Yaafe - audio features extraction

Yaafe is an audio features extraction toolbox.

Easy to use
The user can easily declare the features to extract and their parameters in a text file. Features can be extracted in a batch mode, writing CSV or H5 files. The user can also extract features with Python or Matlab.
Efficient
Yaafe automatically identifies common intermediate representations (spectrum, envelope, autocorrelation, ...) and computes them only once. Extraction is processed block per block so that arbitrarily long files can be processed, and memory occupation is low.

If you wonder about the Yaafe acronym, it’s just Yet Another Audio Feature Extractor.

Project’s status

Current version is Yaafe 0.64, released on 2011-11-07.

Yaafe may evolve in future versions, but current code is pretty stable and feature computation is already reliable. Yaafe is already used in some Music Information Retrieval systems.

Yaafe provides:

  • a great collection of classical audio features, with transformations and temporal integration (see Available features documentation).
  • reading of WAV, OGG, MP3 (and others) audio file formats.
  • CSV and HDF5 output file formats
  • Python and Matlab bindings to extract features straight within your favorite environment.
  • automatic input samplerate conversion, using smarc.
  • an extensible framework that allow you to create your own feature library.
  • a C++ API to efficiently integrate Yaafe in your project

Download

Yaafe‘s source code is released under the terms of the GNU LGPLv3 License. Yaafe source code compiles on Linux and MacOsX platforms.

Download current release: yaafe-v0.64.tgz, see changelog.

View all releases: Sourceforge project’s website

Support

To get help with Yaafe, please use yaafe-users@lists.sourceforge.net mailing list. As project is still in progress, any feedback is welcome !

Cite Yaafe

If you want to cite Yaafe in a publication, please use the following reference:

YAAFE, an Easy to Use and Efficient Audio Feature Extraction Software, B.Mathieu, S.Essid, T.Fillon, J.Prado, G.Richard, proceedings of the 11th ISMIR conference, Utrecht, Netherlands, 2010.

Credits

Yaafe is developed at Telecom Paristech / AAO Team. It uses several great open-source projects like Eigen, Smarc, libsndfile, mpg123, HDF5.

Download Yaafe 0.64

Table Of Contents