>
<item>
<title>A Simultaneous Wirel... |
http://localhost:1200/ieee/author/37281702200 - Failed ❌ |
This problem is related to assets/build/router.json, I don't know how to update this file. Could you give me some advice? |
|
This PR is stale because it has been opened for more than 3 weeks with no activity. Comment or this will be closed in 7 days. |
lewisjiang
reviewed
May 4, 2024
|
Successfully generated as following: http://localhost:1200/ieee/journal/8782710 - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Open Journal of Signal Processing</title>
<link>https://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=8782710</link>
<atom:link href="http://localhost:1200/ieee/journal/8782710" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Open Journal of Signal Processing - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Sat, 04 May 2024 13:09:07 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Synthbuster: Towards Detection of Diffusion Model Generated Images</title>
<description><p><span><big>Synthbuster: Towards Detection of Diffusion Model Generated Images</big></span><br></p><p><span><small><i>Quentin Bammey</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337714</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Synthetically-generated images are getting increasingly popular. Diffusion models have advanced to the stage where even non-experts can generate photo-realistic images from a simple text prompt. They expand creative horizons but also open a Pandora's box of potential disinformation risks. In this context, the present corpus of synthetic image detection techniques, primarily focusing on older generative models like Generative Adversarial Networks, finds itself ill-equipped to deal with this emerging trend. Recognizing this challenge, we introduce a method specifically designed to detect synthetic images produced by diffusion models. Our approach capitalizes on the inherent frequency artefacts left behind during the diffusion process. Spectral analysis is used to highlight the artefacts in the Fourier transform of a residual image, which are used to distinguish real from fake images. The proposed method can detect diffusion-model-generated images even under mild
jpeg
compression, and generalizes relatively well to unknown models. By pioneering this novel approach, we aim to fortify forensic methodologies and ignite further research into the detection of AI-generated images.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334046/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334046/</guid>
</item>
<item>
<title>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</title>
<description><p><span><big>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</big></span><br></p><p><span><small><i>Yanbin Zou; Jingna Fan; Zekai Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we consider using time-of-arrival (TOA) measurements from a single moving receiver to locate a moving target at constant velocity that emits a periodic signal with unknown signal period. First, we give the TOA measurement model and deduce the Cram
$\acute{\text{e}}$
r-Rao lower bounds (CRLB). Then, we formulate a nonlinear least squares (NLS) problem to estimate the unknown parameters. We use semidefinite programming (SDP) techniques to relax the nonconvex NLS problem. However, it is shown that the SDP localization algorithm cannot provide a high-quality solution. Subsequently, we develop a fixed point iteration (FPI) method to improve the performance of the SDP algorithm. In addition, we also consider the presence of receiver position errors and develop the corresponding localization algorithm. Numerical simulations are conducted to demonstrate the localization performance of the proposed algorithms by comparing them with the CRLB.
Index Term
-Fixed point iteration (FPI), semidefinite programming (SDP), single moving receiver, target localization, time-of-arrival (TOA).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336384/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336384/</guid>
</item>
<item>
<title>The Neural-SRP Method for Universal Robust Multi-Source Tracking</title>
<description><p><span><big>The Neural-SRP Method for Universal Robust Multi-Source Tracking</big></span><br></p><p><span><small><i>Eric Grinstein; Christopher M. Hicks; Toon van Waterschoot; Mike Brookes; Patrick A. Naylor</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340057</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Neural networks have achieved state-of-the-art performance on the task of acoustic Direction-of-Arrival (DOA) estimation using microphone arrays. Neural models can be classified as end-to-end or hybrid, each class showing advantages and disadvantages. This work introduces Neural-SRP, an end-to-end neural network architecture for DOA estimation inspired by the classical Steered Response Power (SRP) method, which overcomes limitations of current neural models. We evaluate the architecture on multiple scenarios, namely, multi-source DOA tracking and single-source DOA tracking under the presence of directional and diffuse noise. The experiments demonstrate that our proposed method compares favourably in terms of computational and localization performance with established neural methods on various recorded and simulated benchmark datasets.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345765/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345765/</guid>
</item>
<item>
<title>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</title>
<description><p><span><big>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</big></span><br></p><p><span><small><i>Alexander Venus; Erik Leitinger; Stefan Tertinek; Klaus Witrisal</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper presents a neural-enhanced probabilistic model and corresponding factor graph-based sum-product algorithm for robust localization and tracking in multipath-prone environments. The introduced hybrid probabilistic model consists of physics-based and data-driven measurement models capturing the information contained in both, the line-of-sight (LOS) component as well as in multipath components (NLOS components). The physics-based and data-driven models are embedded in a joint Bayesian framework allowing to derive from first principles a factor graph-based algorithm that fuses the information of these models. The proposed algorithm uses radio signal measurements from multiple base stations to robustly estimate the mobile agent's position together with all model parameters. It provides high localization accuracy by exploiting the position-related information of the LOS component via the physics-based model and robustness by exploiting the geometric imprint of multipath components independent of the propagation channel via the data-driven model. In a challenging numerical experiment involving obstructed LOS situations to all anchors, we show that the proposed sequential algorithm significantly outperforms state-of-the-art methods and attains the posterior Cramér-Rao lower bound even with training data limited to local regions.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336409/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336409/</guid>
</item>
<item>
<title>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</title>
<description><p><span><big>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</big></span><br></p><p><span><small><i>AMIT EFRAIM; Joseph M. Francos</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In solving a model fitting problem, the existence of outliers in the set of measurements can have a devastating effect on the solution accuracy. Traditionally, in order to overcome this problem, robust point cloud registration algorithms are composed of transformation hypothesis generation, followed by hypothesis evaluation aimed at selecting the best hypothesized estimate. Hypotheses evaluation is commonly performed using the sample consensus criterion. However, since this criterion accounts only for the consensus size, it fails when the maximal sample consensus is incorrect. We propose a new hypothesis evaluation approach, generalizing the sample consensus approach, where instead of the sample consensus, the transformation that maximizes the point clouds feature correlation is selected as the best hypothesis. The feature vector at each point contains information such as on local geometry and semantic context. Utilizing this information in the hypotheses evaluation and selection procedure allows for a correct decision even when the hypothesis yielding the maximal sample consensus is false. Consequently, the probability of selecting the correct model increases. We show both mathematically and empirically that substituting the sample consensus criterion with the proposed point cloud feature correlation hypothesis test (PC-FCHT) lowers the probability of large registration errors, compared to using the special case of sample consensus. The proposed PC-FCHT is applicable to any algorithm that follows the hypothesis generation and evaluation scheme, potentially improving the success rates of a wide family of point cloud registration algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345750/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345750/</guid>
</item>
<item>
<title>Joint PAPR and OBP Reduction for NC-OFDM Systems</title>
<description><p><span><big>Joint PAPR and OBP Reduction for NC-OFDM Systems</big></span><br></p><p><span><small><i>Hsuan-Fu Wang; Fang-Biau Ueng; Bo-Heng Yeh</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3329757</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The spectrum resource is always a critical issue for wireless communications since it directly impacts the data rate and capacity. However, the problem of spectrum resource scarcity always exists. Moreover, spectrum resource scarcity becomes more severe as new communication technologies and wireless applications sprout. Noncontiguous orthogonal frequency division multiplexing (NC-OFDM) is a multicarrier method for bandwidth utilization. Unfortunately, this system has two fatal defects: high peak-to-average power ratio (PAPR) and considerable out-of-band power (OBP), which are detrimental to the system's performance. To solve these two problems, we propose a convex optimization-based method for joint PAPR and OBP reduction in NC-OFDM Systems. The strategy is to permit the secondary user to utilize the unoccupied spectrum of the primary user with dynamic spectrum sharing (DSS) based on a cognitive radio network (CRN). To this end, a flexible system operating over noncontiguous bands and DSS scenarios is necessary. The simulation results have shown that our method could effectively improve the overall performance and outperform other schemes, i.e., projections onto convex sets (POCS) and alternating projections onto convex and non-convex sets (APOCNCS), without harming the transmission of the primary system. The collaboration between secondary and primary systems is viable with the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10305259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10305259/</guid>
</item>
<item>
<title>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</title>
<description><p><span><big>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</big></span><br></p><p><span><small><i>Gal Itzhak; Israel Cohen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339433</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article presents a Kronecker-product (KP) beamforming approach incorporating sparse concentric circular arrays (SCCAs). The locations of the microphones on the SCCA are optimized concerning the broadband array directivity over a wide range of direction-of-arrival (DOA) deviations of a desired signal. A maximum directivity factor (MDF) sub-beamformer is derived accordingly with the optimal locations. Then, we propose two global beamformers obtained as a Kronecker product of a uniform linear array (ULA) and the SCCA sub-beamformer. The global beamformers differ by the type of the ULA, which is designed either as an MDF sub-beamformer along the
$\mathsf {x}$
-axis or as a maximum white noise gain sub-beamformer along the
$\mathsf {y}$
-axis. We analyze the performance of the proposed beamformers in terms of the directivity factor, the white noise gain, and their spatial beampatterns. Compared to traditional beamformers, the proposed beamformers exhibit considerably larger tolerance to DOA deviations concerning both the azimuth and elevation angles. Experimental results with speech signals in noisy and reverberant environments demonstrate that the proposed approach outperforms traditional beamformers regarding the perceptual evaluation of speech quality (PESQ) and short-time objective intelligibility (STOI) scores when the desired speech signals deviate from the nominal DOA.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342869/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342869/</guid>
</item>
<item>
<title>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</title>
<description><p><span><big>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</big></span><br></p><p><span><small><i>Karn N. Watcharasupat; Chih-Wei Wu; Yiwei Ding; Iroro Orife; Aaron J. Hipple; Phillip A. Williams; Scott Kramer; Alexander Lerch; William Wolcott</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339428</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Cinematic audio source separation is a relatively new subtask of audio source separation, with the aim of extracting the dialogue, music, and effects stems from their mixture. In this work, we developed a model generalizing the Bandsplit RNN for any complete or overcomplete partitions of the frequency axis. Psychoacoustically motivated frequency scales were used to inform the band definitions which are now defined with redundancy for more reliable feature extraction. A loss function motivated by the signal-to-noise ratio and the sparsity-promoting property of the 1-norm was proposed. We additionally exploit the information-sharing property of a common-encoder setup to reduce computational complexity during both training and inference, improve separation performance for hard-to-generalize classes of sounds, and allow flexibility during inference time with detachable decoders. Our best model sets the state of the art on the Divide and Remaster dataset with performance above the ideal ratio mask for the dialogue stem.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342812/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342812/</guid>
</item>
<item>
<title>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</title>
<description><p><span><big>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</big></span><br></p><p><span><small><i>Miguel Ferrer; María de Diego; Alberto Gonzalez</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340106</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The LMS algorithm is widely employed in adaptive systems due to its robustness, simplicity, and reasonable performance. However, it is well known that this algorithm suffers from a slow convergence speed when dealing with colored reference signals. Numerous variants and alternative algorithms have been proposed to address this issue, though all of them entail an increase in computational cost. Among the proposed alternatives, the affine projection algorithm stands out. This algorithm has the peculiarity of starting from
$N$
data vectors of the reference signal. It transforms these vectors into as many data vectors suitably normalized in energy and mutually orthogonal. In this work, we propose a version of the LMS algorithm that, similar to the affine projection algorithm, starts from
$N$
data vectors of the reference signal but corrects them by using only a scalar factor that functions as a convergence step. Our goal is to align the behavior of this algorithm with the behavior of the affine projection algorithm without significantly increasing the computational cost of the LMS.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345730/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345730/</guid>
</item>
<item>
<title>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</title>
<description><p><span><big>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</big></span><br></p><p><span><small><i>Zongyao Li; Ren Togo; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340616</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article proposes a method for transferring knowledge of semantic segmentation from a labeled source domain to an unlabeled target domain without using the source-domain data. Such a problem is called source-data-free domain adaptation, in which a pre-trained source-domain model and the unlabeled target-domain data are used to transfer the label knowledge across the domains. Like most previous methods, our method uses pseudo labels for distilling and transferring the source-domain knowledge. On the basis of the pseudo-label learning, our method improves the domain adaptation performance in two innovative ways: 1) reducing the domain differences by source-data-free style transfer and 2) exploring the style diversity within the target domain by style modification. To this end, we introduce two additional modules: 1) an inter-domain style transfer module which aligns the feature statistics of the source and target domains before producing the pseudo labels thereby improving the pseudo labels' accuracy, and 2) an intra-domain style modification module which modifies the image styles within the target domain for learning intra-domain style-invariant features. Our method with the two modules outperforms previous source-data-free domain adaptation methods in two commonly used benchmarks. Moreover, our method is well compatible with the previous methods for further improvement.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10356748/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10356748/</guid>
</item>
<item>
<title>Contactless Skin Blood Perfusion Imaging via Multispectral Information, Spectral Unmixing and Multivariable Regression</title>
<description><p><span><big>Contactless Skin Blood Perfusion Imaging via Multispectral Information, Spectral Unmixing and Multivariable Regression</big></span><br></p><p><span><small><i>Liliana Granados-Castro; Omar Gutierrez-Navarro; Aldo Rodrigo Mejia-Rodriguez; Daniel Ulises Campos-Delgado</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2024.3381892</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Noninvasive methods for assessing in-vivo skin blood perfusion parameters, such as hemoglobin oxygenation, are crucial for diagnosing and monitoring microvascular diseases. This approach is particularly beneficial for patients with compromised skin, where standard contact-based clinical devices are inappropriate. For this goal, we propose the analysis of multimodal data from an occlusion protocol applied to 18 healthy participants, which includes multispectral imaging of the whole hand and reference photoplethysmography information from the thumb. Multispectral data analysis was conducted using two different blind linear unmixing methods: principal component analysis (PCA), and extended blind endmember and abundance extraction (EBEAE). Perfusion maps for oxygenated and deoxygenated hemoglobin changes in the hand were generated using linear multivariable regression models based on the unmixing methods. Our results showed high accuracy, with
$\text {R}^{2}$
-adjusted values, up to 0.90
$\pm$
0.08. Further analysis revealed that using more than four characteristic components during spectral unmixing did not improve the fit of the model. Bhattacharyya distance results showed that the fitted models with EBEAE were more sensitive to hemoglobin changes during occlusion stages, up to four times higher than PCA. Our study concludes that multispectral imaging with EBEAE is effective in quantifying changes in oxygenated hemoglobin levels, especially when using 3 to 4 characteristic components. Our proposed method holds promise for the noninvasive diagnosis and monitoring of superficial microvascular alterations across extensive anatomical regions.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10480236/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10480236/</guid>
</item>
<item>
<title>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</title>
<description><p><span><big>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</big></span><br></p><p><span><small><i>Tobias Kabzinski; Peter Jax</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337721</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Quasi-continuous system identification of time-variant linear acoustic systems can be applied in various audio signal processing applications when numerous acoustic transfer functions must be measured. A prominent application is measuring head-related transfer functions. We treat the underlying multiple-input-multiple-output (MIMO) system identification problem in a state-space model as a joint estimation problem for states, representing impulse responses, and state-space model parameters using the expectation maximization (EM) algorithm. We address limitations of prior work by imposing different model structures, especially for dependencies within a (transformed) state vector. This results in block diagonal matrix structures, for which we derive M-step update rules. Making assumptions about this model structure and choosing a block size for a given application define the computational complexity. In examples, we found that applying this framework yields improvements of up to 10 dB in relative system distance in comparison to a conventional method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334061/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334061/</guid>
</item>
<item>
<title>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</title>
<description><p><span><big>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</big></span><br></p><p><span><small><i>Damir Rakhimov; Martin Haardt</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337729</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this paper, we present an analytical performance assessment of 2-D Tensor ESPRIT in terms of physical parameters. We show that the error in the
$r$
-mode depends only on two components, irrespective of the dimensionality of the problem. We obtain analytical expressions in closed form for the mean squared error (MSE) in each dimension as a function of the signal-to-noise (SNR) ratio, the array steering matrices, the number of antennas, the number of snapshots, the selection matrices, and the signal correlation. The derived expressions allow a better understanding of the difference in performance between the tensor and the matrix versions of the ESPRIT algorithm. The simulation results confirm the coincidence between the presented analytical expression and the curves obtained via Monte Carlo trials. We analyze the behavior of each of the two error components in different scenarios.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334446/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334446/</guid>
</item>
<item>
<title>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</title>
<description><p><span><big>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</big></span><br></p><p><span><small><i>Yaman Kındap; Simon Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343341</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this work we study linear vector stochastic differential equation (SDE) models driven by the generalised hyperbolic (GH) Lévy process for inference in continuous-time non-Gaussian filtering problems. The GH family of stochastic processes offers a flexible framework for modelling of non-Gaussian, heavy-tailed characteristics and includes the normal inverse-Gaussian, variance-gamma and Student-t processes as special cases. We present continuous-time simulation methods for the solution of vector SDE models driven by GH processes and novel inference methodologies using a variant of sequential Markov chain Monte Carlo (MCMC). As an example a particular formulation of Langevin dynamics is studied within this framework. The model is applied to both a synthetically generated data set and a real-world financial series to demonstrate its capabilities.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360268/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360268/</guid>
</item>
<item>
<title>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</title>
<description><p><span><big>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</big></span><br></p><p><span><small><i>James M. Cozens; Simon J. Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344048</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper proposes a probabilistic approach for extracting time-varying and irregular time signature information from polyphonic audio extracts, subsequently providing beat and bar line positions given inferred time signature divisions. This is achieved via dynamically evaluating the beat tempo as a function of time through finding an optimal compromise in beat and bar alignment in the time and tempo domains. Time signature divisions are determined based on a new representation, termed the Metrogram, that presents time-varying information regarding rhythmic and metric periodicities in the Tempogram. Our methodology is characterised by its ability to provide a distribution over metric interpretations, offering insights into the diverse ways music can be rhythmically perceived. Results indicate high-level accuracy for a variety of polyphonic extracts containing irregular, complex, irrational, and time-varying time signatures. Accuracy rivalling state-of-the-art methodologies is also reported in a beat tracking task performed on the standard Ballroom Dataset. The paper offers insights into the field of dynamic time signature recognition and beat tracking, offering a valuable and versatile resource for the analysis, composition, and performance of music.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363392/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363392/</guid>
</item>
<item>
<title>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</title>
<description><p><span><big>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</big></span><br></p><p><span><small><i>Yuto Watanabe; Ren Togo; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343335</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Although text-guided image manipulation approaches have demonstrated highly accurate performance for editing the appearance of images in a virtual or simple scenario, their real-world applications face significant challenges. The primary cause of these challenges is the misalignment in the distribution of training and real-world data, which leads to unstable text-guided image manipulation. In this work, we propose a novel framework called TolerantGAN and tackle the new task of real-world text-guided image manipulation independent of the training data. To achieve this, we introduce two key concepts of a border smoothly connection module (BSCM) and a manipulation direction-based attention module (MDAM). BSCM smoothens the misalignment in the distribution of training and real-world data. MDAM extracts only regions highly relevant for image manipulation and assists in reconstructing unobserved objects in the training data. For in-the-wild input images of various classes, TolerantGAN robustly outperforms the state-of-the-art methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360283/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360283/</guid>
</item>
<item>
<title>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</title>
<description><p><span><big>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</big></span><br></p><p><span><small><i>Anastasia Avdeeva; Aleksei Gusev; Tseren Andzhukaev; Artem Ivanov</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343342</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Whispered speech is a quiet voice without vocalization. One of the common cases of using whispered speech is a technique that can help overcome stuttering. But whispered speech can be uncomfortable and difficult to understand in everyday communication. To address these problems, we propose a method of low-delayed whisper-to-speech voice conversion, which can be useful in real life communication of people with disordered speech. As part of our research, we study the impact of streaming Automatic Speech Recognition models on the quality of voice conversion, comparing different streaming models and methods for model adaptation to streaming settings, and showing the importance of using such models in cases of low-delayed voice conversion.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360259/</guid>
</item>
<item>
<title>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</title>
<description><p><span><big>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</big></span><br></p><p><span><small><i>Denis C. Ilie-Ablachim; Andra Băltoiu; Bogdan Dumitrescu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344313</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We propose sparse representation and dictionary learning algorithms for dictionaries whose atoms are characterized by Gaussian probability distributions around some central atoms. This extension of the space covered by the atoms permits a better characterization of localized features of the signals. The representations are computed by trading-off representation error and the probability of the actual atoms used in the representation. We propose two types of algorithms: a greedy one, similar in spirit with Orthogonal Matching Pursuit, and one based on L1-regularization. We apply our new algorithms to unsupervised anomaly detection, where the representation error is used as anomaly score. The flexibility of our approach appears to improve more the representations of the many normal signals than those of the few outliers, at least for an anomaly type called dependency, thus improving the detection quality. Comparison with both standard dictionary learning algorithms and established anomaly detection methods is favorable.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365333/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365333/</guid>
</item>
<item>
<title>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</title>
<description><p><span><big>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</big></span><br></p><p><span><small><i>Yuya Moroto; Yingrui Ye; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344079</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>There are various sentiment theories for categorizing human sentiments into several discrete sentiment categories, which means that the theory used for training sentiment prediction methods does not always match that used in the test phase. As a solution to this problem, zero-shot visual sentiment prediction methods have been proposed to predict unseen sentiments for which no images are available in the training phase. However, the training of these previous zero-shot methods relies on a single sentiment theory, which limits their ability to handle sentiments from other theories. Thus, this article proposes a more robust zero-shot visual sentiment prediction method that can handle cross-domain sentiments defined in different sentiment theories. Specifically, by focusing on the fact that sentiments are abstract concepts common to humans regardless of whether their theories are different, we incorporate knowledge distillation into our method to construct a teacher–student model that can train the implicit relationships between sentiments defined in different sentiment theories. Furthermore, to enhance sentiment discrimination capability and strengthen the implicit relationships between sentiments, we introduce a novel sentiment loss between the teacher and student models. In this way, our model becomes robust to unseen sentiments by exploiting the implicit relationships between sentiments. The contributions of this article are the introduction of knowledge distillation and a novel sentiment loss between the teacher and student models for zero-shot visual sentiment prediction, and improved performance of zero-shot visual sentiment prediction. Experiments on several open datasets demonstrate the effectiveness of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363382/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363382/</guid>
