Blind speech separation using canonical correlation and performance analysis

Abstract

Several methods have been explained for blind source separation (BSS) in the literature. Those methods fail when considered for separation of speech signals. This paper mainly focuses on blind speech signal separation from the observations using canonical correlation. The performance of the proposed method is evaluated in terms of signal to interference ratio (SIR) and time domain waveforms of separated speech signals. It is found that proposed technique will improve the SIR values compared with principal component analysis (PCA) and independent component analysis (ICA) based algorithms.