BASS-dB: the Blind Audio Source Separation evaluation database
Summary of content
This database aims at helping evaluation of Blind Audio Source Separation algorithms.

As such, it contains a list of tables summarizing the separation results of existing algorithms on short selected mixtures, classified according to the number of channels and sources, the mixing setup (synthetic mixing with gains or filters or simulated real recording) and the amount of prior information. Clicking on the elements of a table gives details about algorithms (authors, program files), mixtures (sound files of mixtures and sources, mixing/recording parameters) and results (sound files of estimated sources and detailed performance measures).

The database also contains full multitrack recordings and mixing filters that can be used to produce new synthetic mixtures and test algorithms on larger datasets.

Finally, forms are provided to help you submit your own algorithm and separation results on the selected mixtures.