</item>
<item>
<title>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</title>
<description><p><span><big>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</big></span><br></p><p><span><small><i>Chengen Liu; Geert Leus; Elvin Isufi</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339376</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The edge flow reconstruction task consists of retreiving edge flow signals from corrupted or incomplete measurements. This is typically solved by a regularized optimization problem on higher-order networks such as simplicial complexes and the corresponding regularizers are chosen based on prior knowledge. Tailoring this prior to the setting of interest can be challenging or it may not even be possible. Thus, we consider to learn this prior knowledge via a model-based deep learning approach. We propose a new regularized optimization problem for the simplicial edge flow reconstruction task, the simplicial ElasticNet, which combines the advantages of the
$\ell _{1}$
and
$\ell _{2}$
norms. We solve the simplicial ElasticNet problem via the multi-block alternating direction method of multipliers (ADMM) algorithm and provide conditions on its convergence. By unrolling the ADMM iterative steps, we develop a model-based neural network with a low requirement on the number of training data. This unrolling network replaces the fixed parameters in the iterative algorithm by learnable weights, thus exploiting the neural network's learning capability while preserving the iterative algorithm's interpretability. We enhance this unrolling network via simplicial convolutional filters to aggregate information from the edge flow neighbors, ultimately, improving the network learning expressivity. Extensive experiments on real-world and synthetic datasets validate the proposed approaches and show considerable improvements over both baselines and traditional non-model-based neural networks.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342735/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342735/</guid>
</item>
<item>
<title>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</title>
<description><p><span><big>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</big></span><br></p><p><span><small><i>Jim Beckers; Bart Van Erp; Ziyue Zhao; Kirill Kondrashov; Bert De Vries</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337718</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Bayesian model reduction provides an efficient approach for comparing the performance of all nested sub-models of a model, without re-evaluating any of these sub-models. Until now, Bayesian model reduction has been applied mainly in the computational neuroscience community on simple models. In this paper, we formulate and apply Bayesian model reduction to perform principled pruning of Bayesian neural networks, based on variational free energy minimization. Direct application of Bayesian model reduction, however, gives rise to approximation errors. Therefore, a novel iterative pruning algorithm is presented to alleviate the problems arising with naive Bayesian model reduction, as supported experimentally on the publicly available UCI datasets for different inference algorithms. This novel parameter pruning scheme solves the shortcomings of current state-of-the-art pruning methods that are used by the signal processing community. The proposed approach has a clear stopping criterion and minimizes the same objective that is used during training. Next to these benefits, our experiments indicate better model performance in comparison to state-of-the-art pruning schemes.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334001/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334001/</guid>
</item>
<item>
<title>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</title>
<description><p><span><big>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</big></span><br></p><p><span><small><i>Anh Minh Truong; Wilfried Philips; Peter Veelaert</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340064</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Many image-based rendering (IBR) methods rely on depth estimates obtained from structured light or time-of-flight depth sensors to synthesize novel views from sparse camera networks. However, these estimates often contain missing or noisy regions, resulting in an incorrect mapping between source and target views. This situation makes the fusion process more challenging, as the visual information is misaligned, inconsistent, or missing. In this work, we first implement a lightweight network based on the transformer, which is well-known for its capability to model long-range relationships within input data, to extract spatial features from color images. These features are then used to enhance the quality of completed depth maps. Furthermore, we combine a sequential deep neural network with a spatial attention mechanism to effectively fuse the projected features from multiple source viewpoints. This approach enables us to integrate information from an arbitrary number of source viewpoints as well as improve accuracy in synthesized views. Experimental results on challenging datasets demonstrate that our method achieves superior synthesized image quality compared to state-of-the-art (SOTA) methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345792/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345792/</guid>
</item>
<item>
<title>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</title>
<description><p><span><big>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</big></span><br></p><p><span><small><i>Reza Mirzaeifard; Naveen K. D. Venkategowda; Vinay Chakravarthi Gogineni; Stefan Werner</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344395</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper investigates quantile regression in the presence of non-convex and non-smooth sparse penalties, such as the minimax concave penalty (MCP) and smoothly clipped absolute deviation (SCAD). The non-smooth and non-convex nature of these problems often leads to convergence difficulties for many algorithms. While iterative techniques such as coordinate descent and local linear approximation can facilitate convergence, the process is often slow. This sluggish pace is primarily due to the need to run these approximation techniques until full convergence at each step, a requirement we term as a
secondary convergence iteration
. To accelerate the convergence speed, we employ the alternating direction method of multipliers (ADMM) and introduce a novel single-loop smoothing ADMM algorithm with an increasing penalty parameter, named SIAD, specifically tailored for sparse-penalized quantile regression. We first delve into the convergence properties of the proposed SIAD algorithm and establish the necessary conditions for convergence. Theoretically, we confirm a convergence rate of
$o({k^{-\frac{1}{4}}})$
for the sub-gradient bound of the augmented Lagrangian, where
$k$
denotes the number of iterations. Subsequently, we provide numerical results to showcase the effectiveness of the SIAD algorithm. Our findings highlight that the SIAD method outperforms existing approaches, providing a faster and more stable solution for sparse-penalized quantile regression.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365338/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365338/</guid>
</item>
<item>
<title>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</title>
<description><p><span><big>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</big></span><br></p><p><span><small><i>Oliver Lang; Christian Hofbauer; Reinhard Feger; Mario Huemer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343308</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>A promising waveform candidate for future joint sensing and communication systems is orthogonal frequency-division multiplexing (OFDM). For such systems, supporting multiple transmit antennas requires multiplexing methods for the generation of orthogonal transmit signals, where equidistant subcarrier interleaving (ESI) is the most popular multiplexing method. In this work, we analyze a multiplexing method called Doppler-division multiplexing (DDM). This method applies a phase shift from OFDM symbol to OFDM symbol to separate signals transmitted by different Tx antennas along the velocity axis of the range-Doppler map. The main focus of this work lies on the implications of DDM on the communication task. It will be shown that for DDM, the channels observed in the communication receiver are heavily time-varying, preventing any meaningful transmission of data when not taken into account. In this work, a communication system designed to combat these time-varying channels is proposed, which includes methods for data estimation, synchronization, and channel estimation. Bit error ratio (BER) simulations demonstrate the superiority of this communications system compared to ESI-based systems.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360223/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360223/</guid>
</item>
<item>
<title>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</title>
<description><p><span><big>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</big></span><br></p><p><span><small><i>Nils L. Westhausen; Bernd T. Meyer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343320</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we introduce a causal low-latency low-complexity approach for binaural multichannel blind speaker separation in noisy reverberant conditions. The model, referred to as Group Communication Binaural Filter and Sum Network (GCBFSnet) predicts complex filters for filter-and-sum beamforming in the time-frequency domain. We apply Group Communication (GC), i.e., latent model variables are split into groups and processed with a shared sequence model with the aim of reducing the complexity of a simple model only containing one convolutional and one recurrent module. With GC we are able to reduce the size of the model by up to 83% and the complexity up to 73% compared to the model without GC, while mostly retaining performance. Even for the smallest model configuration, GCBFSnet matches the performance of a low-complexity TasNet baseline in most metrics despite the larger size and higher number of required operations of the baseline.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360275/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360275/</guid>
</item>
<item>
<title>Reverse Ordering Techniques for Attention-Based Channel Prediction</title>
<description><p><span><big>Reverse Ordering Techniques for Attention-Based Channel Prediction</big></span><br></p><p><span><small><i>Valentina Rizzello; Benedikt Böck; Michael Joham; Wolfgang Utschick</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344024</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Channel state information (CSI) is crucial for enhancing the performance of wireless systems by allowing to adjust the transmission strategies based on the current channel conditions. However, obtaining precise CSI is difficult because of the fast-changing channel conditions caused by multi-path fading. An inaccurate CSI hinders the performance of various adaptive wireless systems, highlighting the need for channel prediction techniques to effectively mitigate the drawbacks of outdated CSI. Conventional methods typically depend on assumptions regarding user velocity or require knowledge of the Doppler frequency. In contrast to existing approaches, we aim for a more robust and practical solution by training neural networks without making any assumptions about user velocity, relying solely on noisy channel observations during training. Specifically, we adapt both the sequence-to-sequence with attention (Seq2Seq-attn) and transformer models for channel prediction. Additionally, a new technique called reverse positional encoding is introduced in the transformer model to improve the robustness of the model against varying sequence lengths. Similarly, the encoder outputs of the Seq2Seq-attn model are reversed prior to the application of attention mechanisms. By means of simulations, we show that these proposed techniques enable the models to effectively capture relationships within sequences of channel snapshots without increasing the complexity. Importantly, this capability remains robust across varying sequence lengths, representing a substantial improvement over existing methodologies.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363354/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363354/</guid>
</item>
<item>
<title>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</title>
<description><p><span><big>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</big></span><br></p><p><span><small><i>Sarina Meyer; Xiaoxiao Miao; Ngoc Thang Vu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344375</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Speaker anonymization is the task of modifying a speech recording such that the original speaker cannot be identified anymore. Since the first Voice Privacy Challenge in 2020, along with the release of a framework, the popularity of this research topic is continually increasing. However, the comparison and combination of different anonymization approaches remains challenging due to the complexity of evaluation and the absence of user-friendly research frameworks. We therefore propose an efficient speaker anonymization and evaluation framework based on a modular and easily extendable structure, almost fully in Python. The framework facilitates the orchestration of several anonymization approaches in parallel and allows for interfacing between different techniques. Furthermore, we propose modifications to common evaluation methods which improves the quality of the evaluation and reduces their computation time by 65 to 95%, depending on the metric. Our code is fully open source.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365329/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365329/</guid>
</item>
<item>
<title>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</title>
<description><p><span><big>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</big></span><br></p><p><span><small><i>Alessandro Ilic Mezza; Matteo Amerena; Alberto Bernardini; Augusto Sarti</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343318</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Real-time audio communications over IP have become essential to our daily lives. Packet-switched networks, however, are inherently prone to jitter and data losses, thus creating a strong need for effective packet loss concealment (PLC) techniques. Though solutions based on deep learning have made significant progress in that direction as far as speech is concerned, extending the use of such methods to applications of Networked Music Performance (NMP) presents significant challenges, including high fidelity requirements, higher sampling rates, and stringent temporal constraints associated to the simultaneous interaction between remote musicians. In this article, we present PARCnet, a hybrid PLC method that utilizes a feed-forward neural network to estimate the time-domain residual signal of a parallel linear autoregressive model. Objective metrics and a listening test show that PARCnet provides state-of-the-art results while enabling real-time operation on CPU.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360264/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360264/</guid>
</item>
<item>
<title>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</title>
<description><p><span><big>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</big></span><br></p><p><span><small><i>Katerina Zmolikova; Michael Syskind Pedersen; Jesper Jensen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Supervised learning-based speech enhancement methods often work remarkably well in acoustic situations represented in the training corpus but generalize poorly to out-of-domain situations, i.e. situations not seen during training. This stands in the way of further improvement of these methods in realistic scenarios, as collecting paired noisy-clean recordings in the target application domain is typically not feasible. Recording noisy-only in-domain data is, though, much more practical. In this article, we tackle the problem of unsupervised domain adaptation in speech enhancement. Specifically, we propose a way to use in-domain noisy-only data in the training of a neural network to improve upon a model trained solely on out-of-domain paired data. For this, we make use of masked spectrogram prediction, a technique from self-supervised learning that aims to interpolate masked regions of a spectrogram. We hypothesize that masked spectrogram prediction encourages learning of features that represent well both speech and noise components of the noisy signals. These features can then be used to train a more robust speech enhancement system. We evaluate the proposed method on the VoiceBank-DEMAND and LibriFSD50k databases, with WSJ0-CHiME3 serving as the out-of-domain database. We show that the proposed method outperforms both the out-of-domain system and the baseline approaches, i.e. RemixIT and noisy-target training, and also combines well with the previously proposed RemixIT method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360251/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360251/</guid>
</item>
<item>
<title>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</title>
<description><p><span><big>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</big></span><br></p><p><span><small><i>Alexander Karataev; Christian Forsch; Laura Cottatellucci</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3348343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We consider a cell-free massive multiple-input multiple-output (CF-MaMIMO) communication system in the uplink transmission and propose a novel algorithm for blind or semi-blind joint channel estimation and data detection (JCD). We formulate the problem in the framework of bilinear inference and develop a solution based on the expectation propagation (EP) method for both channel estimation and data detection. We propose a new approximation of the joint a posteriori distribution of the channel and data whose representation as a factor graph enables the application of the EP approach using the message-passing technique, local low-complexity computations at the nodes, and an effective modeling of channel-data interplay. The derived algorithm, called bilinear-EP JCD, allows for a distributed implementation among access points (APs) and the central processing unit (CPU) and has polynomial complexity. Our simulation results show that it outperforms other EP-based state-of-the-art polynomial time algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10378663/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10378663/</guid>
</item>
<item>
<title>Adversarial Representation Learning for Robust Privacy Preservation in Audio</title>
<description><p><span><big>Adversarial Representation Learning for Robust Privacy Preservation in Audio</big></span><br></p><p><span><small><i>Shayan Gharib; Minh Tran; Diep Luong; Konstantinos Drossos; Tuomas Virtanen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>S... |
http://localhost:1200/ieee/journal/4427201/earlyaccess - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Sensors Journal</title>
<link>https://ieeexplore.ieee.org/xpl/tocresult.jsp?isnumber=4427201</link>
<atom:link href="http://localhost:1200/ieee/journal/4427201/earlyaccess" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Sensors Journal - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Sat, 04 May 2024 13:09:10 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>The Impact of GIMMF Pitch on Sensitivity of Curvature Sensor Based on Anti-resonance</title>
<description><p><span><big>The Impact of GIMMF Pitch on Sensitivity of Curvature Sensor Based on Anti-resonance</big></span><br></p><p><span><small><i>Yaxun Zhang; Yuxin Liu; Zhiliang Huang; Xiaouyun Tang; Lei Jin; Zhihai Liu; Yu Zhang; Libo Yuan</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3386720</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>We propose and demonstrate a high sensitivity curvature sensor based on the cascaded structure of single mode fiber-graded index multimode fiber-hollow core fiber-single mode fiber, and study the influence of the incident light angle on the sensitivity of anti-resonance curvature sensor. The angle of light entering the hollow core fiber can be modulated by the graded index multimode fiber. Different angles affect the intensity of the spectrum produced by the anti-resonance mechanism. The intensity of the resonance peak generated by the anti-resonance mechanism is sensitive to curvature. The sensor achieves a large dynamic range in the range of 0-9.21 m
-1
and high sensitivity curvature measurement. The maximum sensitivity of the curvature sensor is -5.00 dB/m
-1
. And there is hardly any temperature crosstalk sensitivity. The proposed curvature sensor has the advantages of large dynamic range, high sensitivity and low temperature crosstalk.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10500302/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10500302/</guid>
</item>
<item>
<title>Multi-sensor multiple extended objects tracking based on the message passing</title>
<description><p><span><big>Multi-sensor multiple extended objects tracking based on the message passing</big></span><br></p><p><span><small><i>Yuansheng Li; Tao Shen; Lin Gao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3384560</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Multiple extended objects (EOs) tracking has attracted a lot of attention due to the fast development of high-resolution sensors. A remarkable feature of EO, compared to the traditional point target, is that an EO normally produces more than one measurement, resulting in challenges for finding the associations among objects and measurements. In this paper, we are interested in tracking an unknown number of EOs with multiple sensors by resorting to the message passing method. The existence probability and belief of each EO are explicitly estimated, where the belief is modeled by a mixture gamma Gaussian inverse Wishart (GGIW) distribution, so as to jointly estimate the measurement rate (MR), centroid state and extension. The marginal posterior of EOs is approximately by belief, which is obtained by running the loopy sum-product algorithm (LSPA) on a suitably devised factor graph. As a result, the computational load of the proposed algorithm increases linearly with respect to the number of targets, thus admitting the scalability. Simulation experiments are carried out to verify the performance of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10495766/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10495766/</guid>
</item>
<item>
<title>An Intestine based Bio-Compatible Humidity Sensor for Environmental and Medical Measurements</title>
<description><p><span><big>An Intestine based Bio-Compatible Humidity Sensor for Environmental and Medical Measurements</big></span><br></p><p><span><small><i>Emrehan Yavsan; Muhammet Rojhat Kara; Mehmet Akif Erismis</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3367032</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The bio-compatible devices suitable for recycling and bio-degrading can be achieved with organic materials in nature. In this work, a bio-compatible capacitive humidity sensor is presented for reducing the amount of electronic waste and contributing to the sustainability of natural resources and the future. The sensor consists of 3 layers. The first layer is the processed intestine layer of cattle. Bio-compatibility is achieved with this layer. In addition to being a highly absorbing tissue, the intestine has been used for centuries for long-term preservation of meat based food. Correspondingly, the developed sensor is found to be more durable and long-lasting than other natural-material based humidity sensors in the literature. The other layers of the sensor are interdigitated copper electrodes and a 0.2 mm thick thin film strip. Thin film strip increases mechanical strength as well as flexibility. The developed sensor prototype was subjected to various tests in the humidity range of 20%-90%. In these tests, the hysteresis characteristic of the sensor, its response-recovery time, and its long-term stability and short-term step responses were examined. Moreover, as a possible application in medicine, the sensor can be used to detect breathing cycles. The sensor’s response and recovery times were measured as 8.72’ and 4.47’, respectively, possibly attributed to the stabilization of our test setup, while the sensor successfully detected deep, normal and fast breathing. Despite being kept in an uncontrolled environment, the sensor continued to operate consistently for breath measurements after 56 weeks, which is more than a year.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10445282/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10445282/</guid>
</item>
<item>
<title>Deep neural network assisted terahertz metasurface sensors for the detection of lung cancer biomarkers</title>
<description><p><span><big>Deep neural network assisted terahertz metasurface sensors for the detection of lung cancer biomarkers</big></span><br></p><p><span><small><i>Jie Chen; Fangrong Hu; Xiaoya Ma; Mo Yang; Shangjun Lin; An Su</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3384578</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>MicroRNA (miRNA) is closely related to various cancers, and the change in its expression level is closely related to the development and death of tumor cells. Here, we design and manufacture a terahertz (THz) metasurface sensor to realize the concentration detection and category identification of the miRNAs related to lung cancer. We established two spectral classification datasets, one contains 9 concentrations of miRNAs, and the other contains 3 categories of lung cancer-related miRNAs. And then, we used a deep neural network (DNN) algorithm to classify these spectral datasets and obtained a LOD (Limit-of-Detection) of 100 aM for miRNAs. Moreover, this method can identify the categories of miRNAs. Compared with the other five machine learning (ML) algorithms, the proposed neural network framework achieves the best classification results. This work provides a new way for the detection and identification of trace nucleic acid and biomarkers of many cancers.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10495770/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10495770/</guid>
</item>
<item>
<title>A Casting Surface Dataset and Benchmark for Subtle and Confusable Defect Detection in Complex Contexts</title>
<description><p><span><big>A Casting Surface Dataset and Benchmark for Subtle and Confusable Defect Detection in Complex Contexts</big></span><br></p><p><span><small><i>Qishan Wang; Shuyong Gao; Li Xiong; Aili Liang; Kaidong Jiang; Wenqiang Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3387082</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Industrial anomaly detection (IAD) algorithms are essential for implementing automated quality inspection. Dataset diversity serves as the foundation for developing comprehensive detection algorithms. Existing IAD datasets focus on the diversity of objects and defects, overlooking the diversity of domains within the real data. To bridge this gap, this study proposes the Casting Surface Defect Detection (CSDD) dataset, containing 12,647 high-resolution gray images and pixel-precise ground truth labels for all defect samples. Compared to existing datasets, CSDD has the following two characteristics: 1.) The target samples are unaligned and have complex and variable context information. 2.) The defects in the CSDD dataset samples are subtle and confusable by factors such as oil contamination, processing features, and machining marks, illustrating the challenge of detecting real casting defects in an industrial context. Based on this dataset, we observe that current state-of-the-art (SOTA) IAD methods face challenges when there is considerable variation in sample context information. Furthermore, these methods encounter difficulties when abnormal samples are scarce, particularly those samples with subtle and confusable defects. To address this issue, we propose a novel method called Realistic Synthetic Anomalies (RSA), which enhances the model’s capacity to construct a normal sample distribution by generating a large number of realistic synthetic anomalies. Experimental results demonstrate that the model trained to classify synthetic anomalies from normal samples achieves the highest accuracy for casting surface defect detection and significantly improves detection accuracy for subtle and confusable defects. The CSDD dataset and code of RSA are available at https://github.com/18894269590/RSA.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10502267/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10502267/</guid>
</item>
<item>
<title>PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards</title>
<description><p><span><big>PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards</big></span><br></p><p><span><small><i>Elias Arbash; Margret Fuchs; Behnood Rasti; Sandra Lorenz; Pedram Ghamisi; Richard Gloaguen</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3380826</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Addressing the critical theme of recycling electronic waste (E-waste), this contribution is dedicated to developing advanced automated data processing pipelines as a basis for decision-making and process control. Aligning with the broader goals of the circular economy and the United Nations (UN) Sustainable Development Goals (SDG), our work leverages non-invasive analysis methods utilizing RGB and hyperspectral imaging data to provide both quantitative and qualitative insights into the E-waste stream composition for optimizing recycling efficiency. In this paper, we introduce ’PCB-Vision’; a pioneering RGB-hyperspectral printed circuit board (PCB) benchmark dataset, comprising 53 RGB images of high spatial resolution paired with their corresponding high spectral resolution hyperspectral data cubes in the visible and near-infrared (VNIR) range. Grounded in open science principles, our dataset provides a comprehensive resource for researchers through high-quality ground truths, focusing on three primary PCB components: integrated circuits (IC), capacitors, and connectors. We provide extensive statistical investigations on the proposed dataset together with the performance of several state-of-the-art (SOTA) models, including U-Net, Attention UNet, Residual U-Net, LinkNet, and DeepLabv3+. By openly sharing this multiscene benchmark dataset along with the baseline codes, we hope to foster transparent, traceable, and comparable developments of advanced data processing across various scientific communities, including, but not limited to, computer vision and remote sensing. Emphasizing our commitment to supporting a collaborative and inclusive scientific community, all materials, including code, data, ground truth, and masks, will be accessible at https://github.com/hifexplo/PCBVision.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10485259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10485259/</guid>
