Extension of the ”TIme-Frequency Ratio Of Mixtures” blind source separation method to more than 2 channels

In a recent paper, we proposed a new blind source separation (BSS) method, which uses timefrequency (TF) information to extract two source signals from two linear instantaneous mixtures of these sources. In this new paper, we introduce an extension of the latter method, intended for the general situation when mixtures of source signals are available. Unlike previously reported TF BSS methods, the proposed approach only requires slight differences in the TF distributions of the considered signals: it mainly requests the sources to be ”visible”, i.e. to each occur alone in one local area of the TF plane. By using TF ratios of mixed signals, it automatically determines these single-source TF areas and identifies the corresponding parts of the mixing matrix. We present in detail the proposed method and give experimental results concerning mixtures of speech and music signals, thus showing that this approach yields very good performance.