/BenchmarkSourceSeparation

Project for a class validation at ATIAM master's degree at IRCAM on a survey of source separation on self-made studio recording.

Primary LanguageMATLAB

BenchmarkSourceSeparation

Project for a class validation at ATIAM master's degree at IRCAM on a survey of source separation on self-made studio recording.

This project is made with the supervision of Umut Şimşekli and Benoit Fabre and realised by

  • Valentin Bilot
  • Gabriel Dias Neto
  • Clement Le-Moine
  • Guilhem Marion
  • Yann Teytaut

Presentation

This project presents a quick survey of the problem of source separation. To this end, we will investigate three state-of-the-art signal processing algorithms and compare them for remixing purpose with two means: objective measures and perceptive experiment. We show that these results are not correlated and that low- separated sources with low artifacts tend to be better perceived than high-separated with artifacts sources.

Usage

Extraction

The script extract.py in audio/ extracts files in order to feed the experiment script.

Experiment

The script MainExperiment.py launch the experiment, you can change parameter by editing the file.

Usage: Python3 MainExperiment.py <ROOM or ANECHOIC>

Database

The scripts anal_ROOM.py and anal_ANECHOIC.py allow to fill the database and compute statistical tests onto.

Usage: anal_ROOM.py [options] <subject to fill>

Options:
  -h, --help            show this help message and exit
  -d NAME, --name=NAME  Name of the database.
  -p PPRINT, --print=PPRINT
                        1 if you want to print the data.

Licence

This project is entirely open source, and may by used for any purpose.