</item>
<item>
<title>AuNP-decorated textile as chemo resistive sensor for acetone detection</title>
<description><p><span><big>AuNP-decorated textile as chemo resistive sensor for acetone detection</big></span><br></p><p><span><small><i>S. Casalinuovo; D. Caschera; S. Quaranta; D. Caputo</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2023.3348693</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This paper presents the development of chemo resistive sensors for the detection of volatile organic compounds (VOCs). The proposed sensor is based on citrate-functionalized gold nanoparticles (AuNPs) serving as a sensitive layer deposited on cotton fabric. Impedance variations due to VOC/substrate interaction are used as a detection principle. Specifically, this work focuses on acetone detection after exposing the AuNP-decorated cotton to a CH3COCH3 aqueous solution. Such an interaction resulted in a reduction of the total impedance (i.e., magnitude) of the system. This behavior can be ascribed to Van der Waals forces existing between the C=O group and the citrate moieties adsorbed on the gold nanoparticles, which favor charge injection to the substrate. Response to water was also tested for comparison, assuring that the solvent interacts with the sensitive layer by a different adsorption mechanism, not influencing the overall results. Sensor selectivity was also verified by considering ethanol (representative of alcohol group). Indeed, impedance curves reflect the different type of chemical interaction between the analyte and the substrate. In addition, sensor limit of detection for acetone was found to be 1% v/v, in the considered frequency range. Furthermore, sensor performance in terms of reusability was evaluated, showing that the Au-cotton ability in VOCs detection could be restored after about 90 min with a percentage up to 97 % in the frequency of 1Hz. These results can be considered the starting point for the development of portable, sensitive and user-friendly devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10384362/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10384362/</guid>
</item>
<item>
<title>Failure Mechanism Analysis and Experiment of MEMS VRG under High-g Shock</title>
<description><p><span><big>Failure Mechanism Analysis and Experiment of MEMS VRG under High-g Shock</big></span><br></p><p><span><small><i>Jin Wang; Qi Cai; Wenqiang Wei; Rang Cui; Yunbo Shi; Chong Shen; Huiliang Cao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3389744</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This paper investigates the structure response and failure mechanism of a micro-electromechanical system (MEMS) vibrating ring gyroscope (VRG) under high-g shock. A principle for categorizing different types of loading shock according to the VRG’s frequency scales is devised. A dynamic response model of the VRG structure to diverse types of shock based on quasi-static, vibration and elastic wave theories is established. The failure mechanism of the VRG is analyzed for the typical failures of MEMS devices, including pull-in, fracture and delamination. Based on the established dynamic response model, the inertial displacement that leads to pull-in failure, the fracture-sensitive positions of the VRG structure under different shocks, and the physical model of delamination failure under shock are derived. Through the shock experiments, the critical amplitudes of the loading shock that cause the failures of the VRG are obtained, the reliability model of the VRG under high-g shock is established, and the model established by analyzing the failure mechanism of the VRG is verified from the results of the experiments.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10506452/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10506452/</guid>
</item>
<item>
<title>Affinity-based capturing, release, and glycoprofiling of PSA cancer biomarker using miniaturized micropillar-based platform</title>
<description><p><span><big>Affinity-based capturing, release, and glycoprofiling of PSA cancer biomarker using miniaturized micropillar-based platform</big></span><br></p><p><span><small><i>Ahmad Usman; Ali Asghar Eftekhar; Ali Adibi</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3389992</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The work presents affinity-based biomolecular capturing and release employing a PDMS micropillar-based microfluidic chip. The microfluidic chip has been equipped with a micropillar-based capture chamber specifically optimized to provide enhanced surface area for biomolecular capturing without having any fluid and material clogging. A novel multi-step surface functionalization immunoassay protocol to optimize on-the-flow surface functionalization and capturing efficiency of the biomolecules has been developed and demonstrated. The novel immunoassay protocol involved conjugation of the anti-PSA IgG antibodies with a photo-cleavable (PC)-biotin-PEG3-NHS-ester, capturing of free-PSA (f-PSA) cancer biomarker, the capability of multiplexed glycoprofiling for elucidating the PTMs of the f-PSA using the Sambucus Nigra Lectin (SNA) and Maackia Amurensis Lectin II (MAA-II) lectins, and on-demand release of the captured f-PSA biomarkers. The work provides a proof-of-concept demonstration of affinity-based capturing, multiplexing, glycoprofiling, and release of f-PSA biomarkers with a minimum limit of detection of ~10 pg/ml and a limit of quantification of ~20 pg/ml using a low-cost, disposable microfluidic chip.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10506344/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10506344/</guid>
</item>
<item>
<title>Design of MTM-based Multi-band Micro-Biosensor in Terahertz region as perfect absorber for Early-Stage Leukemia Diagnosis with sensitivity 18626373 TH...</title>
<description><p><span><big>Design of MTM-based Multi-band Micro-Biosensor in Terahertz region as perfect absorber for Early-Stage Leukemia Diagnosis with sensitivity 18626373 THz/RIU</big></span><br></p><p><span><small><i>Musa N. Hamza; Mohammad Tariqul Islam</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383522</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this article, a novel highly sensitive biosensor based on perfect metamaterial absorbers is presented. A detailed study is presented for several different models, different types of substrate materials, different types of resonant materials and substrate thicknesses showing very good sensitivity to any changes. The proposed biosensor exhibits high sensitivity to different polarization and incidence angles, ensuring its performance in terms of signal-to-noise ratio. The proposed biosensor was carefully compared with previously designed sensors and biosensors. The proposed biosensor exhibits amazing sensitivity, such as a quality factor of 41.1, a figure of merit (FOM) of 172466416 RIU-1, and a sensitivity (S) of 18626373 THz/RIU. The high sensitivity of the biosensor allows early detection of leukemia, demonstrating significant differences between leukemia and normal blood. Another interesting result of this article is the use of terahertz waves for imaging. Microwave imaging is performed for electric fields, magnetic fields, and power.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10494209/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10494209/</guid>
</item>
<item>
<title>LEPS: A lightweight and effective single-stage detector for pothole segmentation</title>
<description><p><span><big>LEPS: A lightweight and effective single-stage detector for pothole segmentation</big></span><br></p><p><span><small><i>Xiaoning Huang; Jintao Cheng; Qiuchi Xiang; Jun Dong; Jin Wu; Rui Fan; Xiaoyu Tang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2023.3330335</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Currently, the problem of potholes on urban roads is becoming increasingly severe. Identifying and locating road potholes promptly has become a major challenge for urban construction. Therefore, we proposed a lightweight and effective instance segmentation method called LEPS (Lightweight and Effective Pothole Segmentation Detector) for road pothole detection. To extract the image edge information and gradient information of potholes from the feature map more efficiently, we proposed a module that performs the convolutional super-position supplemented with a convolutional kernel to enhance spatial details for the backbone. We have designed a novel module applied to the Neck layer, improving the detection performance while reducing the parameters. To enable accurate segmentation of fine-grained features, we optimized the ProtoNet, which enables the segmentation head to generate high-quality masks for more accurate prediction. We have fully demonstrated the effectiveness of the method through a large number of comparative experiments. Our detector has excellent performance on two authoritative example datasets, URTIRI and COCO, and can successfully be applied to visual sensors to accurately detect and segment road potholes in real environments, accurately LEPS reached 0.892 and 0.648 in terms of Mask for AP50 and AP50:95, which is improved by 4.6% and 20.6% compared to the original model. These results demonstrate its strong competitiveness when compared to other models. Comprehensively, LEPS improves detection accuracy while maintaining lightweight, which allows the model to meet the practical application requirements of edge computing devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10315064/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10315064/</guid>
</item>
<item>
<title>Design of Ag@ZnO Microreactor with High Sensitivity and Selectivity for Triethylamine Detection</title>
<description><p><span><big>Design of Ag@ZnO Microreactor with High Sensitivity and Selectivity for Triethylamine Detection</big></span><br></p><p><span><small><i>Feifan Fan; Junyang Hao; Yuefeng Gu; Fangfang Xue; Zhicheng Zhu; Tiancheng Wu; Qiuhong Li</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3386602</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Triethylamine (TEA) is a crucial raw material in industrial processes. However, exposure to this gas poses significant health risks to humans. Therefore, the development of sensors that can effectively detect TEA with high performance is crucial. In this study, we synthesized Ag@ZnO hollow spheres using microwave hydrothermal and template methods. The Ag@ZnO microreactor structure was formed by decorating Ag particles on the inner wall of ZnO hollow spheres. This unique structure effectively extended the contact time between Ag particles and gas molecules, and noteworthily enhanced the sensing properties. The gas sensing measurements demonstrated that Ag@ZnO sensor exhibited ultra-high selectivity and superior sensitivity to TEA. The response value to 100 ppm triethylamine was 763 at the optimal temperature of 200 °C, which was approximately 34 times higher than pure ZnO hollow spheres, and 3 times higher than that of ZnO@Ag where Ag particles were loaded on the outer wall of ZnO hollow spheres.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10499990/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10499990/</guid>
</item>
<item>
<title>UMHE: Unsupervised Multispectral Homography Estimation</title>
<description><p><span><big>UMHE: Unsupervised Multispectral Homography Estimation</big></span><br></p><p><span><small><i>Jeongmin Shin; Jiwon Kim; Seokjun Kwon; Namil Kim; Soonmin Hwang; Yukyung Choi</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383453</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Multispectral image alignment plays a crucial role in exploiting complementary information between different spectral images. Homography-based image alignment can be a practical solution considering a tradeoff between runtime and accuracy. Existing methods, however, have difficulty with multispectral images due to the additional spectral gap or require expensive human labels to train models. To solve these problems, this paper presents a comprehensive study on multispectral homography estimation in an unsupervised learning manner. We propose a curriculum data augmentation, an effective solution for models learning spectrum-agnostic representation by providing diverse input pairs. We also propose to use the phase congruency loss that explicitly calculates the reconstruction between images based on low-level structural information in the frequency domain. To encourage multispectral alignment research, we introduce a novel FLIR corresponding dataset that has manually labeled local correspondences between multispectral images. Our model achieves state-of-the-art alignment performance on the proposed FLIR correspondence dataset among supervised and unsupervised methods while running at
151 FPS
. Furthermore, our model shows good generalization ability on the M3FD dataset without finetuning.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10494213/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10494213/</guid>
</item>
<item>
<title>Characterization of Bubble Size of Gas-Liquid Two-Phase Flow Using Ultrasonic Signals</title>
<description><p><span><big>Characterization of Bubble Size of Gas-Liquid Two-Phase Flow Using Ultrasonic Signals</big></span><br></p><p><span><small><i>Ningde Jin; Huiwen Lu; Jiachen Zhang; Weikai Ren</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3388012</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Refining the bubble size distribution characteristics in gas-liquid two-phase flow can enhance the precision of detecting flow parameters. In this study, gas-liquid two-phase bubble flow experiments were first carried out in a vertical upward pipeline with an internal diameter of 20 mm. A high-speed camera was used to take instantaneous snapshots of the bubbles flowing through the fluid sampler, capturing images of the size distribution of the dispersed bubbles within the fluid sampler at different flow conditions. Ultrasonic attenuation information when the bubbles flowing through the measured area was obtained using the pulsed transmission ultrasonic sensor. The magnitude and sign fractal scaling exponents of ultrasonic attenuation signals were obtained by the detrended fluctuation analysis (DFA) method, which exhibits a strong correlation with the bubble diameter. The results show that the nonlinear dynamic characteristics of bubble size evolution in gas-liquid two-phase bubble flow can be effectively characterized by ultrasonic measurement signals, which provides a crucial foundation for different bubble size of void fraction prediction model of gas-liquid two-phase flow.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10505146/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10505146/</guid>
</item>
<item>
<title>A Stretchable and Wearable Ultrasonic Transducer Array for Bladder Volume Monitoring Application</title>
<description><p><span><big>A Stretchable and Wearable Ultrasonic Transducer Array for Bladder Volume Monitoring Application</big></span><br></p><p><span><small><i>Cong Pu; Ben Fu; Lehang Guo; Huixiong Xu; Chang Peng</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3382244</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Monitoring bladder volume is an essential application for patients with voiding dysfunction. Current wearable ultrasonic bladder monitors are rigid, bulky and cannot conform to human skin. This study presents a stretchable and wearable ultrasonic transducer array that can both conform to non-planar skin surfaces and continuously monitor bladder volume. The wearable transducer array mainly consists of 4 × 4 piezoelectric transducer elements, stretchable serpentine-based electrodes for electrical interconnection, electrically conductive adhesive as both matching and backing layers, and silicone elastomer for encapsulation. The transducer array has a center frequency of 3.9 MHz, -6 dB acoustic bandwidth of 57.6%, and up to 40% elastic stretchability. The volume of balloon-bladder phantom was estimated by a least square ellipsoid fitting method. With the true volume of balloon-bladder phantoms ranging from 10 mL to 405 mL, the proposed transducer array has a mean absolute percentage error of 9.4%, a mean absolute error of 11.9 mL and a coefficient of determination of 0.975, which demonstrates the capability of the proposed stretchable and wearable ultrasonic transducer array for bladder volume monitoring application.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10489841/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10489841/</guid>
</item>
<item>
<title>ASMGCN: Attention-Based Semantic-Guided Multi-Stream Graph Convolution Network for Skeleton Action Recognition</title>
<description><p><span><big>ASMGCN: Attention-Based Semantic-Guided Multi-Stream Graph Convolution Network for Skeleton Action Recognition</big></span><br></p><p><span><small><i>Moyan Zhang; Zhenzhen Quan; Wei Wang; Zhe Chen; Xiaoshan Guo; Yujun Li</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3388154</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In recent years, the field of action recognition using spatio-temporal graph convolution models for human skeletal data has made significant progress. However, current methodologies tend to prioritize spatial graph convolution, which leads to an underutilization of valuable information present in skeletal data. It limits the model’s ability to effectively capture complex data patterns, especially in time series data, ultimately impacting recognition accuracy significantly. To address the above issues, this paper introduces an Attention-based Semantic-guided Multi-stream Graph Convolution Network (ASMGCN), which can extract the deep features in skeletal data more fully. Specifically, ASMGCN incorporates a novel temporal convolutional module featuring an attention mechanism and a multiscale residual network, which can dynamically adjust the weights between skeleton graphs at different time points, enabling better capture of relational features. In addition, semantic information is introduced into the loss function, enhancing the model’s ability to distinguish similar actions. Furthermore, the coordinate information of different joints within the same frame is explored to generate new relative position features known as centripetal and centrifugal streams based on the center of gravity. These features are integrated with the original position and motion features of skeleton, including joints and bones, enriching the inputs to the GCN network. Experimental results on the NW-UCLA, NTU RGB+D (NTU60) and NTU RGB+D 120 (NTU120) datasets demonstrate that ASMGCN outperforms other state-of-the-art (SOTA) Human Action Recognition (HAR) methods, signifying its potential in advancing the field of action recognition using skeletal data.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10505150/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10505150/</guid>
</item>
<item>
<title>Partial discharge positioning method in air insulated substation with vehicle mounted UHF sensor array based on RSSI and regularization</title>
<description><p><span><big>Partial discharge positioning method in air insulated substation with vehicle mounted UHF sensor array based on RSSI and regularization</big></span><br></p><p><span><small><i>Ruoyu Wang; Baiqiang Yin; Lifen Yuan; Shudong Wang; Chengwei Ding; Xiaodong Lv</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3387665</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Monitoring technology for partial discharges (PD) in Air-Insulated Substations (AIS) has undergone development and practical application. Partial discharge positioning based on Received Signal Strength Indication (RSSI) faces a variety of interference factors, including complex environments, multipath effects of signals, expensive fees and huge size of monitoring equipment. These factors jointly lead to the decline of PD positioning accuracy. In order to solve these problems, this paper proposes a substation patrol vehicle positioning technology based on RSSI and vehicle-mounted UHF sensor array. Initially, at a specific distance, multiple measurements, Gaussian filtering, and averaging are performed on a specific PD signal, to determine the RSSI at this distance, and consequently identify the parameters within the Shadowing model. Subsequently, a centralized and balanced preprocessing of the positioning equation is to be performed. Then, derive regularization parameters through the L-curve method and employ the Tikhonov regularization technique to determine the coordinates of the PD source. Simulations and experiments indicate that under simulated disturbances of 0-15% within a space of 30m × 30m × 20m, the proposed method maintains a positioning error within 3.6m. The method is cost-effective and operationally straightforward.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10504736/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10504736/</guid>
</item>
<item>
<title>Finger-tapping Motion Recognition Based on Skin Surface Deformation Using Wrist-mounted Piezoelectric Film Sensors</title>
<description><p><span><big>Finger-tapping Motion Recognition Based on Skin Surface Deformation Using Wrist-mounted Piezoelectric Film Sensors</big></span><br></p><p><span><small><i>Shumma Jomyo; Akira Furui; Tatsuhiko Matsumoto; Tomomi Tsunoda; Toshio Tsuji</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3386333</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The miniaturization of computers has led to the development of wearable devices in the form of watches and eyeglasses. Consequently, the narrower screen size has raised the issue of operability for text input. This problem can be resolved using external input devices, such as physical keyboards. However, this can impair portability and accessibility. This study proposes a finger-tapping motion recognition system using wrist-mounted piezoelectric film sensors to realize an input interface with high wearability and not limited by screen size. In the proposed system, biodegradable piezoelectric film sensors, which are highly compatible with biological signal measurement, are attached to the palmar and dorsal surfaces of the wrist to measure minute skin surface deformation during tapping. The system detects the occurrence of tapping movements for each finger by preprocessing the measured signals and calculating the total activity of all channels. It also recognizes the type of finger movement based on machine learning. In the experiment, we measured ten different signals, including five-finger flexion and extension, for eleven subjects, to evaluate the effectiveness of the proposed method. According to the experimental results, tapping recognition accuracy for time-series data was 77.5%, assuming character input. In addition, the time difference between the detected and actual taps was approximately 50 ms on average. Therefore, the proposed method can be utilized as an input interface for wristband-type wearable devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10500304/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10500304/</guid>
</item>
<item>
<title>CLORP: Cross-Layer Opportunistic Routing Protocol for Underwater Sensor Networks Based on Multi-Agent Reinforcement Learning</title>
<description><p><span><big>CLORP: Cross-Layer Opportunistic Routing Protocol for Underwater Sensor Networks Based on Multi-Agent Reinforcement Learning</big></span><br></p><p><span><small><i>Shuai Liu; Jingjing Wang; Wei Shi; Guangjie Han; Shefeng Yan; Jiaheng Li</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383035</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>With the development of the Internet of Underwater Things (IoUT), both academia and industry have significant emphasized underwater wireless sensor networks (UWSNs). To address the issues of slow convergence, high latency, and limited energy in existing intelligent routing protocols in UWSNs, a cross-layer opportunistic routing protocol (CLORP) for underwater sensor networks based on multi-agent reinforcement learning (MARL) is proposed in this paper. First, CLORP combines the decision-making capability of multi-agent reinforcement learning with the idea of opportunistic routing to sequentially select a set of neighbors with larger values as potential forwarding nodes, thereby increasing the packet transmission success rate. Second, in the design of the MARL reward function, two reward functions for successful and unsuccessful packet transmission are designed jointly with cross-layer information to improve the routing protocol’s performance. Finally, two algorithmic optimization strategies, adaptive learning rate and Q-value initialization based on location and number of neighbors, are proposed to facilitate the faster adaptation of agents to the dynamic changes of network topology and accelerate CLORP convergence. The experimental results demonstrate that CLORP can increase algorithm convergence speed by 13.2%, reduce network energy consumption by 25%, and decrease network latency by 31.2%.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10492498/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10492498/</guid>
</item>
<item>
<title>Three-Dimensional Array SAR Sparse Imaging Based on Hybrid Regularization</title>
<description><p><span><big>Three-Dimensional Array SAR Sparse Imaging Based on Hybrid Regularization</big></span><br></p><p><span><small><i>Jing Gao; Yangyang Wang; Jinjie Yao; Xu Zhan; Guohao Sun; Jiansheng Bai</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3386901</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>With the development and maturity of compressed sensing theories, sparse signal processing has been widely applied to synthetic aperture radar (SAR) imaging. L
1
regularization, as an effective sparse reconstruction model, can enhance sparsity of the scene. However, due to the convexity of the L
1
regularization function, the L
1
-based sparse reconstruction frequently introduces bias estimation, resulting in an underestimation of target amplitudes. Furthermore, the 3-D SAR imaging scene has diverse features, which are difficult to characterize solely with the L
1
regularization. Therefore, in this paper, we propose a novel imaging framework for 3-D array SAR imaging, which combines hybrid regularization function and an improved variable splitting with the method of multipliers (IVSMM). Firstly, we present the hybrid regularization function, which combines smoothly clipped absolute deviation (SCAD) and high-dimensional total variation (HTV) to reduce bias effects and preserve the regional features of the target. Then, we present IVSMM to solve the optimization problem of the hybrid regularization, effectively reducing the computational complexity required for 3-D imaging. Finally, the excellent reconstruction performance of the proposed method is validated through a series of simulations and experimental data.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10502305/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10502305/</guid>
</item>
<item>
<title>Water Quality Assessment Tool for On-site Water Quality Monitoring</title>
<description><p><span><big>Water Quality Assessment Tool for On-site Water Quality Monitoring</big></span><br></p><p><span><small><i>Segun O. Olatinwo; Trudi H. Joubert; Damilola D. Olatinwo</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383887</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Reliable water quality monitoring requires on-site processing and assessment of water quality data in near real-time. This helps to promptly detect changes in water quality, prevent biodiversity loss, safeguard the health and well-being of communities, and mitigate agricultural problems. To this end, we proposed a Highway-Bidirectional Long Short-term Memory (Highway-BiLSTM)-based water quality classification tool for potential integration into an edge-enabled water quality monitoring system to facilitate on-site water quality classification. The performance of the proposed classifier was validated by comparing it with several baseline water quality classifiers. The proposed classifier outperformed the baseline water classifier in terms of accuracy, precision, sensitivity, F1-score, and confusion matrix. Specifically, the proposed water classifier surpassed the random forest (RF) classifier with 2% accuracy, precision, sensitivity, and F1-score. Moreover, the proposed classifier achieved an increase of 4% in accuracy, precision, sensitivity, and F1-score for classifying water quality compared with the Gradient Boosting classifier. Additionally, the proposed method has 4% increase in accuracy, sensitivity, F1-score, and 3% increase in precision compared to the support vector machine (SVM) water quality classifier. The proposed method outperformed the artificial neural network (ANN) classifier by 1% accuracy, precision, sensitivity, and F1-score. Finally, the proposed method demonstrated rare errors in accurately classifying complex water quality samples. These findings suggest that our proposed method could be used to effectively classify water quality to aid accurate decision making and environmental management.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10499201/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10499201/</guid>
</item>
<item>
<title>An Integrated Self-Powered Wheel-Speed Monitoring System utilizing Piezoelectric-Electromagnetic−Triboelectric Hybrid Generator</title>
<description><p><span><big>An Integrated Self-Powered Wheel-Speed Monitoring System utilizing Piezoelectric-Electromagnetic−Triboelectric Hybrid Generator</big></span><br></p><p><span><small><i>Xiaohui Lu; Zhaojie Zhang; Wentao Ruan; Hengyu Li; Da Zhao; Kuankuan Wang; Bingzhao Gao; Baichuan Leng; Xin Yu; Bangcheng Zhang; Tinghai Cheng</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3384569</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In the intelligent vehicle systems, vehicle speed information is collected and transmitted by wheel speed sensors, which is related to the safety, energy efficiency, and comfort of the vehicle. This work proposes an integrated self-powered wheel-speed monitoring system (SWMS), which can achieve real-time sensing and wireless transmission multifunction by employing an integrated Piezoelectric-Electromagnetic−Triboelectric hybrid generator (PETHG) for harvesting the mechanical energy from rotating wheels. In particular, the triboelectric generation unit in SWMS includes an energy-channel (E-TENG) and a sensing-channel (S-TENG), the E-TENG was used for harvesting the rotational energy of the wheel, and the S-TENG was used for monitoring the rotation speed of the wheel. In experiment test, the output performance of the hybrid generator was assessed using a rotating electrical motor to simulate different wheel speed inputs. At 600 rpm, the instantaneous power of the hybrid generator reached 168.2 mW, which can power the low-power MCU and waveform conversion circuit in the established wheel speed monitoring system. the wheel speed recorded by MCU were compared with actual rotation speed yielding an average error rate of only 1.9 %. This work offers a solution for self-powered sensors for intelligent driving.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10495772/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10495772/</guid>
