Blind Source Separation of Convolutive Mixtures of Speech in Frequency Domain

Shoji MAKINO  Hiroshi SAWADA  Ryo MUKAI  Shoko ARAKI  

IEICE TRANSACTIONS on Fundamentals of Electronics, Communications and Computer Sciences   Vol.E88-A   No.7   pp.1640-1655
Publication Date: 2005/07/01
Online ISSN: 
DOI: 10.1093/ietfec/e88-a.7.1640
Print ISSN: 0916-8508
Type of Manuscript: INVITED PAPER (Special Section on Multi-channel Acoustic Signal Processing)
blind source separation,  convolutive mixtures,  independent component analysis,  frequency-domain BSS,  microphone array,  adaptive beamformer,  

Full Text: PDF(1.7MB)>>
Buy this Article

This paper overviews a total solution for frequency-domain blind source separation (BSS) of convolutive mixtures of audio signals, especially speech. Frequency-domain BSS performs independent component analysis (ICA) in each frequency bin, and this is more efficient than time-domain BSS. We describe a sophisticated total solution for frequency-domain BSS, including permutation, scaling, circularity, and complex activation function solutions. Experimental results of 22, 33, 44, 68, and 22 (moving sources), (#sources#microphones) in a room are promising.