Participants: Alexey Ozerov and Emmanuel Vincent

Data processed: Only the mixtures recorded in domestic environment (from CHiME corpus)

Algorithm description:
   The algorithm is exactly as described in [2], which is a particular instance of general source separation framework described in [1].
   Moreover, the separated signals are exactly those we used in our submission to CHiME challenge.

Average running time: Our Matlab implementation on 2.2 GHz CPU, runs 2.6 minutes to process 1.5 second expert
                      (this time include noise model adaptation from neighboring background)

Strategy (from those in the task description):
   Strategy 4: process the whole 5 min recording using knowledge of the sentence positions

URL: http://www.irisa.fr/metiss/ozerov/sisec_2011/Ozerov_Two-channel_mixtures_of_speech_and_real-world_background_noise.zip

References:

[1] A. Ozerov, E. Vincent and F. Bimbot, "A general flexible framework for the handling of prior information in audio source separation,"  IEEE Trans. on Audio, Speech and Lang. Proc. (to appear)

[2] A. Ozerov and E. Vincent, "Using the FASST source separation toolbox for noise robust speech recognition,"  International Workshop on Machine Listening in Multisource Environments (CHiME 2011), pages 86-87, Florence, Italy, September, 2011.