</item>
<item>
<title>Event-based Estimation of Hand Forces from High-Density Surface EMG on a Parallel Ultra-Low-Power Microcontroller</title>
<description><p><span><big>Event-based Estimation of Hand Forces from High-Density Surface EMG on a Parallel Ultra-Low-Power Microcontroller</big></span><br></p><p><span><small><i>Marcello Zanghieri; Pierangelo Maria Rapa; Mattia Orlandi; Elisa Donati; Luca Benini; Simone Benatti</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3359917</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Modeling hand kinematics and dynamics is a key goal for research on Human-Machine Interfaces, with surface electromyography (sEMG) being the most commonly used sensing modality. Though under-researched, sEMG regression-based modeling of hand movements and forces is promising for finer control than allowed by mapping to fixed gestures. We present an event-based sEMG encoding for multi-finger force estimation implemented on a microcontroller unit (MCU). We are the first to target the HYSER High-Density (HD)-sEMG dataset in multi-day conditions closest to a real scenario without a fixed force pattern. Our Mean Absolute Error of (8.4 ± 2.8)% of the Maximum Voluntary Contraction (MVC) is on par with State-of-the-Art (SoA) works on easier settings such as within-day, single-finger, or fixed-exercise. We deploy our solution for HYSER’s hardest task on a parallel ultra-low power MCU, getting an energy consumption below 6.5 uJ per sample, 2.8× to 11× more energy-efficient than SoA single-core solutions, and a latency below 280 us per sample, shorter than HYSER’s HD-sEMG sampling period, thus compatible with real-time operation on embedded devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10422755/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10422755/</guid>
</item>
<item>
<title>Ge-on-Si avalanche photodiodes with photon trapping nanostructures for sensing and optical quantum applications</title>
<description><p><span><big>Ge-on-Si avalanche photodiodes with photon trapping nanostructures for sensing and optical quantum applications</big></span><br></p><p><span><small><i>Shaoteng Wu; Hao Zhou; Li He; Zhaozhen Wang; Qimiao Chen; Lin Zhang; Chuan Seng Tan</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2023.3271215</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>High-sensitivity Ge/Si avalanche photodiodes (APDs) have recently gained attention for their application in sensing and optical communication due to their low cost and CMOS compatible process. However, compared to commercial III–V compound APDs, Ge/Si APDs usually suffer from the issue of relatively low primary responsivity. In this paper, we report Ge-on-Si separate absorption, charge, and multiplication avalanche photodiodes (SACM-APDs) with photon-trapping nanostructures to enhance light absorption. Besides, by optimizing the depth of the holes, the photon trapping structure could reduce the dark current without compromising the avalanche effect as confirmed by both simulations and experimental results. As a result, the responsivity of the photon trapping APDs increases by 20-50 % from that of control APDs at the 1,550 nm wavelength band. Furthermore, a quantum efficiency higher than 80% could be achieved at 1550 nm when the photon trapping Ge-on-Si APD is on Si-on-insulator (SOI) platforms as predicted by simulations. Our results demonstrate that the photon trapping Ge/Si APDs exhibit superior dark current, light absorption and gain than those of the control devices, which have the potential applications in sensing and optical quantum communications.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10114637/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10114637/</guid>
</item>
<item>
<title>CFBG-based linear displacement sensor enabled by MXene’s photothermal effect</title>
<description><p><span><big>CFBG-based linear displacement sensor enabled by MXene’s photothermal effect</big></span><br></p><p><span><small><i>Xiaokang Li; Binchuan Sun; Ting Xue; Kangzhe Zhao; Xinrong Li; Dong Li; Yajun Jiang; Bobo Du; Dexing Yang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3392266</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Accurate measuring of mechanical displacement is essential for most technologies. Here, we introduce a novel chirped fiber Bragg grating (CFBG) displacement sensor that enables ultra-sensitive measurement of a transverse displacement by mapping it into the phase-shift distribution of the CFBG. The displacement sensor is implemented by introducing a nonpermanent phase-shift on the CFBG with the assistance of MXene’s excellent photothermal effect. The phase-shift distribution can be used as a marker on the scale, and a high-resolution spectrometer observes the phase-shift markers so that a displacement sensor can be realized. The results of the proof-of-concept experiments show that the sensitivity of the sensor can reach 0.88 nm/mm with a linearity coefficient of 0.998 over a 10 mm range, and the displacement resolution of the sensor is about 12 μm. Our scheme is intrinsically stable and could be implemented as a compact sensor, using a cost-effective sensor design. This sensor can offer aerospace, biomedical, and machining applications, especially in harsh and closed environments with strong electromagnetic interference.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10510241/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10510241/</guid>
</item>
<item>
<title>Reflective Markers Assisted Indoor LiDAR Self-Positioning Algorithm based on Joint Optimization with FIM and GDOP</title>
<description><p><span><big>Reflective Markers Assisted Indoor LiDAR Self-Positioning Algorithm based on Joint Optimization with FIM and GDOP</big></span><br></p><p><span><small><i>Liangbo Xie; Jiahao Hu; Nan Du; Mu Zhou; Yong Wang; Wei Nie</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2023.3294934</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Simultaneous Localization and Mapping (SLAM) is one of the key technologies in autonomous driving and has been wildly used in mobile robot exploration and unmanned aerial vehicle navigation. LiDAR (Light Detection and Ranging) SLAM depends on matching the data perceived by LiDAR, finding the best match for consistency to determine the LiDAR’s pose, and completing the map construction. However, most existing matching methods are inefficient. Moreover, each processing is based on the previous estimation result, estimation errors accumulate over time, resulting in cumulative drift error. To address issues of low efficiency in extracting key feature points, low self-positioning accuracy and reliability, this paper proposes a LiDAR self-positioning method based on reflective markers (RMs). Firstly, the relationship between the positioning accuracy and the RMs’ coordinates are calculated through Fisher information matrix (FIM). Then, Geometric Dilution of Precision (GDOP) is introduced to comprehensively analyze impacts of the geometric positions and quantity of the RMs on the positioning lower error bound of the LiDAR. The layout rules of RMs are designed according to the vertical angle resolution of the LiDAR resolution. We leverage the high reflectivity of RMs to calculate the LiDAR’s global position. By solving the distance, horizontal angle, and vertical angle information obtained from multiple RMs, high-precision self-positioning is achieved during the LiDAR’s motion. Experimental results show that, compared to the LOAM and LeGO-LOAM algorithms, the LiDAR self-positioning accuracy can be improved by approximately 50% in indoor environments, especially in indoor scenarios involving moving up and down stairs.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10186281/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10186281/</guid>
</item>
<item>
<title>Ultrahigh-sensitivity fiber-optic immunosensor based on the high-order-harmonic Vernier effect for cytokeratin 19 fragment detection</title>
<description><p><span><big>Ultrahigh-sensitivity fiber-optic immunosensor based on the high-order-harmonic Vernier effect for cytokeratin 19 fragment detection</big></span><br></p><p><span><small><i>Haiming Qiu; Jiajun Tian; Ying Gao; Cheng Zhou; Yong Yao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3385361</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Herein, we propose and demonstrate a fiber-optic immunosensor based on the high-order-harmonic Vernier effect (HVE) and cascaded dual Fabry–Pérot interferometers (FPIs) for detecting the cytokeratin 19 fragment 21-1 (CYFRA 21-1). The proposed sensor was fabricated by splicing a single-mode fiber (SMF), a hollow-core photonic crystal fiber (HCPCF), and an SMF pigtail. The high-order HVE occurs between the HCPCF (reference FPI) and SMF segment (sensing FPI). To enhance the interaction between the sensing FPI and external medium, graphene oxide with a large specific surface area and abundant oxygen-containing functional groups is deposited on the SMF pigtail end face. Subsequently, the SMF pigtail end surface is immobilized with the anti-CYFRA 21-1 antibody to detect CYFRA 21-1. Antigen–antibody binding changes the sensing layer thickness and roughness, resulting in changes in the reflectance spectrum. HVE can substantially improve the refractive-index sensitivity of a sensor. In our experiments, we obtained a high sensitivity of 162,000 nm/RIU, which is two-orders-of-magnitude higher than that obtained without HVE (1,270 nm/RIU). Further, the special structure of the proposed immunosensor results in a low noise level (±0.07 nm). On this basis, the proposed immunosensor was used for CYFRA 21-1 detection, and a limit of detection of 1.6 fg/mL was achieved. Thus, the proposed immunosensor has the advantages of ultrahigh sensitivity, high fabrication tolerance, simple structure, strong stability, and good robustness. To the best of our knowledge, this study is the first to apply HVE to fiber-optic immunosensors.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10502344/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10502344/</guid>
</item>
<item>
<title>A Multi-matrix E-nose with Optimal Multi-ranged AFE Circuit for Human Volatilome Fingerprinting</title>
<description><p><span><big>A Multi-matrix E-nose with Optimal Multi-ranged AFE Circuit for Human Volatilome Fingerprinting</big></span><br></p><p><span><small><i>Antonio Vincenzo Radogna; Giuseppe Grassi; Stefano D’Amico; Pietro Aleardo Siciliano; Angiola Forleo; Simonetta Capone</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2023.3343762</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Since hundreds of volatile organic compounds (VOCs) produced by cell metabolism and released into the blood are excreted through exhaled breath or body fluids, the volatile composition (volatilome) of human samples reflects a subject’s state of health and early signals any abnormal deviation from healthy to disease. The chemical volatilomic profile of biological matrices can be transduced in a digital fingerprint by low cost and easy-to-use electronic nose (e-nose) devices based on gas sensor arrays. The e-noses can be used to aid clinical diagnosis supporting conventional diagnostic methods that sometimes require expensive or invasive medical procedures and delays in diagnoses. In this paper, an e-nose devoted to the human volatilome fingerprinting is presented. The device, code-named SPYROX, adopts an array of 8 metal-oxide (MOX) gas sensors and it is able to analyze response signals from different matrices (multi-matrix samples), dealing with exhaled... |
http://localhost:1200/ieee/author/37281702200 - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>Ana Garcia Armada on IEEE Xplore</title>
<link>https://ieeexplore.ieee.org/author/37281702200</link>
<atom:link href="http://localhost:1200/ieee/author/37281702200" rel="self" type="application/rss+xml"></atom:link>
<description>Ana Garcia Armada (Fellow, IEEE) is a Professor at Universidad Carlos III de Madrid (UC3M), Spain. She has been a Visiting Scholar at Stanford University, Bell Labs, and University of Southampton. She has published more than 250 papers in conferences and journals. She holds five patents. Her research mainly focuses on signal processing applied to wireless communications. She has been a member of the organizing committee of several conferences, including IEEE Globecom 2021 as the General Chair. She has received several awards from UC3M, the third place Bell Labs Prize 2014, the Outstanding Service Award from the IEEE ComSoc Signal Processing and Communications Electronics Technical Committee and the Outstanding Service Award from the IEEE ComSoc Women in Communications Engineering Standing Committee. She has received the IEEE ComSoc/KICS Exemplary Global Service Award in 2022. She serves on the editorial boards for IEEE Transactions on Communications, IEEE Open Journal of the Communications Society, and ITU Journal on Future and Evolving Technologies. - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Sat, 04 May 2024 13:09:11 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Understanding the effects of phase noise in orthogonal frequency division multiplexing (OFDM)</title>
<description><p><span><big>Understanding the effects of phase noise in orthogonal frequency division multiplexing (OFDM)</big></span><br></p><p><span><small><i>A. Garcia Armada</i></small></span><br><span><small><i>https://doi.org/10.1109/11.948268</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>Phase noise must be carefully considered when designing an OFDM-based communication system since an accurate prediction of the tolerable phase noise can allow the system and RF engineers to relax specifications. This paper analyzes the performance of OFDM systems under phase noise and its dependence on the number of sub-carriers both in the presence and absence of a phase correction mechanism. Besides some practical results are provided so as to give some insight into the phase noise spectral specifications that should be required of the local oscillator.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/948268/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/948268/</guid>
</item>
<item>
<title>Phase noise and sub-carrier spacing effects on the performance of an OFDM communication system</title>
<description><p><span><big>Phase noise and sub-carrier spacing effects on the performance of an OFDM communication system</big></span><br></p><p><span><small><i>A.G. Armada; M. Calvo</i></small></span><br><span><small><i>https://doi.org/10.1109/4234.658613</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>This letter analyzes the phase noise effects on an orthogonal frequency division multiplexing (OFDM) signal and its dependence with the sub-carrier spacing. Pilot-based channel estimation, which has been suggested as a means of combating the channel effects, can also correct the phase noise effects under some circumstances, which are investigated.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/658613/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/658613/</guid>
</item>
<item>
<title>Fair Design of Plug-in Electric Vehicles Aggregator for V2G Regulation</title>
<description><p><span><big>Fair Design of Plug-in Electric Vehicles Aggregator for V2G Regulation</big></span><br></p><p><span><small><i>José Joaquín Escudero-Garzas; Ana Garcia-Armada; Gonzalo Seco-Granados</i></small></span><br><span><small><i>https://doi.org/10.1109/TVT.2012.2212218</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>Plug-in electric vehicles (PEVs) have recently attracted much attention due to their potential to reduce
$\hbox{CO}_{2}$
emissions and transportation costs and can be grouped into entities (aggregators) to provide ancillary services such as frequency regulation. In this paper, the application of aggregators to frequency regulation by making fair use of their energy storage capacity is addressed. When the power grid requires frequency regulation service to the aggregator to adjust the grid frequency, the PEVs participating in providing the service can either draw energy (as it is usually done to charge the vehicle) or deliver energy to the grid by means of the vehicle-to-grid (V2G) interface. Under the general framework of optimizing the aggregator profit, different methods, such as state-dependent allocation and the water-filling approach, are proposed to achieve a final state of charge (SOC) of the PEVs that satisfy the desired fairness criteria once the regulation service has been carried out.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/6262497/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/6262497/</guid>
</item>
<item>
<title>Task Scheduling for Mobile Edge Computing Using Genetic Algorithm and Conflict Graphs</title>
<description><p><span><big>Task Scheduling for Mobile Edge Computing Using Genetic Algorithm and Conflict Graphs</big></span><br></p><p><span><small><i>Ahmed A. Al-Habob; Octavia A. Dobre; Ana García Armada; Sami Muhaidat</i></small></span><br><span><small><i>https://doi.org/10.1109/TVT.2020.2995146</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>In this paper, we consider parallel and sequential task offloading to multiple mobile edge computing servers. The task consists of a set of inter-dependent sub-tasks, which are scheduled to servers to minimize both offloading latency and failure probability. Two algorithms are proposed to solve the scheduling problem, which are based on genetic algorithm and conflict graph models, respectively. Simulation results show that these algorithms provide performance close to the optimal solution, which is obtained through exhaustive search. Furthermore, although parallel offloading uses orthogonal channels, results demonstrate that the sequential offloading yields a reduced offloading failure probability when compared to the parallel offloading. On the other hand, parallel offloading provides less latency. However, as the dependency among sub-tasks increases, the latency gap between parallel and sequential schemes decreases.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/9094341/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/9094341/</guid>
</item>
<item>
<title>SNR gap approximation for M-PSK-Based bit loading</title>
<description><p><span><big>SNR gap approximation for M-PSK-Based bit loading</big></span><br></p><p><span><small><i>Ana Garcia-Armada</i></small></span><br><span><small><i>https://doi.org/10.1109/TWC.2006.1576527</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>Adaptive OFDM has the potential of providing bandwidth-efficient communications in hostile propagation environments. Currently, bit loading algorithms use M-ary quadrature amplitude modulation of the OFDM sub-carriers, where the number of bits per symbol modulating each of them is obtained in order to maximize the performance. SNR gap approximation for M-QAM signaling makes the algorithms simpler to implement. However, in some circumstances it may be preferable to use. M-ary phase shift keying. In this letter an approximation is derived for M-PSK similar to the SNR gap of M-QAM so that bit loading algorithms can be extended to this type of modulation. In addition, the performance obtained when using M-PSK is compared to that of M-QAM in a practical situation.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/1576527/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/1576527/</guid>
</item>
<item>
<title>OFDM performance in amplifier nonlinearity</title>
<description><p><span><big>OFDM performance in amplifier nonlinearity</big></span><br></p><p><span><small><i>S. Merchan; A.G. Armada; J.L. Garcia</i></small></span><br><span><small><i>https://doi.org/10.1109/11.713060</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>The activities of the current European RACE and ACTS projects have led to an increasing interest in OFDM (orthogonal frequency division multiplexing) as a means of combating impulsive noise and multipath effects and making fuller use of the available bandwidth of the system. This paper analyses the performance of OFDM signals in amplifier nonlinearity. In particular, bit error rate (BER) degradation as a result of amplitude limiting or clipping are analysed. In the presence of both nonlinear distortion and additive Gaussian noise, optimized output power back off is provided to balance the requirements of minimum BER and power amplifier efficiency. For this purpose, an OFDM system has been built using the SPW (Signal Processing Worksystem) simulator.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/713060/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/713060/</guid>
</item>
<item>
<title>Design and implementation of synchronization and AGC for OFDM-based WLAN receivers</title>
<description><p><span><big>Design and implementation of synchronization and AGC for OFDM-based WLAN receivers</big></span><br></p><p><span><small><i>V.P.G. Jimenez; M.J.F.-G. Garcia; F.J.G. Serrano; A.G. Armada</i></small></span><br><span><small><i>https://doi.org/10.1109/TCE.2004.1362493</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>An efficient implementation of several tasks at the receiver becomes crucial in OFDM-based high-speed WLAN systems, such as automatic gain control, time and frequency synchronization, and offset tracking. This paper deals with fixed point constraints and accuracy requirements for implementation of those algorithms. Also, a complete set of thresholds for the practical implementation of time and frequency synchronization sub-blocks is obtained. Moreover, a technique to mitigate the remaining frequency offset after coarse acquisition is proposed, yielding a good trade-off between performance and complexity. Finally, we propose the implementation of a simple and effective automatic gain control procedure.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/1362493/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/1362493/</guid>
</item>
<item>
<title>New Technologies and Trends for Next Generation Mobile Broadcasting Services</title>
<description><p><span><big>New Technologies and Trends for Next Generation Mobile Broadcasting Services</big></span><br></p><p><span><small><i>Alejandro de la Fuente; Raquel Perez Leal; Ana Garcia Armada</i></small></span><br><span><small><i>https://doi.org/10.1109/MCOM.2016.1600216RP</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>It is expected that by the year 2020, video services will account for more than 70 percent of mobile traffic. It is worth noting that broadcasting is a mechanism that efficiently delivers the same content to many users, not only focusing on venue casting, but also distributing many other media such as software updates and breaking news. Although broadcasting services are available in LTE and LTE-A networks, new improvements are needed in some areas to handle the demands expected in the near future. In this article we review the actual situation and some of the techniques that will make the broadcast service more dynamic and scalable, meeting the demands of its evolution toward the next generation. Resource allocation techniques for broadcast/multicast services, integration with new waveforms in 5th generation mobile communications (5G), initiatives for spectrum sharing and aggregation, or the deployment of small cells placed together with the existing macro cells, are some enhancements that are examined in detail, providing directions for further development. With this evolution, 5G broadcasting will be a driver to achieve the spectral efficiency needed for the 1000 times traffic growth that is expected in upcoming years, leading to new applications in 5G networks that are specifically focused on mobile video services.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/7593432/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/7593432/</guid>
</item>
<item>
<title>VLC-Based Networking: Feasibility and Challenges</title>
<description><p><span><big>VLC-Based Networking: Feasibility and Challenges</big></span><br></p><p><span><small><i>Alain R. Ndjiongue; Telex M. N. Ngatched; Octavia A. Dobre; Ana G. Armada</i></small></span><br><span><small><i>https://doi.org/10.1109/MNET.001.1900428</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>VLC has emerged as a prominent technology to address the radio spectrum shortage. It is characterized by the unlicensed and unexploited high bandwidth, and provides the system with cost-effective advantages because of the dual-use of light bulbs for illumination and communication and the low complexity design. It is considered to be utilized in various telecommunication systems, including 5G, and represents the key technology for light-fidelity. To this end, VLC has to be integrated into the existing telecommunication networks. Therefore, its analysis as a network technology is momentous. In this article, we consider the feasibility of using VLC as a network technology and discuss the challenges related to the implementation of a VLC-based network, as well as the integration of VLC into existing conventional networks and its inclusion in standards.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/8970387/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/8970387/</guid>
</item>
<item>
<title>Analysis of RIS-Based Terrestrial-FSO Link Over G-G Turbulence With Distance and Jitter Ratios</title>
<description><p><span><big>Analysis of RIS-Based Terrestrial-FSO Link Over G-G Turbulence With Distance and Jitter Ratios</big></span><br></p><p><span><small><i>Alain R. Ndjiongue; Telex. M. N. Ngatched; Octavia A. Dobre; Ana Garcia Armada; Harald Haas</i></small></span><br><span><small><i>https://doi.org/10.1109/JLT.2021.3108532</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>One of the main problems faced by communication systems is the presence of skip-zones in the targeted areas. With the deployment of the fifth-generation mobile network, solutions are proposed to solve the signal loss due to obstruction by buildings, mountains, and atmospheric or weather conditions. Among these solutions, reconfigurable intelligent surfaces (RIS), which are newly proposed modules, may be exploited to reflect the incident signal in the direction of dead zones, increase communication coverage, and make the channel smarter and controllable. This paper tackles the skip-zone problem in terrestrial free-space optical (T-FSO) systems using a single-element RIS. Considering link distances and jitter ratios at the RIS position, we carry out a performance analysis of RIS-aided T-FSO links affected by turbulence and pointing errors, for both heterodyne detection and intensity modulation-direct detection techniques. Turbulence is modeled using the Gamma-Gamma distribution. We analyze the model and provide exact closed-form expressions of the probability density function, cumulative distribution function, and moment generating function of the end-to-end signal-to-noise ratio. Capitalizing on these statistics, we evaluate the system performance through the outage probability, ergodic channel capacity, and average bit error rate for selected binary modulation schemes. Numerical results, validated through simulations, obtained for different RIS positions and link distances ratio values, reveal that RIS-based T-FSO performs better when the RIS module is located near the transmitter.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/9525295/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/9525295/</guid>
</item>
</channel>
</rss> |
Derekmini
changed the title
got to ofetch in /ieee/journal.ts, /ieee/author.ts and /ieee/earlyaccess.ts/ieee/journal.ts, /ieee/author.ts and /ieee/earlyaccess.ts
Derekmini
changed the title
/ieee/journal.ts, /ieee/author.ts and /ieee/earlyaccess.ts
|
Successfully generated as following: http://localhost:1200/ieee/journal/8782710 - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Open Journal of Signal Processing</title>
<link>https://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=8782710</link>
<atom:link href="http://localhost:1200/ieee/journal/8782710" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Open Journal of Signal Processing - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Tue, 14 May 2024 10:45:01 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Synthbuster: Towards Detection of Diffusion Model Generated Images</title>
<description><p><span><big>Synthbuster: Towards Detection of Diffusion Model Generated Images</big></span><br></p><p><span><small><i>Quentin Bammey</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337714</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Synthetically-generated images are getting increasingly popular. Diffusion models have advanced to the stage where even non-experts can generate photo-realistic images from a simple text prompt. They expand creative horizons but also open a Pandora's box of potential disinformation risks. In this context, the present corpus of synthetic image detection techniques, primarily focusing on older generative models like Generative Adversarial Networks, finds itself ill-equipped to deal with this emerging trend. Recognizing this challenge, we introduce a method specifically designed to detect synthetic images produced by diffusion models. Our approach capitalizes on the inherent frequency artefacts left behind during the diffusion process. Spectral analysis is used to highlight the artefacts in the Fourier transform of a residual image, which are used to distinguish real from fake images. The proposed method can detect diffusion-model-generated images even under mild
jpeg
compression, and generalizes relatively well to unknown models. By pioneering this novel approach, we aim to fortify forensic methodologies and ignite further research into the detection of AI-generated images.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334046/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334046/</guid>
</item>
<item>
<title>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</title>
<description><p><span><big>Moving Target At Constant Velocity Localization Using TOA Measurements From Single Moving Receiver With Unknown Signal Period</big></span><br></p><p><span><small><i>Yanbin Zou; Jingna Fan; Zekai Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we consider using time-of-arrival (TOA) measurements from a single moving receiver to locate a moving target at constant velocity that emits a periodic signal with unknown signal period. First, we give the TOA measurement model and deduce the Cram
$\acute{\text{e}}$
r-Rao lower bounds (CRLB). Then, we formulate a nonlinear least squares (NLS) problem to estimate the unknown parameters. We use semidefinite programming (SDP) techniques to relax the nonconvex NLS problem. However, it is shown that the SDP localization algorithm cannot provide a high-quality solution. Subsequently, we develop a fixed point iteration (FPI) method to improve the performance of the SDP algorithm. In addition, we also consider the presence of receiver position errors and develop the corresponding localization algorithm. Numerical simulations are conducted to demonstrate the localization performance of the proposed algorithms by comparing them with the CRLB.
