Two time-frequency ratio-based blind source separation methods for time-delayed mixtures

Abstract

We propose two time-frequency (TF) blind source separar tion (BSS) methods suited to attenuated and delayed (AD) mixtures. They consist in identifying the columns of the (filtered permuted) mixing matrix in Constant-Time TF zones where they detect that a single source occurs, using Time-Frequency Ratios Of Mixtures (hence their name AD-TIFROM-CT). We thus identify columns of scale coefficients and time shifts. Unlike various previously reported TF-BSS approaches, these methods set very limited constraints on the source sparsity and overlap. They are especially suited to non-stationary sources. © Springer-Verlag Berlin Heidelberg 2006.