Blind source separation for convolutive audio mixing

Abstract

This paper describes an efficient Blind Source Separation of speech and music, speech and music which are considered as convolutive mixtures. The convolutive mixed signals consist of source signals and same amount of delay or echo of the same source signal. Convolutive BSS of stereo mixtures is the challenging task in the audio signal processing application.BSS is a technique for estimating original source signal from their mixtures of signals. The mixed signals were decomposed by 1D multilevel discrete wavelet decomposition. Decomposition levels are changed and the signals to noise ratio (SNR) are calulated.After that ICA has been performed and the sources are separated. © 2011 Springer-Verlag.