Index Term
-Fixed point iteration (FPI), semidefinite programming (SDP), single moving receiver, target localization, time-of-arrival (TOA).</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336384/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336384/</guid>
</item>
<item>
<title>The Neural-SRP Method for Universal Robust Multi-Source Tracking</title>
<description><p><span><big>The Neural-SRP Method for Universal Robust Multi-Source Tracking</big></span><br></p><p><span><small><i>Eric Grinstein; Christopher M. Hicks; Toon van Waterschoot; Mike Brookes; Patrick A. Naylor</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340057</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Neural networks have achieved state-of-the-art performance on the task of acoustic Direction-of-Arrival (DOA) estimation using microphone arrays. Neural models can be classified as end-to-end or hybrid, each class showing advantages and disadvantages. This work introduces Neural-SRP, an end-to-end neural network architecture for DOA estimation inspired by the classical Steered Response Power (SRP) method, which overcomes limitations of current neural models. We evaluate the architecture on multiple scenarios, namely, multi-source DOA tracking and single-source DOA tracking under the presence of directional and diffuse noise. The experiments demonstrate that our proposed method compares favourably in terms of computational and localization performance with established neural methods on various recorded and simulated benchmark datasets.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345765/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345765/</guid>
</item>
<item>
<title>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</title>
<description><p><span><big>A Neural-Enhanced Factor Graph-Based Algorithm for Robust Positioning in Obstructed LOS Situations</big></span><br></p><p><span><small><i>Alexander Venus; Erik Leitinger; Stefan Tertinek; Klaus Witrisal</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3338113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper presents a neural-enhanced probabilistic model and corresponding factor graph-based sum-product algorithm for robust localization and tracking in multipath-prone environments. The introduced hybrid probabilistic model consists of physics-based and data-driven measurement models capturing the information contained in both, the line-of-sight (LOS) component as well as in multipath components (NLOS components). The physics-based and data-driven models are embedded in a joint Bayesian framework allowing to derive from first principles a factor graph-based algorithm that fuses the information of these models. The proposed algorithm uses radio signal measurements from multiple base stations to robustly estimate the mobile agent's position together with all model parameters. It provides high localization accuracy by exploiting the position-related information of the LOS component via the physics-based model and robustness by exploiting the geometric imprint of multipath components independent of the propagation channel via the data-driven model. In a challenging numerical experiment involving obstructed LOS situations to all anchors, we show that the proposed sequential algorithm significantly outperforms state-of-the-art methods and attains the posterior Cramér-Rao lower bound even with training data limited to local regions.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10336409/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10336409/</guid>
</item>
<item>
<title>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</title>
<description><p><span><big>On Minimizing the Probability of Large Errors in Robust Point Cloud Registration</big></span><br></p><p><span><small><i>AMIT EFRAIM; Joseph M. Francos</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340111</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In solving a model fitting problem, the existence of outliers in the set of measurements can have a devastating effect on the solution accuracy. Traditionally, in order to overcome this problem, robust point cloud registration algorithms are composed of transformation hypothesis generation, followed by hypothesis evaluation aimed at selecting the best hypothesized estimate. Hypotheses evaluation is commonly performed using the sample consensus criterion. However, since this criterion accounts only for the consensus size, it fails when the maximal sample consensus is incorrect. We propose a new hypothesis evaluation approach, generalizing the sample consensus approach, where instead of the sample consensus, the transformation that maximizes the point clouds feature correlation is selected as the best hypothesis. The feature vector at each point contains information such as on local geometry and semantic context. Utilizing this information in the hypotheses evaluation and selection procedure allows for a correct decision even when the hypothesis yielding the maximal sample consensus is false. Consequently, the probability of selecting the correct model increases. We show both mathematically and empirically that substituting the sample consensus criterion with the proposed point cloud feature correlation hypothesis test (PC-FCHT) lowers the probability of large registration errors, compared to using the special case of sample consensus. The proposed PC-FCHT is applicable to any algorithm that follows the hypothesis generation and evaluation scheme, potentially improving the success rates of a wide family of point cloud registration algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345750/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345750/</guid>
</item>
<item>
<title>Joint PAPR and OBP Reduction for NC-OFDM Systems</title>
<description><p><span><big>Joint PAPR and OBP Reduction for NC-OFDM Systems</big></span><br></p><p><span><small><i>Hsuan-Fu Wang; Fang-Biau Ueng; Bo-Heng Yeh</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3329757</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The spectrum resource is always a critical issue for wireless communications since it directly impacts the data rate and capacity. However, the problem of spectrum resource scarcity always exists. Moreover, spectrum resource scarcity becomes more severe as new communication technologies and wireless applications sprout. Noncontiguous orthogonal frequency division multiplexing (NC-OFDM) is a multicarrier method for bandwidth utilization. Unfortunately, this system has two fatal defects: high peak-to-average power ratio (PAPR) and considerable out-of-band power (OBP), which are detrimental to the system's performance. To solve these two problems, we propose a convex optimization-based method for joint PAPR and OBP reduction in NC-OFDM Systems. The strategy is to permit the secondary user to utilize the unoccupied spectrum of the primary user with dynamic spectrum sharing (DSS) based on a cognitive radio network (CRN). To this end, a flexible system operating over noncontiguous bands and DSS scenarios is necessary. The simulation results have shown that our method could effectively improve the overall performance and outperform other schemes, i.e., projections onto convex sets (POCS) and alternating projections onto convex and non-convex sets (APOCNCS), without harming the transmission of the primary system. The collaboration between secondary and primary systems is viable with the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10305259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10305259/</guid>
</item>
<item>
<title>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</title>
<description><p><span><big>Kronecker-Product Beamforming With Sparse Concentric Circular Arrays</big></span><br></p><p><span><small><i>Gal Itzhak; Israel Cohen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339433</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article presents a Kronecker-product (KP) beamforming approach incorporating sparse concentric circular arrays (SCCAs). The locations of the microphones on the SCCA are optimized concerning the broadband array directivity over a wide range of direction-of-arrival (DOA) deviations of a desired signal. A maximum directivity factor (MDF) sub-beamformer is derived accordingly with the optimal locations. Then, we propose two global beamformers obtained as a Kronecker product of a uniform linear array (ULA) and the SCCA sub-beamformer. The global beamformers differ by the type of the ULA, which is designed either as an MDF sub-beamformer along the
$\mathsf {x}$
-axis or as a maximum white noise gain sub-beamformer along the
$\mathsf {y}$
-axis. We analyze the performance of the proposed beamformers in terms of the directivity factor, the white noise gain, and their spatial beampatterns. Compared to traditional beamformers, the proposed beamformers exhibit considerably larger tolerance to DOA deviations concerning both the azimuth and elevation angles. Experimental results with speech signals in noisy and reverberant environments demonstrate that the proposed approach outperforms traditional beamformers regarding the perceptual evaluation of speech quality (PESQ) and short-time objective intelligibility (STOI) scores when the desired speech signals deviate from the nominal DOA.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342869/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342869/</guid>
</item>
<item>
<title>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</title>
<description><p><span><big>A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation</big></span><br></p><p><span><small><i>Karn N. Watcharasupat; Chih-Wei Wu; Yiwei Ding; Iroro Orife; Aaron J. Hipple; Phillip A. Williams; Scott Kramer; Alexander Lerch; William Wolcott</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339428</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Cinematic audio source separation is a relatively new subtask of audio source separation, with the aim of extracting the dialogue, music, and effects stems from their mixture. In this work, we developed a model generalizing the Bandsplit RNN for any complete or overcomplete partitions of the frequency axis. Psychoacoustically motivated frequency scales were used to inform the band definitions which are now defined with redundancy for more reliable feature extraction. A loss function motivated by the signal-to-noise ratio and the sparsity-promoting property of the 1-norm was proposed. We additionally exploit the information-sharing property of a common-encoder setup to reduce computational complexity during both training and inference, improve separation performance for hard-to-generalize classes of sounds, and allow flexibility during inference time with detachable decoders. Our best model sets the state of the art on the Divide and Remaster dataset with performance above the ideal ratio mask for the dialogue stem.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342812/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342812/</guid>
</item>
<item>
<title>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</title>
<description><p><span><big>Low Cost Variable Step-Size LMS With Maximum Similarity to the Affine Projection Algorithm</big></span><br></p><p><span><small><i>Miguel Ferrer; María de Diego; Alberto Gonzalez</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340106</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The LMS algorithm is widely employed in adaptive systems due to its robustness, simplicity, and reasonable performance. However, it is well known that this algorithm suffers from a slow convergence speed when dealing with colored reference signals. Numerous variants and alternative algorithms have been proposed to address this issue, though all of them entail an increase in computational cost. Among the proposed alternatives, the affine projection algorithm stands out. This algorithm has the peculiarity of starting from
$N$
data vectors of the reference signal. It transforms these vectors into as many data vectors suitably normalized in energy and mutually orthogonal. In this work, we propose a version of the LMS algorithm that, similar to the affine projection algorithm, starts from
$N$
data vectors of the reference signal but corrects them by using only a scalar factor that functions as a convergence step. Our goal is to align the behavior of this algorithm with the behavior of the affine projection algorithm without significantly increasing the computational cost of the LMS.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345730/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345730/</guid>
</item>
<item>
<title>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</title>
<description><p><span><big>Source-Data-Free Cross-Domain Knowledge Transfer for Semantic Segmentation</big></span><br></p><p><span><small><i>Zongyao Li; Ren Togo; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340616</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This article proposes a method for transferring knowledge of semantic segmentation from a labeled source domain to an unlabeled target domain without using the source-domain data. Such a problem is called source-data-free domain adaptation, in which a pre-trained source-domain model and the unlabeled target-domain data are used to transfer the label knowledge across the domains. Like most previous methods, our method uses pseudo labels for distilling and transferring the source-domain knowledge. On the basis of the pseudo-label learning, our method improves the domain adaptation performance in two innovative ways: 1) reducing the domain differences by source-data-free style transfer and 2) exploring the style diversity within the target domain by style modification. To this end, we introduce two additional modules: 1) an inter-domain style transfer module which aligns the feature statistics of the source and target domains before producing the pseudo labels thereby improving the pseudo labels' accuracy, and 2) an intra-domain style modification module which modifies the image styles within the target domain for learning intra-domain style-invariant features. Our method with the two modules outperforms previous source-data-free domain adaptation methods in two commonly used benchmarks. Moreover, our method is well compatible with the previous methods for further improvement.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10356748/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10356748/</guid>
</item>
<item>
<title>Contactless Skin Blood Perfusion Imaging via Multispectral Information, Spectral Unmixing and Multivariable Regression</title>
<description><p><span><big>Contactless Skin Blood Perfusion Imaging via Multispectral Information, Spectral Unmixing and Multivariable Regression</big></span><br></p><p><span><small><i>Liliana Granados-Castro; Omar Gutierrez-Navarro; Aldo Rodrigo Mejia-Rodriguez; Daniel Ulises Campos-Delgado</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2024.3381892</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Noninvasive methods for assessing in-vivo skin blood perfusion parameters, such as hemoglobin oxygenation, are crucial for diagnosing and monitoring microvascular diseases. This approach is particularly beneficial for patients with compromised skin, where standard contact-based clinical devices are inappropriate. For this goal, we propose the analysis of multimodal data from an occlusion protocol applied to 18 healthy participants, which includes multispectral imaging of the whole hand and reference photoplethysmography information from the thumb. Multispectral data analysis was conducted using two different blind linear unmixing methods: principal component analysis (PCA), and extended blind endmember and abundance extraction (EBEAE). Perfusion maps for oxygenated and deoxygenated hemoglobin changes in the hand were generated using linear multivariable regression models based on the unmixing methods. Our results showed high accuracy, with
$\text {R}^{2}$
-adjusted values, up to 0.90
$\pm$
0.08. Further analysis revealed that using more than four characteristic components during spectral unmixing did not improve the fit of the model. Bhattacharyya distance results showed that the fitted models with EBEAE were more sensitive to hemoglobin changes during occlusion stages, up to four times higher than PCA. Our study concludes that multispectral imaging with EBEAE is effective in quantifying changes in oxygenated hemoglobin levels, especially when using 3 to 4 characteristic components. Our proposed method holds promise for the noninvasive diagnosis and monitoring of superficial microvascular alterations across extensive anatomical regions.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10480236/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10480236/</guid>
</item>
<item>
<title>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</title>
<description><p><span><big>A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems</big></span><br></p><p><span><small><i>Tobias Kabzinski; Peter Jax</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337721</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Quasi-continuous system identification of time-variant linear acoustic systems can be applied in various audio signal processing applications when numerous acoustic transfer functions must be measured. A prominent application is measuring head-related transfer functions. We treat the underlying multiple-input-multiple-output (MIMO) system identification problem in a state-space model as a joint estimation problem for states, representing impulse responses, and state-space model parameters using the expectation maximization (EM) algorithm. We address limitations of prior work by imposing different model structures, especially for dependencies within a (transformed) state vector. This results in block diagonal matrix structures, for which we derive M-step update rules. Making assumptions about this model structure and choosing a block size for a given application define the computational complexity. In examples, we found that applying this framework yields improvements of up to 10 dB in relative system distance in comparison to a conventional method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334061/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334061/</guid>
</item>
<item>
<title>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</title>
<description><p><span><big>Analytical Performance Assessment of 2-D Tensor ESPRIT in Terms of Physical Parameters</big></span><br></p><p><span><small><i>Damir Rakhimov; Martin Haardt</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337729</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this paper, we present an analytical performance assessment of 2-D Tensor ESPRIT in terms of physical parameters. We show that the error in the
$r$
-mode depends only on two components, irrespective of the dimensionality of the problem. We obtain analytical expressions in closed form for the mean squared error (MSE) in each dimension as a function of the signal-to-noise (SNR) ratio, the array steering matrices, the number of antennas, the number of snapshots, the selection matrices, and the signal correlation. The derived expressions allow a better understanding of the difference in performance between the tensor and the matrix versions of the ESPRIT algorithm. The simulation results confirm the coincidence between the presented analytical expression and the curves obtained via Monte Carlo trials. We analyze the behavior of each of the two error components in different scenarios.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334446/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334446/</guid>
</item>
<item>
<title>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</title>
<description><p><span><big>Generalised Hyperbolic State-Space Models for Inference in Dynamic Systems</big></span><br></p><p><span><small><i>Yaman Kındap; Simon Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343341</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this work we study linear vector stochastic differential equation (SDE) models driven by the generalised hyperbolic (GH) Lévy process for inference in continuous-time non-Gaussian filtering problems. The GH family of stochastic processes offers a flexible framework for modelling of non-Gaussian, heavy-tailed characteristics and includes the normal inverse-Gaussian, variance-gamma and Student-t processes as special cases. We present continuous-time simulation methods for the solution of vector SDE models driven by GH processes and novel inference methodologies using a variant of sequential Markov chain Monte Carlo (MCMC). As an example a particular formulation of Langevin dynamics is studied within this framework. The model is applied to both a synthetically generated data set and a real-world financial series to demonstrate its capabilities.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360268/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360268/</guid>
</item>
<item>
<title>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</title>
<description><p><span><big>Dynamic Time Signature Recognition, Tempo Inference, and Beat Tracking Through the Metrogram Transform</big></span><br></p><p><span><small><i>James M. Cozens; Simon J. Godsill</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344048</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper proposes a probabilistic approach for extracting time-varying and irregular time signature information from polyphonic audio extracts, subsequently providing beat and bar line positions given inferred time signature divisions. This is achieved via dynamically evaluating the beat tempo as a function of time through finding an optimal compromise in beat and bar alignment in the time and tempo domains. Time signature divisions are determined based on a new representation, termed the Metrogram, that presents time-varying information regarding rhythmic and metric periodicities in the Tempogram. Our methodology is characterised by its ability to provide a distribution over metric interpretations, offering insights into the diverse ways music can be rhythmically perceived. Results indicate high-level accuracy for a variety of polyphonic extracts containing irregular, complex, irrational, and time-varying time signatures. Accuracy rivalling state-of-the-art methodologies is also reported in a beat tracking task performed on the standard Ballroom Dataset. The paper offers insights into the field of dynamic time signature recognition and beat tracking, offering a valuable and versatile resource for the analysis, composition, and performance of music.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363392/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363392/</guid>
</item>
<item>
<title>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</title>
<description><p><span><big>TolerantGAN: Text-Guided Image Manipulation Tolerant to Real-World Image</big></span><br></p><p><span><small><i>Yuto Watanabe; Ren Togo; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343335</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Although text-guided image manipulation approaches have demonstrated highly accurate performance for editing the appearance of images in a virtual or simple scenario, their real-world applications face significant challenges. The primary cause of these challenges is the misalignment in the distribution of training and real-world data, which leads to unstable text-guided image manipulation. In this work, we propose a novel framework called TolerantGAN and tackle the new task of real-world text-guided image manipulation independent of the training data. To achieve this, we introduce two key concepts of a border smoothly connection module (BSCM) and a manipulation direction-based attention module (MDAM). BSCM smoothens the misalignment in the distribution of training and real-world data. MDAM extracts only regions highly relevant for image manipulation and assists in reconstructing unobserved objects in the training data. For in-the-wild input images of various classes, TolerantGAN robustly outperforms the state-of-the-art methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360283/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360283/</guid>
</item>
<item>
<title>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</title>
<description><p><span><big>Streaming ASR Encoder for Whisper-to-Speech Online Voice Conversion</big></span><br></p><p><span><small><i>Anastasia Avdeeva; Aleksei Gusev; Tseren Andzhukaev; Artem Ivanov</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343342</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Whispered speech is a quiet voice without vocalization. One of the common cases of using whispered speech is a technique that can help overcome stuttering. But whispered speech can be uncomfortable and difficult to understand in everyday communication. To address these problems, we propose a method of low-delayed whisper-to-speech voice conversion, which can be useful in real life communication of people with disordered speech. As part of our research, we study the impact of streaming Automatic Speech Recognition models on the quality of voice conversion, comparing different streaming models and methods for model adaptation to streaming settings, and showing the importance of using such models in cases of low-delayed voice conversion.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360259/</guid>
</item>
<item>
<title>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</title>
<description><p><span><big>Sparse Representation With Gaussian Atoms and Its Use in Anomaly Detection</big></span><br></p><p><span><small><i>Denis C. Ilie-Ablachim; Andra Băltoiu; Bogdan Dumitrescu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344313</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We propose sparse representation and dictionary learning algorithms for dictionaries whose atoms are characterized by Gaussian probability distributions around some central atoms. This extension of the space covered by the atoms permits a better characterization of localized features of the signals. The representations are computed by trading-off representation error and the probability of the actual atoms used in the representation. We propose two types of algorithms: a greedy one, similar in spirit with Orthogonal Matching Pursuit, and one based on L1-regularization. We apply our new algorithms to unsupervised anomaly detection, where the representation error is used as anomaly score. The flexibility of our approach appears to improve more the representations of the many normal signals than those of the few outliers, at least for an anomaly type called dependency, thus improving the detection quality. Comparison with both standard dictionary learning algorithms and established anomaly detection methods is favorable.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365333/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365333/</guid>
</item>
<item>
<title>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</title>
<description><p><span><big>Zero-Shot Visual Sentiment Prediction via Cross-Domain Knowledge Distillation</big></span><br></p><p><span><small><i>Yuya Moroto; Yingrui Ye; Keisuke Maeda; Takahiro Ogawa; Miki Haseyama</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344079</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>There are various sentiment theories for categorizing human sentiments into several discrete sentiment categories, which means that the theory used for training sentiment prediction methods does not always match that used in the test phase. As a solution to this problem, zero-shot visual sentiment prediction methods have been proposed to predict unseen sentiments for which no images are available in the training phase. However, the training of these previous zero-shot methods relies on a single sentiment theory, which limits their ability to handle sentiments from other theories. Thus, this article proposes a more robust zero-shot visual sentiment prediction method that can handle cross-domain sentiments defined in different sentiment theories. Specifically, by focusing on the fact that sentiments are abstract concepts common to humans regardless of whether their theories are different, we incorporate knowledge distillation into our method to construct a teacher–student model that can train the implicit relationships between sentiments defined in different sentiment theories. Furthermore, to enhance sentiment discrimination capability and strengthen the implicit relationships between sentiments, we introduce a novel sentiment loss between the teacher and student models. In this way, our model becomes robust to unseen sentiments by exploiting the implicit relationships between sentiments. The contributions of this article are the introduction of knowledge distillation and a novel sentiment loss between the teacher and student models for zero-shot visual sentiment prediction, and improved performance of zero-shot visual sentiment prediction. Experiments on several open datasets demonstrate the effectiveness of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363382/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363382/</guid>
</item>
<item>
<title>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</title>
<description><p><span><big>Unrolling of Simplicial ElasticNet for Edge Flow Signal Reconstruction</big></span><br></p><p><span><small><i>Chengen Liu; Geert Leus; Elvin Isufi</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3339376</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>The edge flow reconstruction task consists of retreiving edge flow signals from corrupted or incomplete measurements. This is typically solved by a regularized optimization problem on higher-order networks such as simplicial complexes and the corresponding regularizers are chosen based on prior knowledge. Tailoring this prior to the setting of interest can be challenging or it may not even be possible. Thus, we consider to learn this prior knowledge via a model-based deep learning approach. We propose a new regularized optimization problem for the simplicial edge flow reconstruction task, the simplicial ElasticNet, which combines the advantages of the
