Menu:

Source Separation in Stereo Mixtures


The following tables present the results obtained applying sound source separation methods to commercial stereo mixtures. The first table contains the results employing the method known as ADRess [1], and the second one employing a new method developed by us [2].

Three pop music recordings have been used. The first one is an instantaneous stereo mixture produced by us from the multitrack studio recording, employing only pan-pot techniques. The other two are commercial recordings ripped from the original CD (only 20 sec).

(*) May thanks to Indras and to Miguel Roma for providing us with the multitrack recording of the song "Es por ella".

Results with ADRess

Original Stereo Source 1 Source 2 Source 3 Source 4 Source 5
Es por ella Voice Guitar Drums Piano
2x4 Voice&Drums Guitar 1 Guitar 2
I Say a Little Prayer ... Lead Voice Chorus 1&2 Piano&Drums


Results with the New Method of Separation

Original Stereo Source 1 Source 2 Source 3 Source 4 Source 5
Es por ella Voice Guitar Drums Piano
2x4 Voice&Drums Guitar 1 Guitar 2
I Say a Little Prayer ... Lead Voice Chorus 1 Chorus 2 Drums Piano


[1] D. Barry, B. Lawlor, and E. Coyle, “Sound source separation: Azimuth discrimination and resynthesis”. Proceedings of the 7th Int. Conference on Digital Audio Effects (DAFTX 04), 2004.
[2] M. Cobos, J.J. Lopez, "An automatic sound source separation method for stereo mixtures combining azimuth and smart threshold", in preparation.