On the Use of Time-Frequency Representation in Multicomponent Signal Separation

In this paper, we address the problem of separating unknown multicomponent signals from their instantaneous mixtures. Using linear time-frequency (TF) representation of the mixtures along with vectors classification scheme provide us a simple and efficient technique to separate multicomponent signals. The proposed algorithm can handle monocomponent as well as multicomponent sources and its assumptions about the mixing matrix are more relaxed compared to other existing TF based algorithms. The source separation results for the mixed synthetic signals as well as mixed real audio signals, such as mixture of speech and music, are shown to illustrate the validity and efficiency of the proposed scheme.

• Type: Conference
• Year: 2006
• Where: ICMCS
• Authors: Braham Barkat, Farook Sattar, Karim Abed-Meraim
• Keywords: ICMCS 2006, Multicomponent Signals, Separate Multicomponent Signals, Unknown Multicomponent Signals