Alexey Ozerov LTCI (TELECOM ParisTech & CNRS) - Signal and Image Processing Department alexey.ozerov@telecom-paristech.fr A. Ozerov & C. Févotte ALGORITHM: 1. 100 iterations of the Expectation-Maximization algorithm from [1] for joint estimation of instantaneous mixing matrix A and source Nonnegative Matrix Factorization (NMF) models (spectral patterns W and activation gains H) are performed. 2. The sources are recovered via Wiener filtering, as described in [1], given the estimated model (A, W and H). (Comment: 11 NMF components per source were used for speech sources and 5 components for music sources) INITIALIZATION: 1. An initial mixing matrix A_init is estimated as in [2]. 2. Initial source estimates s_init are obtained via the minimization of the l0 norm (given A_init) of the source STFTs [3]. 3. An initial NMF decomposition (W_init and H_init) is computed from the power spectrograms of initial source estimates s_init minimizing Kullback-Leibler divergence (see e.g., [4]). COMPUTATIONAL TIME Our Matlab implementation on 2.2 GHz CPU runs a. 410 seconds per test excerpt for speech sources b. 180 seconds per test excerpt for music sources REFERENCES: [1] A. Ozerov, and C. Févotte, \"Multichannel nonnegative matrix factorization in convolutive mixtures. With application to blind audio source separation\", In IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP\'09), 2009, (submitted). [2] S. Arberet, R. Gribonval, and F. Bimbot, \"A robust method to count and locate audio sources in a stereophonic linear instantaneous mixture\", In Proc. Int. Conf. on Independent Component Analysis and Blind Source Separation (ICA). (2006) 536-543 [3] E. Vincent, \"Complex nonconvex lp norm minimization for underdetermined source separation\", In Proc. Int. Conf. on Independent Component Analysis and Blind Source Separation (ICA). (2007) [4] C. Févotte, N. Bertin, and J.-L. Durrieu, \"Nonnegative matrix factorization with the Itakura-Saito divergence. with application to music analysis\", Tech. Rep., 2008, In press. http://www.tsi.enst.fr/~fevotte/TechRep/techrep08_is-nmf.pdf.