Example program of ICASSP2010 proceedings: MUSIC DEREVERBERATION USING HARMONIC STRUCTURE SOURCE MODEL AND WIENER FILTER.

setup

The author is using Mac OS 10.11, Python2.7 on Anaconda. Cython is also used for fast numerical computation.

1. Install homebrew (see: http://brew.sh/)

2. Install pyenv

   $ brew install pyenv
   $ echo 'export PYENV_ROOT="${HOME}/.pyenv"' >> ~/.bash_profile
   $ echo 'export PATH="${PYENV_ROOT}/bin:$PATH"' >> ~/.bash_profile
   $ echo 'eval "$(pyenv init -)"' >> ~/.bash_profile
   

3. Install Anaconda 2.7

   pyenv install anaconda2-4.1.1
   pyenv global anaconda2-4.1.1
   

4. Install CMake (see: https://cmake.org/download/)

usage

1. Build Cython C++ extension.

   cmake ./
   make
   

   (If you see errors, please check how CMakeLists.txt works)

2. Prepare sample *.wav file and execute entry python script.

   python prg/icassp2010.py sample.wav
   

   Two output files will be generated. Both are given by inverse STFT of dereverberated spectrogram.

   • sample_derev: dereverberated signal by using inverse linear filtering, which is mentioned at the last of section 2.2 in the proceedings
   • sample_derev_wiener: dereverberated signal by using the proposed Wiener filtering, which is mentioned at section 3.1 in the proceedings