A cascade system for solving permutation and gain problems in frequency-domain BSS

Abstract

This paper presents a novel technique for separating convolutive mixtures of statistically independent non-Gaussian signals. The time-domain convolution is transformed into several instantaneous mixtures in the frequency-domain. The separation of these mixtures is performed in two steps. First, the instantaneous mixture at the frequency of reference is solved using JADE and the other mixtures are then separated using the Mean Square Error (MSE) criterion. We also present a novel method to select the frequency of reference. © Springer-Verlag Berlin Heidelberg 2009.