$\ell _{1}$
and
$\ell _{2}$
norms. We solve the simplicial ElasticNet problem via the multi-block alternating direction method of multipliers (ADMM) algorithm and provide conditions on its convergence. By unrolling the ADMM iterative steps, we develop a model-based neural network with a low requirement on the number of training data. This unrolling network replaces the fixed parameters in the iterative algorithm by learnable weights, thus exploiting the neural network's learning capability while preserving the iterative algorithm's interpretability. We enhance this unrolling network via simplicial convolutional filters to aggregate information from the edge flow neighbors, ultimately, improving the network learning expressivity. Extensive experiments on real-world and synthetic datasets validate the proposed approaches and show considerable improvements over both baselines and traditional non-model-based neural networks.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10342735/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10342735/</guid>
</item>
<item>
<title>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</title>
<description><p><span><big>Principled Pruning of Bayesian Neural Networks Through Variational Free Energy Minimization</big></span><br></p><p><span><small><i>Jim Beckers; Bart Van Erp; Ziyue Zhao; Kirill Kondrashov; Bert De Vries</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3337718</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Bayesian model reduction provides an efficient approach for comparing the performance of all nested sub-models of a model, without re-evaluating any of these sub-models. Until now, Bayesian model reduction has been applied mainly in the computational neuroscience community on simple models. In this paper, we formulate and apply Bayesian model reduction to perform principled pruning of Bayesian neural networks, based on variational free energy minimization. Direct application of Bayesian model reduction, however, gives rise to approximation errors. Therefore, a novel iterative pruning algorithm is presented to alleviate the problems arising with naive Bayesian model reduction, as supported experimentally on the publicly available UCI datasets for different inference algorithms. This novel parameter pruning scheme solves the shortcomings of current state-of-the-art pruning methods that are used by the signal processing community. The proposed approach has a clear stopping criterion and minimizes the same objective that is used during training. Next to these benefits, our experiments indicate better model performance in comparison to state-of-the-art pruning schemes.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10334001/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10334001/</guid>
</item>
<item>
<title>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</title>
<description><p><span><big>Exploiting a Spatial Attention Mechanism for Improved Depth Completion and Feature Fusion in Novel View Synthesis</big></span><br></p><p><span><small><i>Anh Minh Truong; Wilfried Philips; Peter Veelaert</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3340064</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Many image-based rendering (IBR) methods rely on depth estimates obtained from structured light or time-of-flight depth sensors to synthesize novel views from sparse camera networks. However, these estimates often contain missing or noisy regions, resulting in an incorrect mapping between source and target views. This situation makes the fusion process more challenging, as the visual information is misaligned, inconsistent, or missing. In this work, we first implement a lightweight network based on the transformer, which is well-known for its capability to model long-range relationships within input data, to extract spatial features from color images. These features are then used to enhance the quality of completed depth maps. Furthermore, we combine a sequential deep neural network with a spatial attention mechanism to effectively fuse the projected features from multiple source viewpoints. This approach enables us to integrate information from an arbitrary number of source viewpoints as well as improve accuracy in synthesized views. Experimental results on challenging datasets demonstrate that our method achieves superior synthesized image quality compared to state-of-the-art (SOTA) methods.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10345792/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10345792/</guid>
</item>
<item>
<title>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</title>
<description><p><span><big>Smoothing ADMM for Sparse-Penalized Quantile Regression With Non-Convex Penalties</big></span><br></p><p><span><small><i>Reza Mirzaeifard; Naveen K. D. Venkategowda; Vinay Chakravarthi Gogineni; Stefan Werner</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344395</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>This paper investigates quantile regression in the presence of non-convex and non-smooth sparse penalties, such as the minimax concave penalty (MCP) and smoothly clipped absolute deviation (SCAD). The non-smooth and non-convex nature of these problems often leads to convergence difficulties for many algorithms. While iterative techniques such as coordinate descent and local linear approximation can facilitate convergence, the process is often slow. This sluggish pace is primarily due to the need to run these approximation techniques until full convergence at each step, a requirement we term as a
secondary convergence iteration
. To accelerate the convergence speed, we employ the alternating direction method of multipliers (ADMM) and introduce a novel single-loop smoothing ADMM algorithm with an increasing penalty parameter, named SIAD, specifically tailored for sparse-penalized quantile regression. We first delve into the convergence properties of the proposed SIAD algorithm and establish the necessary conditions for convergence. Theoretically, we confirm a convergence rate of
$o({k^{-\frac{1}{4}}})$
for the sub-gradient bound of the augmented Lagrangian, where
$k$
denotes the number of iterations. Subsequently, we provide numerical results to showcase the effectiveness of the SIAD algorithm. Our findings highlight that the SIAD method outperforms existing approaches, providing a faster and more stable solution for sparse-penalized quantile regression.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365338/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365338/</guid>
</item>
<item>
<title>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</title>
<description><p><span><big>Effects of Doppler-Division Multiplexing on OFDM Joint Sensing and Communication Systems</big></span><br></p><p><span><small><i>Oliver Lang; Christian Hofbauer; Reinhard Feger; Mario Huemer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343308</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>A promising waveform candidate for future joint sensing and communication systems is orthogonal frequency-division multiplexing (OFDM). For such systems, supporting multiple transmit antennas requires multiplexing methods for the generation of orthogonal transmit signals, where equidistant subcarrier interleaving (ESI) is the most popular multiplexing method. In this work, we analyze a multiplexing method called Doppler-division multiplexing (DDM). This method applies a phase shift from OFDM symbol to OFDM symbol to separate signals transmitted by different Tx antennas along the velocity axis of the range-Doppler map. The main focus of this work lies on the implications of DDM on the communication task. It will be shown that for DDM, the channels observed in the communication receiver are heavily time-varying, preventing any meaningful transmission of data when not taken into account. In this work, a communication system designed to combat these time-varying channels is proposed, which includes methods for data estimation, synchronization, and channel estimation. Bit error ratio (BER) simulations demonstrate the superiority of this communications system compared to ESI-based systems.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360223/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360223/</guid>
</item>
<item>
<title>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</title>
<description><p><span><big>Binaural Multichannel Blind Speaker Separation With a Causal Low-Latency and Low-Complexity Approach</big></span><br></p><p><span><small><i>Nils L. Westhausen; Bernd T. Meyer</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343320</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>In this article, we introduce a causal low-latency low-complexity approach for binaural multichannel blind speaker separation in noisy reverberant conditions. The model, referred to as Group Communication Binaural Filter and Sum Network (GCBFSnet) predicts complex filters for filter-and-sum beamforming in the time-frequency domain. We apply Group Communication (GC), i.e., latent model variables are split into groups and processed with a shared sequence model with the aim of reducing the complexity of a simple model only containing one convolutional and one recurrent module. With GC we are able to reduce the size of the model by up to 83% and the complexity up to 73% compared to the model without GC, while mostly retaining performance. Even for the smallest model configuration, GCBFSnet matches the performance of a low-complexity TasNet baseline in most metrics despite the larger size and higher number of required operations of the baseline.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360275/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360275/</guid>
</item>
<item>
<title>Reverse Ordering Techniques for Attention-Based Channel Prediction</title>
<description><p><span><big>Reverse Ordering Techniques for Attention-Based Channel Prediction</big></span><br></p><p><span><small><i>Valentina Rizzello; Benedikt Böck; Michael Joham; Wolfgang Utschick</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344024</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Channel state information (CSI) is crucial for enhancing the performance of wireless systems by allowing to adjust the transmission strategies based on the current channel conditions. However, obtaining precise CSI is difficult because of the fast-changing channel conditions caused by multi-path fading. An inaccurate CSI hinders the performance of various adaptive wireless systems, highlighting the need for channel prediction techniques to effectively mitigate the drawbacks of outdated CSI. Conventional methods typically depend on assumptions regarding user velocity or require knowledge of the Doppler frequency. In contrast to existing approaches, we aim for a more robust and practical solution by training neural networks without making any assumptions about user velocity, relying solely on noisy channel observations during training. Specifically, we adapt both the sequence-to-sequence with attention (Seq2Seq-attn) and transformer models for channel prediction. Additionally, a new technique called reverse positional encoding is introduced in the transformer model to improve the robustness of the model against varying sequence lengths. Similarly, the encoder outputs of the Seq2Seq-attn model are reversed prior to the application of attention mechanisms. By means of simulations, we show that these proposed techniques enable the models to effectively capture relationships within sequences of channel snapshots without increasing the complexity. Importantly, this capability remains robust across varying sequence lengths, representing a substantial improvement over existing methodologies.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10363354/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10363354/</guid>
</item>
<item>
<title>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</title>
<description><p><span><big>VoicePAT: An Efficient Open-Source Evaluation Toolkit for Voice Privacy Research</big></span><br></p><p><span><small><i>Sarina Meyer; Xiaoxiao Miao; Ngoc Thang Vu</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3344375</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Speaker anonymization is the task of modifying a speech recording such that the original speaker cannot be identified anymore. Since the first Voice Privacy Challenge in 2020, along with the release of a framework, the popularity of this research topic is continually increasing. However, the comparison and combination of different anonymization approaches remains challenging due to the complexity of evaluation and the absence of user-friendly research frameworks. We therefore propose an efficient speaker anonymization and evaluation framework based on a modular and easily extendable structure, almost fully in Python. The framework facilitates the orchestration of several anonymization approaches in parallel and allows for interfacing between different techniques. Furthermore, we propose modifications to common evaluation methods which improves the quality of the evaluation and reduces their computation time by 65 to 95%, depending on the metric. Our code is fully open source.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10365329/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10365329/</guid>
</item>
<item>
<title>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</title>
<description><p><span><big>Hybrid Packet Loss Concealment for Real-Time Networked Music Applications</big></span><br></p><p><span><small><i>Alessandro Ilic Mezza; Matteo Amerena; Alberto Bernardini; Augusto Sarti</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343318</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Real-time audio communications over IP have become essential to our daily lives. Packet-switched networks, however, are inherently prone to jitter and data losses, thus creating a strong need for effective packet loss concealment (PLC) techniques. Though solutions based on deep learning have made significant progress in that direction as far as speech is concerned, extending the use of such methods to applications of Networked Music Performance (NMP) presents significant challenges, including high fidelity requirements, higher sampling rates, and stringent temporal constraints associated to the simultaneous interaction between remote musicians. In this article, we present PARCnet, a hybrid PLC method that utilizes a feed-forward neural network to estimate the time-domain residual signal of a parallel linear autoregressive model. Objective metrics and a listening test show that PARCnet provides state-of-the-art results while enabling real-time operation on CPU.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360264/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360264/</guid>
</item>
<item>
<title>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</title>
<description><p><span><big>Masked Spectrogram Prediction for Unsupervised Domain Adaptation in Speech Enhancement</big></span><br></p><p><span><small><i>Katerina Zmolikova; Michael Syskind Pedersen; Jesper Jensen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3343343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>Supervised learning-based speech enhancement methods often work remarkably well in acoustic situations represented in the training corpus but generalize poorly to out-of-domain situations, i.e. situations not seen during training. This stands in the way of further improvement of these methods in realistic scenarios, as collecting paired noisy-clean recordings in the target application domain is typically not feasible. Recording noisy-only in-domain data is, though, much more practical. In this article, we tackle the problem of unsupervised domain adaptation in speech enhancement. Specifically, we propose a way to use in-domain noisy-only data in the training of a neural network to improve upon a model trained solely on out-of-domain paired data. For this, we make use of masked spectrogram prediction, a technique from self-supervised learning that aims to interpolate masked regions of a spectrogram. We hypothesize that masked spectrogram prediction encourages learning of features that represent well both speech and noise components of the noisy signals. These features can then be used to train a more robust speech enhancement system. We evaluate the proposed method on the VoiceBank-DEMAND and LibriFSD50k databases, with WSJ0-CHiME3 serving as the out-of-domain database. We show that the proposed method outperforms both the out-of-domain system and the baseline approaches, i.e. RemixIT and noisy-target training, and also combines well with the previously proposed RemixIT method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10360251/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10360251/</guid>
</item>
<item>
<title>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</title>
<description><p><span><big>Bilinear Expectation Propagation for Distributed Semi-Blind Joint Channel Estimation and Data Detection in Cell-Free Massive MIMO</big></span><br></p><p><span><small><i>Alexander Karataev; Christian Forsch; Laura Cottatellucci</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3348343</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>We consider a cell-free massive multiple-input multiple-output (CF-MaMIMO) communication system in the uplink transmission and propose a novel algorithm for blind or semi-blind joint channel estimation and data detection (JCD). We formulate the problem in the framework of bilinear inference and develop a solution based on the expectation propagation (EP) method for both channel estimation and data detection. We propose a new approximation of the joint a posteriori distribution of the channel and data whose representation as a factor graph enables the application of the EP approach using the message-passing technique, local low-complexity computations at the nodes, and an effective modeling of channel-data interplay. The derived algorithm, called bilinear-EP JCD, allows for a distributed implementation among access points (APs) and the central processing unit (CPU) and has polynomial complexity. Our simulation results show that it outperforms other EP-based state-of-the-art polynomial time algorithms.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10378663/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10378663/</guid>
</item>
<item>
<title>Adversarial Representation Learning for Robust Privacy Preservation in Audio</title>
<description><p><span><big>Adversarial Representation Learning for Robust Privacy Preservation in Audio</big></span><br></p><p><span><small><i>Shayan Gharib; Minh Tran; Diep Luong; Konstantinos Drossos; Tuomas Virtanen</i></small></span><br><span><small><i>https://doi.org/10.1109/OJSP.2023.3349113</i></small></span><br><span><small><i>Volume 5</i></small></span><br></p><p><span>S... |
http://localhost:1200/ieee/journal/4427201/earlyaccess - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>IEEE Sensors Journal</title>
<link>https://ieeexplore.ieee.org/xpl/tocresult.jsp?isnumber=4427201</link>
<atom:link href="http://localhost:1200/ieee/journal/4427201/earlyaccess" rel="self" type="application/rss+xml"></atom:link>
<description>IEEE Sensors Journal - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Tue, 14 May 2024 10:45:06 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>The Impact of GIMMF Pitch on Sensitivity of Curvature Sensor Based on Anti-resonance</title>
<description><p><span><big>The Impact of GIMMF Pitch on Sensitivity of Curvature Sensor Based on Anti-resonance</big></span><br></p><p><span><small><i>Yaxun Zhang; Yuxin Liu; Zhiliang Huang; Xiaouyun Tang; Lei Jin; Zhihai Liu; Yu Zhang; Libo Yuan</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3386720</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>We propose and demonstrate a high sensitivity curvature sensor based on the cascaded structure of single mode fiber-graded index multimode fiber-hollow core fiber-single mode fiber, and study the influence of the incident light angle on the sensitivity of anti-resonance curvature sensor. The angle of light entering the hollow core fiber can be modulated by the graded index multimode fiber. Different angles affect the intensity of the spectrum produced by the anti-resonance mechanism. The intensity of the resonance peak generated by the anti-resonance mechanism is sensitive to curvature. The sensor achieves a large dynamic range in the range of 0-9.21 m
-1
and high sensitivity curvature measurement. The maximum sensitivity of the curvature sensor is -5.00 dB/m
-1
. And there is hardly any temperature crosstalk sensitivity. The proposed curvature sensor has the advantages of large dynamic range, high sensitivity and low temperature crosstalk.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10500302/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10500302/</guid>
</item>
<item>
<title>Design of Highly Sensitive Joint Torque Sensor for Collision Detection in Robotic Manipulators</title>
<description><p><span><big>Design of Highly Sensitive Joint Torque Sensor for Collision Detection in Robotic Manipulators</big></span><br></p><p><span><small><i>Sreekanth Makkal Mohandas; Rajesh Kannan Megalingam; Dhananjay Raghavan</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3395788</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Joint torque sensors are widely used in robots for collision detection and continuous monitoring of the joint torque during path planning, manipulation, and tool operation. There is a strong demand for such sensors that ensure efficient collision detection and safe operation by virtue of their responsive sensitivity, overload capacity, and avoidance of resonance. This research investigated a highly sensitive strain gauge-based joint torque sensor design, with a novel pear-shaped hole on its spoke structure. A safety factor of 4 was assumed during the design phase to secure the high overload capacity of the sensor structure. The proffered sensor was designed for robotic applications involving high-frequency vibrations, with minimal risk of achieving mechanical resonance. Finite element method optimization was carried out to maximize the sensitivity, precluding resonance. An instrumentation amplifier-based sensor-mountable data acquisition unit was also designed and fabricated to acquire the amplified strain gauge data. The experimental results showed that the proposed sensor has 38 mV/Nm sensitivity and 0.04 Nm resolution in the measurement range of ±25 Nm torque. The sensor resolution and sensitivity were shown to be 60% and 26.78% better than extant sensors, respectively. The reduction in strain over the sensing area along the length of the spoke structure was constrained to 5.42%, which secured uniform strain distribution under strain gauges, for enhanced sensitivity. The proposed methodology can very well be applied to design sensors with maximum sensitivity for the prescribed vibration constraints.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10522600/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10522600/</guid>
</item>
<item>
<title>Multi-sensor multiple extended objects tracking based on the message passing</title>
<description><p><span><big>Multi-sensor multiple extended objects tracking based on the message passing</big></span><br></p><p><span><small><i>Yuansheng Li; Tao Shen; Lin Gao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3384560</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Multiple extended objects (EOs) tracking has attracted a lot of attention due to the fast development of high-resolution sensors. A remarkable feature of EO, compared to the traditional point target, is that an EO normally produces more than one measurement, resulting in challenges for finding the associations among objects and measurements. In this paper, we are interested in tracking an unknown number of EOs with multiple sensors by resorting to the message passing method. The existence probability and belief of each EO are explicitly estimated, where the belief is modeled by a mixture gamma Gaussian inverse Wishart (GGIW) distribution, so as to jointly estimate the measurement rate (MR), centroid state and extension. The marginal posterior of EOs is approximately by belief, which is obtained by running the loopy sum-product algorithm (LSPA) on a suitably devised factor graph. As a result, the computational load of the proposed algorithm increases linearly with respect to the number of targets, thus admitting the scalability. Simulation experiments are carried out to verify the performance of the proposed method.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10495766/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10495766/</guid>
</item>
<item>
<title>An Intestine based Bio-Compatible Humidity Sensor for Environmental and Medical Measurements</title>
<description><p><span><big>An Intestine based Bio-Compatible Humidity Sensor for Environmental and Medical Measurements</big></span><br></p><p><span><small><i>Emrehan Yavsan; Muhammet Rojhat Kara; Mehmet Akif Erismis</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3367032</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The bio-compatible devices suitable for recycling and bio-degrading can be achieved with organic materials in nature. In this work, a bio-compatible capacitive humidity sensor is presented for reducing the amount of electronic waste and contributing to the sustainability of natural resources and the future. The sensor consists of 3 layers. The first layer is the processed intestine layer of cattle. Bio-compatibility is achieved with this layer. In addition to being a highly absorbing tissue, the intestine has been used for centuries for long-term preservation of meat based food. Correspondingly, the developed sensor is found to be more durable and long-lasting than other natural-material based humidity sensors in the literature. The other layers of the sensor are interdigitated copper electrodes and a 0.2 mm thick thin film strip. Thin film strip increases mechanical strength as well as flexibility. The developed sensor prototype was subjected to various tests in the humidity range of 20%-90%. In these tests, the hysteresis characteristic of the sensor, its response-recovery time, and its long-term stability and short-term step responses were examined. Moreover, as a possible application in medicine, the sensor can be used to detect breathing cycles. The sensor’s response and recovery times were measured as 8.72’ and 4.47’, respectively, possibly attributed to the stabilization of our test setup, while the sensor successfully detected deep, normal and fast breathing. Despite being kept in an uncontrolled environment, the sensor continued to operate consistently for breath measurements after 56 weeks, which is more than a year.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10445282/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10445282/</guid>
</item>
<item>
<title>Deep neural network assisted terahertz metasurface sensors for the detection of lung cancer biomarkers</title>
<description><p><span><big>Deep neural network assisted terahertz metasurface sensors for the detection of lung cancer biomarkers</big></span><br></p><p><span><small><i>Jie Chen; Fangrong Hu; Xiaoya Ma; Mo Yang; Shangjun Lin; An Su</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3384578</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>MicroRNA (miRNA) is closely related to various cancers, and the change in its expression level is closely related to the development and death of tumor cells. Here, we design and manufacture a terahertz (THz) metasurface sensor to realize the concentration detection and category identification of the miRNAs related to lung cancer. We established two spectral classification datasets, one contains 9 concentrations of miRNAs, and the other contains 3 categories of lung cancer-related miRNAs. And then, we used a deep neural network (DNN) algorithm to classify these spectral datasets and obtained a LOD (Limit-of-Detection) of 100 aM for miRNAs. Moreover, this method can identify the categories of miRNAs. Compared with the other five machine learning (ML) algorithms, the proposed neural network framework achieves the best classification results. This work provides a new way for the detection and identification of trace nucleic acid and biomarkers of many cancers.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10495770/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10495770/</guid>
</item>
<item>
<title>Surface and Interface investigation on MoS2-rGO hybrids for room temperature gas sensing</title>
<description><p><span><big>Surface and Interface investigation on MoS2-rGO hybrids for room temperature gas sensing</big></span><br></p><p><span><small><i>Saurabh Rawat; Priyanka Bamola; Karishma; Chanchal Rani; Hemlata Dhoundiyal; Nikita Sharma; Charu Dwivedi; Ujjwal Kumar; Pushp Sen Satyarthi; Mohit Sharam; Rajesh Kumar; Himani Sharma</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3396630</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Novel resistive copper PCB-based sensors have been developed for room temperature NH
3
gas detection at a concentration as low as 10 parts per million (ppm), employing a binary hybrid composition of Molybdenum Disulfide (MoS
2
) and reduced graphene oxide (rGO). These sensors utilize hierarchical structures that outperform unary MoS
2
counterparts, with an 45% enhancement in sensing response at 10 ppm NH
3
