On calculating the inverse of separation matrix in frequency-domain blind source separation

Authors:
Hiroshi Sawada;Shoko Araki;Ryo Mukai;Shoji Makino
Affiliations:
NTT Communication Science Laboratories, NTT Corporation, Kyoto, Japan;NTT Communication Science Laboratories, NTT Corporation, Kyoto, Japan;NTT Communication Science Laboratories, NTT Corporation, Kyoto, Japan;NTT Communication Science Laboratories, NTT Corporation, Kyoto, Japan
Venue:
ICA'06 Proceedings of the 6th international conference on Independent Component Analysis and Blind Signal Separation
Year:
2006

Citing 3
Cited 0

Pattern Classification (2nd Edition)

Pattern Classification (2nd Edition)
Blind source separation combining independent component analysis and beamforming

EURASIP Journal on Applied Signal Processing
Blind separation of speech mixtures via time-frequency masking

IEEE Transactions on Signal Processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

For blind source separation (BSS) of convolutive mixtures, the frequency-domain approach is efficient and practical, because the convolutive mixtures are modeled with instantaneous mixtures at each frequency bin and simple instantaneous independent component analysis (ICA) can be employed to separate the mixtures. However, the permutation and scaling ambiguities of ICA solutions need to be aligned to obtain proper time-domain separated signals. This paper discusses the idea that calculating the inverses of separation matrices obtained by ICA is very important as regards aligning these ambiguities. This paper also shows the relationship between the ICA-based method and the time-frequency masking method for BSS, which becomes clear by calculating the inverses.