INTERSPEECH 2010

We propose a probabilistic factorial sparse coder model for single channel source separation in the magnitude spectrogram domain. The mixture spectrogram is assumed to be the sum of the sources, which are assumed to be generated framewise as the output of sparse coders plus noise. For dictionary training we use an algorithm which can be described as nonnegative matrix factorization with l0 sparseness constraints. In order to infer likely source spectrogram candidates, we approximate the intractable exact inference by maximizing the posterior over a plausible subset of solutions. We compare our system to the factorialmax vector quantization model, where the proposed method shows a superior performance in terms of signaltointerference ratio. Finally, the low computational requirements of the algorithm allows close to real time applications.
Bibliographic reference. Peharz, Robert / Stark, Michael / Pernkopf, Franz / Stylianou, Yannis (2010): "A factorial sparse coder model for single channel source separation", In INTERSPEECH2010, 386389.