concentration. Surface analysis studies reveals surface charges, electronic interaction at the surface interface. The mentioned studies clarify gas-surface interactions, guiding material design for optimized response and sensitivity in gas sensing. It validates sensor performance, enhances understanding, and refines designs for practical applications. Notably, the interface and surface of hybrids has been thoroughly analyzed by X-Ray Photoelectron spectroscopy (XPS), and Brunauer-Emmett-Teller (BET) respectively, enhancing our understanding of the structural and chemical aspects driving the enhanced sensing performance. This increase in sensitivity is underpinned by nanoscale electronic modifications at interfaces, particularly within nano heterojunctions, which not only amplify adsorption but also boosted selectivity. The binary hybrid device, showcasing superior NH
3
detection and specificity over volatile gases like methanol, ethanol, isopropyl alcohol and toxic Carbon monoxide gas emerges as a robust candidate for selective gas sensing. Additionally, the foundation of these advancements is fortified by incorporating theoretical surface potential calculations, underscoring a significant stride towards the advancement of room temperature gas sensing technology.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10528245/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10528245/</guid>
</item>
<item>
<title>A Casting Surface Dataset and Benchmark for Subtle and Confusable Defect Detection in Complex Contexts</title>
<description><p><span><big>A Casting Surface Dataset and Benchmark for Subtle and Confusable Defect Detection in Complex Contexts</big></span><br></p><p><span><small><i>Qishan Wang; Shuyong Gao; Li Xiong; Aili Liang; Kaidong Jiang; Wenqiang Zhang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3387082</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Industrial anomaly detection (IAD) algorithms are essential for implementing automated quality inspection. Dataset diversity serves as the foundation for developing comprehensive detection algorithms. Existing IAD datasets focus on the diversity of objects and defects, overlooking the diversity of domains within the real data. To bridge this gap, this study proposes the Casting Surface Defect Detection (CSDD) dataset, containing 12,647 high-resolution gray images and pixel-precise ground truth labels for all defect samples. Compared to existing datasets, CSDD has the following two characteristics: 1.) The target samples are unaligned and have complex and variable context information. 2.) The defects in the CSDD dataset samples are subtle and confusable by factors such as oil contamination, processing features, and machining marks, illustrating the challenge of detecting real casting defects in an industrial context. Based on this dataset, we observe that current state-of-the-art (SOTA) IAD methods face challenges when there is considerable variation in sample context information. Furthermore, these methods encounter difficulties when abnormal samples are scarce, particularly those samples with subtle and confusable defects. To address this issue, we propose a novel method called Realistic Synthetic Anomalies (RSA), which enhances the model’s capacity to construct a normal sample distribution by generating a large number of realistic synthetic anomalies. Experimental results demonstrate that the model trained to classify synthetic anomalies from normal samples achieves the highest accuracy for casting surface defect detection and significantly improves detection accuracy for subtle and confusable defects. The CSDD dataset and code of RSA are available at https://github.com/18894269590/RSA.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10502267/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10502267/</guid>
</item>
<item>
<title>PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards</title>
<description><p><span><big>PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards</big></span><br></p><p><span><small><i>Elias Arbash; Margret Fuchs; Behnood Rasti; Sandra Lorenz; Pedram Ghamisi; Richard Gloaguen</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3380826</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Addressing the critical theme of recycling electronic waste (E-waste), this contribution is dedicated to developing advanced automated data processing pipelines as a basis for decision-making and process control. Aligning with the broader goals of the circular economy and the United Nations (UN) Sustainable Development Goals (SDG), our work leverages non-invasive analysis methods utilizing RGB and hyperspectral imaging data to provide both quantitative and qualitative insights into the E-waste stream composition for optimizing recycling efficiency. In this paper, we introduce ’PCB-Vision’; a pioneering RGB-hyperspectral printed circuit board (PCB) benchmark dataset, comprising 53 RGB images of high spatial resolution paired with their corresponding high spectral resolution hyperspectral data cubes in the visible and near-infrared (VNIR) range. Grounded in open science principles, our dataset provides a comprehensive resource for researchers through high-quality ground truths, focusing on three primary PCB components: integrated circuits (IC), capacitors, and connectors. We provide extensive statistical investigations on the proposed dataset together with the performance of several state-of-the-art (SOTA) models, including U-Net, Attention UNet, Residual U-Net, LinkNet, and DeepLabv3+. By openly sharing this multiscene benchmark dataset along with the baseline codes, we hope to foster transparent, traceable, and comparable developments of advanced data processing across various scientific communities, including, but not limited to, computer vision and remote sensing. Emphasizing our commitment to supporting a collaborative and inclusive scientific community, all materials, including code, data, ground truth, and masks, will be accessible at https://github.com/hifexplo/PCBVision.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10485259/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10485259/</guid>
</item>
<item>
<title>AuNP-decorated textile as chemo resistive sensor for acetone detection</title>
<description><p><span><big>AuNP-decorated textile as chemo resistive sensor for acetone detection</big></span><br></p><p><span><small><i>S. Casalinuovo; D. Caschera; S. Quaranta; D. Caputo</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2023.3348693</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This paper presents the development of chemo resistive sensors for the detection of volatile organic compounds (VOCs). The proposed sensor is based on citrate-functionalized gold nanoparticles (AuNPs) serving as a sensitive layer deposited on cotton fabric. Impedance variations due to VOC/substrate interaction are used as a detection principle. Specifically, this work focuses on acetone detection after exposing the AuNP-decorated cotton to a CH3COCH3 aqueous solution. Such an interaction resulted in a reduction of the total impedance (i.e., magnitude) of the system. This behavior can be ascribed to Van der Waals forces existing between the C=O group and the citrate moieties adsorbed on the gold nanoparticles, which favor charge injection to the substrate. Response to water was also tested for comparison, assuring that the solvent interacts with the sensitive layer by a different adsorption mechanism, not influencing the overall results. Sensor selectivity was also verified by considering ethanol (representative of alcohol group). Indeed, impedance curves reflect the different type of chemical interaction between the analyte and the substrate. In addition, sensor limit of detection for acetone was found to be 1% v/v, in the considered frequency range. Furthermore, sensor performance in terms of reusability was evaluated, showing that the Au-cotton ability in VOCs detection could be restored after about 90 min with a percentage up to 97 % in the frequency of 1Hz. These results can be considered the starting point for the development of portable, sensitive and user-friendly devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10384362/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10384362/</guid>
</item>
<item>
<title>Few-Shot Class-Incremental Learning with Adjustable Pseudo-Incremental Sessions for Bearing Fault Diagnosis</title>
<description><p><span><big>Few-Shot Class-Incremental Learning with Adjustable Pseudo-Incremental Sessions for Bearing Fault Diagnosis</big></span><br></p><p><span><small><i>Hongyan Zhu; Changqing Shen; Jiaan Wang; Bojian Chen; Dong Wang; Zhongkui Zhu</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3395515</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Rotating machinery may constantly generate new classes of faults in complex operating environments, with a finite set of fault samples that are obtainable. The incremental nature of fault types and the scarcity of samples present great challenges for fault diagnosis models employing deep learning techniques, such as overfitting due to insufficient data and stability–plasticity problem when new fault samples are introduced. A few-shot class-incremental learning (FSCIL) method with adjustable pseudo-incremental sessions (APIS) is introduced in this study to overcome the abovementioned concerns. First, the method extracts fault samples from the sufficient data of the real base session, and it sets the pseudo FSCIL task according to the data format of FSCIL. The pseudo FSCIL task is employed to extract invariant features in the base and incremental session, and the feature space is trained to provide a reference model for the real fault incremental task. Second, the adjuster based on a self-attention mechanism is used to learn more discriminative features. Feature differences are observed between the old fault classifier and a fresh fault prototype. Based on the self-attention mechanism, the distinguishing features between the fault samples in the testing dataset and the old classifier are highlighted. In this way, a fault diagnosis model of the few-shot fault class increment with more generalization ability is acquired. A case study shows that APIS can effectively alleviate stability–plasticity and overfitting problems when dealing with FSCIL bearing fault diagnosis tasks.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10521474/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10521474/</guid>
</item>
<item>
<title>Failure Mechanism Analysis and Experiment of MEMS VRG under High-g Shock</title>
<description><p><span><big>Failure Mechanism Analysis and Experiment of MEMS VRG under High-g Shock</big></span><br></p><p><span><small><i>Jin Wang; Qi Cai; Wenqiang Wei; Rang Cui; Yunbo Shi; Chong Shen; Huiliang Cao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3389744</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>This paper investigates the structure response and failure mechanism of a micro-electromechanical system (MEMS) vibrating ring gyroscope (VRG) under high-g shock. A principle for categorizing different types of loading shock according to the VRG’s frequency scales is devised. A dynamic response model of the VRG structure to diverse types of shock based on quasi-static, vibration and elastic wave theories is established. The failure mechanism of the VRG is analyzed for the typical failures of MEMS devices, including pull-in, fracture and delamination. Based on the established dynamic response model, the inertial displacement that leads to pull-in failure, the fracture-sensitive positions of the VRG structure under different shocks, and the physical model of delamination failure under shock are derived. Through the shock experiments, the critical amplitudes of the loading shock that cause the failures of the VRG are obtained, the reliability model of the VRG under high-g shock is established, and the model established by analyzing the failure mechanism of the VRG is verified from the results of the experiments.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10506452/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10506452/</guid>
</item>
<item>
<title>GAN-CNN-Based Moving Target Detector for Airborne Radar Systems</title>
<description><p><span><big>GAN-CNN-Based Moving Target Detector for Airborne Radar Systems</big></span><br></p><p><span><small><i>Yangguang Zhao; Taohan Sun; Jiawei Zhang; Meiguo Gao</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3397731</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Conventional phased-array airborne radar accomplishes target detection tasks through adaptive filtering and constant false alarm techniques. However, it encounters significant performance degradation in scenarios characterized by nonhomogeneous clutter and low signal-to-noise ratio (SNR). This paper introduces an airborne radar target detection (ARTD) scheme based on a Generative Adversarial Network (GAN) and Convolutional Neural Network (CNN), capitalizing on the distributional properties of targets and clutter in the Beam-Doppler (BD) domain. In contrast to conventional ARTD methods that rely on clutter suppression processing, the proposed approach centers on target and clutter classification. Specifically, the proposed scheme comprises two modules: the reconstruction network and the detection network. The reconstruction module employs an adversarial learning mechanism to reconstruct and denoise the BD-maps with high resolution, while the detection module extracts target information from the reconstructed BD-maps. Results from experimental assessments involving both simulated and actual radar echoes indicate that the proposed GAN-CNN-based detector outperforms the comparison methods in terms of detection capability and accuracy. Moreover, it exhibits a reduced false alarm rate, particularly in scenarios with low SNR.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10529979/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10529979/</guid>
</item>
<item>
<title>Affinity-based capturing, release, and glycoprofiling of PSA cancer biomarker using miniaturized micropillar-based platform</title>
<description><p><span><big>Affinity-based capturing, release, and glycoprofiling of PSA cancer biomarker using miniaturized micropillar-based platform</big></span><br></p><p><span><small><i>Ahmad Usman; Ali Asghar Eftekhar; Ali Adibi</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3389992</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The work presents affinity-based biomolecular capturing and release employing a PDMS micropillar-based microfluidic chip. The microfluidic chip has been equipped with a micropillar-based capture chamber specifically optimized to provide enhanced surface area for biomolecular capturing without having any fluid and material clogging. A novel multi-step surface functionalization immunoassay protocol to optimize on-the-flow surface functionalization and capturing efficiency of the biomolecules has been developed and demonstrated. The novel immunoassay protocol involved conjugation of the anti-PSA IgG antibodies with a photo-cleavable (PC)-biotin-PEG3-NHS-ester, capturing of free-PSA (f-PSA) cancer biomarker, the capability of multiplexed glycoprofiling for elucidating the PTMs of the f-PSA using the Sambucus Nigra Lectin (SNA) and Maackia Amurensis Lectin II (MAA-II) lectins, and on-demand release of the captured f-PSA biomarkers. The work provides a proof-of-concept demonstration of affinity-based capturing, multiplexing, glycoprofiling, and release of f-PSA biomarkers with a minimum limit of detection of ~10 pg/ml and a limit of quantification of ~20 pg/ml using a low-cost, disposable microfluidic chip.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10506344/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10506344/</guid>
</item>
<item>
<title>Design of MTM-based Multi-band Micro-Biosensor in Terahertz region as perfect absorber for Early-Stage Leukemia Diagnosis with sensitivity 18626373 TH...</title>
<description><p><span><big>Design of MTM-based Multi-band Micro-Biosensor in Terahertz region as perfect absorber for Early-Stage Leukemia Diagnosis with sensitivity 18626373 THz/RIU</big></span><br></p><p><span><small><i>Musa N. Hamza; Mohammad Tariqul Islam</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383522</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this article, a novel highly sensitive biosensor based on perfect metamaterial absorbers is presented. A detailed study is presented for several different models, different types of substrate materials, different types of resonant materials and substrate thicknesses showing very good sensitivity to any changes. The proposed biosensor exhibits high sensitivity to different polarization and incidence angles, ensuring its performance in terms of signal-to-noise ratio. The proposed biosensor was carefully compared with previously designed sensors and biosensors. The proposed biosensor exhibits amazing sensitivity, such as a quality factor of 41.1, a figure of merit (FOM) of 172466416 RIU-1, and a sensitivity (S) of 18626373 THz/RIU. The high sensitivity of the biosensor allows early detection of leukemia, demonstrating significant differences between leukemia and normal blood. Another interesting result of this article is the use of terahertz waves for imaging. Microwave imaging is performed for electric fields, magnetic fields, and power.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10494209/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10494209/</guid>
</item>
<item>
<title>LEPS: A lightweight and effective single-stage detector for pothole segmentation</title>
<description><p><span><big>LEPS: A lightweight and effective single-stage detector for pothole segmentation</big></span><br></p><p><span><small><i>Xiaoning Huang; Jintao Cheng; Qiuchi Xiang; Jun Dong; Jin Wu; Rui Fan; Xiaoyu Tang</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2023.3330335</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Currently, the problem of potholes on urban roads is becoming increasingly severe. Identifying and locating road potholes promptly has become a major challenge for urban construction. Therefore, we proposed a lightweight and effective instance segmentation method called LEPS (Lightweight and Effective Pothole Segmentation Detector) for road pothole detection. To extract the image edge information and gradient information of potholes from the feature map more efficiently, we proposed a module that performs the convolutional super-position supplemented with a convolutional kernel to enhance spatial details for the backbone. We have designed a novel module applied to the Neck layer, improving the detection performance while reducing the parameters. To enable accurate segmentation of fine-grained features, we optimized the ProtoNet, which enables the segmentation head to generate high-quality masks for more accurate prediction. We have fully demonstrated the effectiveness of the method through a large number of comparative experiments. Our detector has excellent performance on two authoritative example datasets, URTIRI and COCO, and can successfully be applied to visual sensors to accurately detect and segment road potholes in real environments, accurately LEPS reached 0.892 and 0.648 in terms of Mask for AP50 and AP50:95, which is improved by 4.6% and 20.6% compared to the original model. These results demonstrate its strong competitiveness when compared to other models. Comprehensively, LEPS improves detection accuracy while maintaining lightweight, which allows the model to meet the practical application requirements of edge computing devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10315064/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10315064/</guid>
</item>
<item>
<title>Sparse Wearable Sonomyography Sensor-based Proprioceptive Proportional Control Across Multiple Gestures</title>
<description><p><span><big>Sparse Wearable Sonomyography Sensor-based Proprioceptive Proportional Control Across Multiple Gestures</big></span><br></p><p><span><small><i>Anne Tryphosa Kamatham; Kavita Sharma; Srikumar Venkataraman; Biswarup Mukherjee</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3394029</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Sonomyography (SMG) is a non-invasive technique that uses ultrasound imaging to detect the dynamic activity of muscles. Wearable SMG systems have recently gained popularity due to their potential as human-computer interfaces for their superior performance compared to conventional methods. This paper demonstrates real-time positional proportional control of multiple gestures using a multiplexed 8-channel wearable SMG system. The amplitude-mode ultrasound signals from the SMG system were utilized to detect muscle activity from the forearm of 8 healthy individuals. The derived SMG signals were used to control the on-screen position of the cursor. A target achievement task was performed to analyze the performance of our SMG-based human-machine interface. Our wearable SMG system provided accurate, stable, and intuitive control in real-time by achieving an average success rate greater than 80% with all gestures. Furthermore, the wearable SMG system’s abilities to detect volitional movement and decode movement kinematic information from SMG trajectories using standard performance metrics were evaluated. Our results provide insights to validate SMG as an intuitive human-machine interface.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10522599/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10522599/</guid>
</item>
<item>
<title>Design of Ag@ZnO Microreactor with High Sensitivity and Selectivity for Triethylamine Detection</title>
<description><p><span><big>Design of Ag@ZnO Microreactor with High Sensitivity and Selectivity for Triethylamine Detection</big></span><br></p><p><span><small><i>Feifan Fan; Junyang Hao; Yuefeng Gu; Fangfang Xue; Zhicheng Zhu; Tiancheng Wu; Qiuhong Li</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3386602</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Triethylamine (TEA) is a crucial raw material in industrial processes. However, exposure to this gas poses significant health risks to humans. Therefore, the development of sensors that can effectively detect TEA with high performance is crucial. In this study, we synthesized Ag@ZnO hollow spheres using microwave hydrothermal and template methods. The Ag@ZnO microreactor structure was formed by decorating Ag particles on the inner wall of ZnO hollow spheres. This unique structure effectively extended the contact time between Ag particles and gas molecules, and noteworthily enhanced the sensing properties. The gas sensing measurements demonstrated that Ag@ZnO sensor exhibited ultra-high selectivity and superior sensitivity to TEA. The response value to 100 ppm triethylamine was 763 at the optimal temperature of 200 °C, which was approximately 34 times higher than pure ZnO hollow spheres, and 3 times higher than that of ZnO@Ag where Ag particles were loaded on the outer wall of ZnO hollow spheres.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10499990/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10499990/</guid>
</item>
<item>
<title>UMHE: Unsupervised Multispectral Homography Estimation</title>
<description><p><span><big>UMHE: Unsupervised Multispectral Homography Estimation</big></span><br></p><p><span><small><i>Jeongmin Shin; Jiwon Kim; Seokjun Kwon; Namil Kim; Soonmin Hwang; Yukyung Choi</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383453</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Multispectral image alignment plays a crucial role in exploiting complementary information between different spectral images. Homography-based image alignment can be a practical solution considering a tradeoff between runtime and accuracy. Existing methods, however, have difficulty with multispectral images due to the additional spectral gap or require expensive human labels to train models. To solve these problems, this paper presents a comprehensive study on multispectral homography estimation in an unsupervised learning manner. We propose a curriculum data augmentation, an effective solution for models learning spectrum-agnostic representation by providing diverse input pairs. We also propose to use the phase congruency loss that explicitly calculates the reconstruction between images based on low-level structural information in the frequency domain. To encourage multispectral alignment research, we introduce a novel FLIR corresponding dataset that has manually labeled local correspondences between multispectral images. Our model achieves state-of-the-art alignment performance on the proposed FLIR correspondence dataset among supervised and unsupervised methods while running at
151 FPS
. Furthermore, our model shows good generalization ability on the M3FD dataset without finetuning.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10494213/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10494213/</guid>
</item>
<item>
<title>Characterization of Bubble Size of Gas-Liquid Two-Phase Flow Using Ultrasonic Signals</title>
<description><p><span><big>Characterization of Bubble Size of Gas-Liquid Two-Phase Flow Using Ultrasonic Signals</big></span><br></p><p><span><small><i>Ningde Jin; Huiwen Lu; Jiachen Zhang; Weikai Ren</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3388012</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Refining the bubble size distribution characteristics in gas-liquid two-phase flow can enhance the precision of detecting flow parameters. In this study, gas-liquid two-phase bubble flow experiments were first carried out in a vertical upward pipeline with an internal diameter of 20 mm. A high-speed camera was used to take instantaneous snapshots of the bubbles flowing through the fluid sampler, capturing images of the size distribution of the dispersed bubbles within the fluid sampler at different flow conditions. Ultrasonic attenuation information when the bubbles flowing through the measured area was obtained using the pulsed transmission ultrasonic sensor. The magnitude and sign fractal scaling exponents of ultrasonic attenuation signals were obtained by the detrended fluctuation analysis (DFA) method, which exhibits a strong correlation with the bubble diameter. The results show that the nonlinear dynamic characteristics of bubble size evolution in gas-liquid two-phase bubble flow can be effectively characterized by ultrasonic measurement signals, which provides a crucial foundation for different bubble size of void fraction prediction model of gas-liquid two-phase flow.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10505146/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10505146/</guid>
</item>
<item>
<title>A Stretchable and Wearable Ultrasonic Transducer Array for Bladder Volume Monitoring Application</title>
<description><p><span><big>A Stretchable and Wearable Ultrasonic Transducer Array for Bladder Volume Monitoring Application</big></span><br></p><p><span><small><i>Cong Pu; Ben Fu; Lehang Guo; Huixiong Xu; Chang Peng</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3382244</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Monitoring bladder volume is an essential application for patients with voiding dysfunction. Current wearable ultrasonic bladder monitors are rigid, bulky and cannot conform to human skin. This study presents a stretchable and wearable ultrasonic transducer array that can both conform to non-planar skin surfaces and continuously monitor bladder volume. The wearable transducer array mainly consists of 4 × 4 piezoelectric transducer elements, stretchable serpentine-based electrodes for electrical interconnection, electrically conductive adhesive as both matching and backing layers, and silicone elastomer for encapsulation. The transducer array has a center frequency of 3.9 MHz, -6 dB acoustic bandwidth of 57.6%, and up to 40% elastic stretchability. The volume of balloon-bladder phantom was estimated by a least square ellipsoid fitting method. With the true volume of balloon-bladder phantoms ranging from 10 mL to 405 mL, the proposed transducer array has a mean absolute percentage error of 9.4%, a mean absolute error of 11.9 mL and a coefficient of determination of 0.975, which demonstrates the capability of the proposed stretchable and wearable ultrasonic transducer array for bladder volume monitoring application.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10489841/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10489841/</guid>
</item>
<item>
<title>ASMGCN: Attention-Based Semantic-Guided Multi-Stream Graph Convolution Network for Skeleton Action Recognition</title>
<description><p><span><big>ASMGCN: Attention-Based Semantic-Guided Multi-Stream Graph Convolution Network for Skeleton Action Recognition</big></span><br></p><p><span><small><i>Moyan Zhang; Zhenzhen Quan; Wei Wang; Zhe Chen; Xiaoshan Guo; Yujun Li</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3388154</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In recent years, the field of action recognition using spatio-temporal graph convolution models for human skeletal data has made significant progress. However, current methodologies tend to prioritize spatial graph convolution, which leads to an underutilization of valuable information present in skeletal data. It limits the model’s ability to effectively capture complex data patterns, especially in time series data, ultimately impacting recognition accuracy significantly. To address the above issues, this paper introduces an Attention-based Semantic-guided Multi-stream Graph Convolution Network (ASMGCN), which can extract the deep features in skeletal data more fully. Specifically, ASMGCN incorporates a novel temporal convolutional module featuring an attention mechanism and a multiscale residual network, which can dynamically adjust the weights between skeleton graphs at different time points, enabling better capture of relational features. In addition, semantic information is introduced into the loss function, enhancing the model’s ability to distinguish similar actions. Furthermore, the coordinate information of different joints within the same frame is explored to generate new relative position features known as centripetal and centrifugal streams based on the center of gravity. These features are integrated with the original position and motion features of skeleton, including joints and bones, enriching the inputs to the GCN network. Experimental results on the NW-UCLA, NTU RGB+D (NTU60) and NTU RGB+D 120 (NTU120) datasets demonstrate that ASMGCN outperforms other state-of-the-art (SOTA) Human Action Recognition (HAR) methods, signifying its potential in advancing the field of action recognition using skeletal data.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10505150/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10505150/</guid>
</item>
<item>
<title>Partial discharge positioning method in air insulated substation with vehicle mounted UHF sensor array based on RSSI and regularization</title>
<description><p><span><big>Partial discharge positioning method in air insulated substation with vehicle mounted UHF sensor array based on RSSI and regularization</big></span><br></p><p><span><small><i>Ruoyu Wang; Baiqiang Yin; Lifen Yuan; Shudong Wang; Chengwei Ding; Xiaodong Lv</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3387665</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Monitoring technology for partial discharges (PD) in Air-Insulated Substations (AIS) has undergone development and practical application. Partial discharge positioning based on Received Signal Strength Indication (RSSI) faces a variety of interference factors, including complex environments, multipath effects of signals, expensive fees and huge size of monitoring equipment. These factors jointly lead to the decline of PD positioning accuracy. In order to solve these problems, this paper proposes a substation patrol vehicle positioning technology based on RSSI and vehicle-mounted UHF sensor array. Initially, at a specific distance, multiple measurements, Gaussian filtering, and averaging are performed on a specific PD signal, to determine the RSSI at this distance, and consequently identify the parameters within the Shadowing model. Subsequently, a centralized and balanced preprocessing of the positioning equation is to be performed. Then, derive regularization parameters through the L-curve method and employ the Tikhonov regularization technique to determine the coordinates of the PD source. Simulations and experiments indicate that under simulated disturbances of 0-15% within a space of 30m × 30m × 20m, the proposed method maintains a positioning error within 3.6m. The method is cost-effective and operationally straightforward.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10504736/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10504736/</guid>
</item>
<item>
<title>Finger-tapping Motion Recognition Based on Skin Surface Deformation Using Wrist-mounted Piezoelectric Film Sensors</title>
<description><p><span><big>Finger-tapping Motion Recognition Based on Skin Surface Deformation Using Wrist-mounted Piezoelectric Film Sensors</big></span><br></p><p><span><small><i>Shumma Jomyo; Akira Furui; Tatsuhiko Matsumoto; Tomomi Tsunoda; Toshio Tsuji</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3386333</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>The miniaturization of computers has led to the development of wearable devices in the form of watches and eyeglasses. Consequently, the narrower screen size has raised the issue of operability for text input. This problem can be resolved using external input devices, such as physical keyboards. However, this can impair portability and accessibility. This study proposes a finger-tapping motion recognition system using wrist-mounted piezoelectric film sensors to realize an input interface with high wearability and not limited by screen size. In the proposed system, biodegradable piezoelectric film sensors, which are highly compatible with biological signal measurement, are attached to the palmar and dorsal surfaces of the wrist to measure minute skin surface deformation during tapping. The system detects the occurrence of tapping movements for each finger by preprocessing the measured signals and calculating the total activity of all channels. It also recognizes the type of finger movement based on machine learning. In the experiment, we measured ten different signals, including five-finger flexion and extension, for eleven subjects, to evaluate the effectiveness of the proposed method. According to the experimental results, tapping recognition accuracy for time-series data was 77.5%, assuming character input. In addition, the time difference between the detected and actual taps was approximately 50 ms on average. Therefore, the proposed method can be utilized as an input interface for wristband-type wearable devices.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10500304/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10500304/</guid>
</item>
<item>
<title>CLORP: Cross-Layer Opportunistic Routing Protocol for Underwater Sensor Networks Based on Multi-Agent Reinforcement Learning</title>
<description><p><span><big>CLORP: Cross-Layer Opportunistic Routing Protocol for Underwater Sensor Networks Based on Multi-Agent Reinforcement Learning</big></span><br></p><p><span><small><i>Shuai Liu; Jingjing Wang; Wei Shi; Guangjie Han; Shefeng Yan; Jiaheng Li</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383035</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>With the development of the Internet of Underwater Things (IoUT), both academia and industry have significant emphasized underwater wireless sensor networks (UWSNs). To address the issues of slow convergence, high latency, and limited energy in existing intelligent routing protocols in UWSNs, a cross-layer opportunistic routing protocol (CLORP) for underwater sensor networks based on multi-agent reinforcement learning (MARL) is proposed in this paper. First, CLORP combines the decision-making capability of multi-agent reinforcement learning with the idea of opportunistic routing to sequentially select a set of neighbors with larger values as potential forwarding nodes, thereby increasing the packet transmission success rate. Second, in the design of the MARL reward function, two reward functions for successful and unsuccessful packet transmission are designed jointly with cross-layer information to improve the routing protocol’s performance. Finally, two algorithmic optimization strategies, adaptive learning rate and Q-value initialization based on location and number of neighbors, are proposed to facilitate the faster adaptation of agents to the dynamic changes of network topology and accelerate CLORP convergence. The experimental results demonstrate that CLORP can increase algorithm convergence speed by 13.2%, reduce network energy consumption by 25%, and decrease network latency by 31.2%.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10492498/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10492498/</guid>
</item>
<item>
<title>Three-Dimensional Array SAR Sparse Imaging Based on Hybrid Regularization</title>
<description><p><span><big>Three-Dimensional Array SAR Sparse Imaging Based on Hybrid Regularization</big></span><br></p><p><span><small><i>Jing Gao; Yangyang Wang; Jinjie Yao; Xu Zhan; Guohao Sun; Jiansheng Bai</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3386901</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>With the development and maturity of compressed sensing theories, sparse signal processing has been widely applied to synthetic aperture radar (SAR) imaging. L
1
regularization, as an effective sparse reconstruction model, can enhance sparsity of the scene. However, due to the convexity of the L
1
regularization function, the L
1
-based sparse reconstruction frequently introduces bias estimation, resulting in an underestimation of target amplitudes. Furthermore, the 3-D SAR imaging scene has diverse features, which are difficult to characterize solely with the L
1
regularization. Therefore, in this paper, we propose a novel imaging framework for 3-D array SAR imaging, which combines hybrid regularization function and an improved variable splitting with the method of multipliers (IVSMM). Firstly, we present the hybrid regularization function, which combines smoothly clipped absolute deviation (SCAD) and high-dimensional total variation (HTV) to reduce bias effects and preserve the regional features of the target. Then, we present IVSMM to solve the optimization problem of the hybrid regularization, effectively reducing the computational complexity required for 3-D imaging. Finally, the excellent reconstruction performance of the proposed method is validated through a series of simulations and experimental data.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10502305/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10502305/</guid>
</item>
<item>
<title>Deep Learning Enabled Two Directional Stretchable Strain Sensor B ased on A Single Multimode Fiber</title>
<description><p><span><big>Deep Learning Enabled Two Directional Stretchable Strain Sensor B ased on A Single Multimode Fiber</big></span><br></p><p><span><small><i>Juan Kang; Sijin He; Osamah Alsalman; Xiao Liu; Chen Zhu</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3395613</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In this work, we have demonstrated a specklegram proprioceptive sensor capable of sensing large tensile strains based on a single silica multimode fiber (MMF) embedded in a soft silicone pad. Taking advantage of the rich information contained in the output speckle patterns from the MMF and a convolutional neural network-based regression demodulation algorithm, the sensor shows an extended strain sensing range up to 10%, and the root mean squared error is determined to be less than 0.05%. Additionally, by integrating with a classification model, the sensor is also able to distinguish the direction of the applied tensile strains, i.e., the axial or the longitudinal directions, with an accuracy of 100%. The proposed sensor has the advantages of low cost, ease of fabrication, and large strain capability, and could find wide applications in the field of soft robotics.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10522602/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10522602/</guid>
</item>
<item>
<title>Water Quality Assessment Tool for On-site Water Quality Monitoring</title>
<description><p><span><big>Water Quality Assessment Tool for On-site Water Quality Monitoring</big></span><br></p><p><span><small><i>Segun O. Olatinwo; Trudi H. Joubert; Damilola D. Olatinwo</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3383887</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>Reliable water quality monitoring requires on-site processing and assessment of water quality data in near real-time. This helps to promptly detect changes in water quality, prevent biodiversity loss, safeguard the health and well-being of communities, and mitigate agricultural problems. To this end, we proposed a Highway-Bidirectional Long Short-term Memory (Highway-BiLSTM)-based water quality classification tool for potential integration into an edge-enabled water quality monitoring system to facilitate on-site water quality classification. The performance of the proposed classifier was validated by comparing it with several baseline water quality classifiers. The proposed classifier outperformed the baseline water classifier in terms of accuracy, precision, sensitivity, F1-score, and confusion matrix. Specifically, the proposed water classifier surpassed the random forest (RF) classifier with 2% accuracy, precision, sensitivity, and F1-score. Moreover, the proposed classifier achieved an increase of 4% in accuracy, precision, sensitivity, and F1-score for classifying water quality compared with the Gradient Boosting classifier. Additionally, the proposed method has 4% increase in accuracy, sensitivity, F1-score, and 3% increase in precision compared to the support vector machine (SVM) water quality classifier. The proposed method outperformed the artificial neural network (ANN) classifier by 1% accuracy, precision, sensitivity, and F1-score. Finally, the proposed method demonstrated rare errors in accurately classifying complex water quality samples. These findings suggest that our proposed method could be used to effectively classify water quality to aid accurate decision making and environmental management.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/10499201/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/10499201/</guid>
</item>
<item>
<title>An Integrated Self-Powered Wheel-Speed Monitoring System utilizing Piezoelectric-Electromagnetic−Triboelectric Hybrid Generator</title>
<description><p><span><big>An Integrated Self-Powered Wheel-Speed Monitoring System utilizing Piezoelectric-Electromagnetic−Triboelectric Hybrid Generator</big></span><br></p><p><span><small><i>Xiaohui Lu; Zhaojie Zhang; Wentao Ruan; Hengyu Li; Da Zhao; Kuankuan Wang; Bingzhao Gao; Baichuan Leng; Xin Yu; Bangcheng Zhang; Tinghai Cheng</i></small></span><br><span><small><i>https://doi.org/10.1109/JSEN.2024.3384569</i></small></span><br><span><small><i>Volume 24</i></small></span><br></p><p><span>In the intelligent vehicle systems, vehicle speed information is collected and transmitted by wheel speed sensors, which is related to the safety, energy efficiency, and comfort of the vehicle. This work proposes an integrated self-powered wheel-speed monitoring system (SWMS), which can achieve real-time sensing and wireless transmission multifunction by employing an integrated Piezoelectric-Electromagnetic−Triboelectric hybrid generator (PETHG) for harvesting the mechanical energy from rotating wheels. In particular, the triboelectric generation unit in SWMS includes an energy-channel (E-TENG) and a sensing-channel (S-TENG), the E-TENG was used for harvesting the rotational energy of the wheel, and the S-TENG was used for monitoring the rotation speed of the wheel. In experiment test, the output performance of the hybrid generator was assessed using a rotating electrical motor to simulate different wheel speed inputs. At 600 rpm, the instantaneous power of the hybrid generato... |
http://localhost:1200/ieee/author/37281702200 - Success ✔️<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
<channel>
<title>Ana Garcia Armada on IEEE Xplore</title>
<link>https://ieeexplore.ieee.org/author/37281702200</link>
<atom:link href="http://localhost:1200/ieee/author/37281702200" rel="self" type="application/rss+xml"></atom:link>
<description>Ana Garcia Armada (Fellow, IEEE) is a Professor at Universidad Carlos III de Madrid (UC3M), Spain. She has been a Visiting Scholar at Stanford University, Bell Labs, and University of Southampton. She has published more than 250 papers in conferences and journals. She holds five patents. Her research mainly focuses on signal processing applied to wireless communications. She has been a member of the organizing committee of several conferences, including IEEE Globecom 2021 as the General Chair. She has received several awards from UC3M, the third place Bell Labs Prize 2014, the Outstanding Service Award from the IEEE ComSoc Signal Processing and Communications Electronics Technical Committee and the Outstanding Service Award from the IEEE ComSoc Women in Communications Engineering Standing Committee. She has received the IEEE ComSoc/KICS Exemplary Global Service Award in 2022. She serves on the editorial boards for IEEE Transactions on Communications, IEEE Open Journal of the Communications Society, and ITU Journal on Future and Evolving Technologies. - Made with love by RSSHub(https://github.com/DIYgod/RSSHub)</description>
<generator>RSSHub</generator>
<webMaster>i@diygod.me (DIYgod)</webMaster>
<language>en</language>
<lastBuildDate>Tue, 14 May 2024 10:45:07 GMT</lastBuildDate>
<ttl>5</ttl>
<item>
<title>Understanding the effects of phase noise in orthogonal frequency division multiplexing (OFDM)</title>
<description><p><span><big>Understanding the effects of phase noise in orthogonal frequency division multiplexing (OFDM)</big></span><br></p><p><span><small><i>A. Garcia Armada</i></small></span><br><span><small><i>https://doi.org/10.1109/11.948268</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>Phase noise must be carefully considered when designing an OFDM-based communication system since an accurate prediction of the tolerable phase noise can allow the system and RF engineers to relax specifications. This paper analyzes the performance of OFDM systems under phase noise and its dependence on the number of sub-carriers both in the presence and absence of a phase correction mechanism. Besides some practical results are provided so as to give some insight into the phase noise spectral specifications that should be required of the local oscillator.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/948268/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/948268/</guid>
</item>
<item>
<title>Phase noise and sub-carrier spacing effects on the performance of an OFDM communication system</title>
<description><p><span><big>Phase noise and sub-carrier spacing effects on the performance of an OFDM communication system</big></span><br></p><p><span><small><i>A.G. Armada; M. Calvo</i></small></span><br><span><small><i>https://doi.org/10.1109/4234.658613</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>This letter analyzes the phase noise effects on an orthogonal frequency division multiplexing (OFDM) signal and its dependence with the sub-carrier spacing. Pilot-based channel estimation, which has been suggested as a means of combating the channel effects, can also correct the phase noise effects under some circumstances, which are investigated.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/658613/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/658613/</guid>
</item>
<item>
<title>Fair Design of Plug-in Electric Vehicles Aggregator for V2G Regulation</title>
<description><p><span><big>Fair Design of Plug-in Electric Vehicles Aggregator for V2G Regulation</big></span><br></p><p><span><small><i>José Joaquín Escudero-Garzas; Ana Garcia-Armada; Gonzalo Seco-Granados</i></small></span><br><span><small><i>https://doi.org/10.1109/TVT.2012.2212218</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>Plug-in electric vehicles (PEVs) have recently attracted much attention due to their potential to reduce
$\hbox{CO}_{2}$
emissions and transportation costs and can be grouped into entities (aggregators) to provide ancillary services such as frequency regulation. In this paper, the application of aggregators to frequency regulation by making fair use of their energy storage capacity is addressed. When the power grid requires frequency regulation service to the aggregator to adjust the grid frequency, the PEVs participating in providing the service can either draw energy (as it is usually done to charge the vehicle) or deliver energy to the grid by means of the vehicle-to-grid (V2G) interface. Under the general framework of optimizing the aggregator profit, different methods, such as state-dependent allocation and the water-filling approach, are proposed to achieve a final state of charge (SOC) of the PEVs that satisfy the desired fairness criteria once the regulation service has been carried out.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/6262497/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/6262497/</guid>
</item>
<item>
<title>Task Scheduling for Mobile Edge Computing Using Genetic Algorithm and Conflict Graphs</title>
<description><p><span><big>Task Scheduling for Mobile Edge Computing Using Genetic Algorithm and Conflict Graphs</big></span><br></p><p><span><small><i>Ahmed A. Al-Habob; Octavia A. Dobre; Ana García Armada; Sami Muhaidat</i></small></span><br><span><small><i>https://doi.org/10.1109/TVT.2020.2995146</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>In this paper, we consider parallel and sequential task offloading to multiple mobile edge computing servers. The task consists of a set of inter-dependent sub-tasks, which are scheduled to servers to minimize both offloading latency and failure probability. Two algorithms are proposed to solve the scheduling problem, which are based on genetic algorithm and conflict graph models, respectively. Simulation results show that these algorithms provide performance close to the optimal solution, which is obtained through exhaustive search. Furthermore, although parallel offloading uses orthogonal channels, results demonstrate that the sequential offloading yields a reduced offloading failure probability when compared to the parallel offloading. On the other hand, parallel offloading provides less latency. However, as the dependency among sub-tasks increases, the latency gap between parallel and sequential schemes decreases.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/9094341/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/9094341/</guid>
</item>
<item>
<title>SNR gap approximation for M-PSK-Based bit loading</title>
<description><p><span><big>SNR gap approximation for M-PSK-Based bit loading</big></span><br></p><p><span><small><i>Ana Garcia-Armada</i></small></span><br><span><small><i>https://doi.org/10.1109/TWC.2006.1576527</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>Adaptive OFDM has the potential of providing bandwidth-efficient communications in hostile propagation environments. Currently, bit loading algorithms use M-ary quadrature amplitude modulation of the OFDM sub-carriers, where the number of bits per symbol modulating each of them is obtained in order to maximize the performance. SNR gap approximation for M-QAM signaling makes the algorithms simpler to implement. However, in some circumstances it may be preferable to use. M-ary phase shift keying. In this letter an approximation is derived for M-PSK similar to the SNR gap of M-QAM so that bit loading algorithms can be extended to this type of modulation. In addition, the performance obtained when using M-PSK is compared to that of M-QAM in a practical situation.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/1576527/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/1576527/</guid>
</item>
<item>
<title>OFDM performance in amplifier nonlinearity</title>
<description><p><span><big>OFDM performance in amplifier nonlinearity</big></span><br></p><p><span><small><i>S. Merchan; A.G. Armada; J.L. Garcia</i></small></span><br><span><small><i>https://doi.org/10.1109/11.713060</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>The activities of the current European RACE and ACTS projects have led to an increasing interest in OFDM (orthogonal frequency division multiplexing) as a means of combating impulsive noise and multipath effects and making fuller use of the available bandwidth of the system. This paper analyses the performance of OFDM signals in amplifier nonlinearity. In particular, bit error rate (BER) degradation as a result of amplitude limiting or clipping are analysed. In the presence of both nonlinear distortion and additive Gaussian noise, optimized output power back off is provided to balance the requirements of minimum BER and power amplifier efficiency. For this purpose, an OFDM system has been built using the SPW (Signal Processing Worksystem) simulator.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/713060/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/713060/</guid>
</item>
<item>
<title>Design and implementation of synchronization and AGC for OFDM-based WLAN receivers</title>
<description><p><span><big>Design and implementation of synchronization and AGC for OFDM-based WLAN receivers</big></span><br></p><p><span><small><i>V.P.G. Jimenez; M.J.F.-G. Garcia; F.J.G. Serrano; A.G. Armada</i></small></span><br><span><small><i>https://doi.org/10.1109/TCE.2004.1362493</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>An efficient implementation of several tasks at the receiver becomes crucial in OFDM-based high-speed WLAN systems, such as automatic gain control, time and frequency synchronization, and offset tracking. This paper deals with fixed point constraints and accuracy requirements for implementation of those algorithms. Also, a complete set of thresholds for the practical implementation of time and frequency synchronization sub-blocks is obtained. Moreover, a technique to mitigate the remaining frequency offset after coarse acquisition is proposed, yielding a good trade-off between performance and complexity. Finally, we propose the implementation of a simple and effective automatic gain control procedure.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/1362493/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/1362493/</guid>
</item>
<item>
<title>New Technologies and Trends for Next Generation Mobile Broadcasting Services</title>
<description><p><span><big>New Technologies and Trends for Next Generation Mobile Broadcasting Services</big></span><br></p><p><span><small><i>Alejandro de la Fuente; Raquel Perez Leal; Ana Garcia Armada</i></small></span><br><span><small><i>https://doi.org/10.1109/MCOM.2016.1600216RP</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>It is expected that by the year 2020, video services will account for more than 70 percent of mobile traffic. It is worth noting that broadcasting is a mechanism that efficiently delivers the same content to many users, not only focusing on venue casting, but also distributing many other media such as software updates and breaking news. Although broadcasting services are available in LTE and LTE-A networks, new improvements are needed in some areas to handle the demands expected in the near future. In this article we review the actual situation and some of the techniques that will make the broadcast service more dynamic and scalable, meeting the demands of its evolution toward the next generation. Resource allocation techniques for broadcast/multicast services, integration with new waveforms in 5th generation mobile communications (5G), initiatives for spectrum sharing and aggregation, or the deployment of small cells placed together with the existing macro cells, are some enhancements that are examined in detail, providing directions for further development. With this evolution, 5G broadcasting will be a driver to achieve the spectral efficiency needed for the 1000 times traffic growth that is expected in upcoming years, leading to new applications in 5G networks that are specifically focused on mobile video services.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/7593432/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/7593432/</guid>
</item>
<item>
<title>Analysis of RIS-Based Terrestrial-FSO Link Over G-G Turbulence With Distance and Jitter Ratios</title>
<description><p><span><big>Analysis of RIS-Based Terrestrial-FSO Link Over G-G Turbulence With Distance and Jitter Ratios</big></span><br></p><p><span><small><i>Alain R. Ndjiongue; Telex. M. N. Ngatched; Octavia A. Dobre; Ana Garcia Armada; Harald Haas</i></small></span><br><span><small><i>https://doi.org/10.1109/JLT.2021.3108532</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>One of the main problems faced by communication systems is the presence of skip-zones in the targeted areas. With the deployment of the fifth-generation mobile network, solutions are proposed to solve the signal loss due to obstruction by buildings, mountains, and atmospheric or weather conditions. Among these solutions, reconfigurable intelligent surfaces (RIS), which are newly proposed modules, may be exploited to reflect the incident signal in the direction of dead zones, increase communication coverage, and make the channel smarter and controllable. This paper tackles the skip-zone problem in terrestrial free-space optical (T-FSO) systems using a single-element RIS. Considering link distances and jitter ratios at the RIS position, we carry out a performance analysis of RIS-aided T-FSO links affected by turbulence and pointing errors, for both heterodyne detection and intensity modulation-direct detection techniques. Turbulence is modeled using the Gamma-Gamma distribution. We analyze the model and provide exact closed-form expressions of the probability density function, cumulative distribution function, and moment generating function of the end-to-end signal-to-noise ratio. Capitalizing on these statistics, we evaluate the system performance through the outage probability, ergodic channel capacity, and average bit error rate for selected binary modulation schemes. Numerical results, validated through simulations, obtained for different RIS positions and link distances ratio values, reveal that RIS-based T-FSO performs better when the RIS module is located near the transmitter.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/9525295/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/9525295/</guid>
</item>
<item>
<title>VLC-Based Networking: Feasibility and Challenges</title>
<description><p><span><big>VLC-Based Networking: Feasibility and Challenges</big></span><br></p><p><span><small><i>Alain R. Ndjiongue; Telex M. N. Ngatched; Octavia A. Dobre; Ana G. Armada</i></small></span><br><span><small><i>https://doi.org/10.1109/MNET.001.1900428</i></small></span><br><span><small><i>Volume </i></small></span><br></p><p><span>VLC has emerged as a prominent technology to address the radio spectrum shortage. It is characterized by the unlicensed and unexploited high bandwidth, and provides the system with cost-effective advantages because of the dual-use of light bulbs for illumination and communication and the low complexity design. It is considered to be utilized in various telecommunication systems, including 5G, and represents the key technology for light-fidelity. To this end, VLC has to be integrated into the existing telecommunication networks. Therefore, its analysis as a network technology is momentous. In this article, we consider the feasibility of using VLC as a network technology and discuss the challenges related to the implementation of a VLC-based network, as well as the integration of VLC into existing conventional networks and its inclusion in standards.</span><br></p></description>
<link>https://ieeexplore.ieee.org/document/8970387/</link>
<guid isPermaLink="false">https://ieeexplore.ieee.org/document/8970387/</guid>
</item>
</channel>
</rss> |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Involved Issue / 该 PR 相关 Issue
Close #
Example for the Proposed Route(s) / 路由地址示例
New RSS Route Checklist / 新 RSS 路由检查表
PuppeteerNote / 说明
gottoofetchin/ieee/journal.tsand/ieee/earlyaccess.ts;routes/ieee/recent.tshas been deprecated.