APSIPA Transactions on Signal and Information Processing > Vol 8 > Issue 1

A review of blind source separation methods: two converging routes to ILRMA originating from ICA and NMF

Hiroshi Sawada, NTT Corporation, Japan, sawada.hiroshi@lab.ntt.co.jp , Nobutaka Ono, Tokyo Metropolitan University, Japan, Hirokazu Kameoka, NTT Corporation, Japan, Daichi Kitamura, Kagawa College, Japan, Hiroshi Saruwatari, The University of Tokyo, Japan
 
Suggested Citation
Hiroshi Sawada, Nobutaka Ono, Hirokazu Kameoka, Daichi Kitamura and Hiroshi Saruwatari (2019), "A review of blind source separation methods: two converging routes to ILRMA originating from ICA and NMF", APSIPA Transactions on Signal and Information Processing: Vol. 8: No. 1, e12. http://dx.doi.org/10.1017/ATSIP.2019.5

Publication Date: 14 May 2019
© 2019 Hiroshi Sawada, Nobutaka Ono, Hirokazu Kameoka, Daichi Kitamura and Hiroshi Saruwatari
 
Subjects
 
Keywords
Blind source separation (BSS), Time-frequency-channel tensor, Independent component analysis (ICA), Nonnegative matrix factorization (NMF), Majorization-minimization algorithm with auxiliary function
 

Share

Journal details

Open Access

This is published under the terms of the Creative Commons Attribution licence.

Downloaded: 7647 times

In this article:
I. INTRODUCTION 
II. MODELS 
III. OPTIMIZATION 
IV. EXPERIMENT 
V. CONCLUSION 

Abstract

This paper describes several important methods for the blind source separation of audio signals in an integrated manner. Two historically developed routes are featured. One started from independent component analysis and evolved to independent vector analysis (IVA) by extending the notion of independence from a scalar to a vector. In the other route, nonnegative matrix factorization (NMF) has been extended to multichannel NMF (MNMF). As a convergence point of these two routes, independent low-rank matrix analysis has been proposed, which integrates IVA and MNMF in a clever way. All the objective functions in these methods are efficiently optimized by majorization-minimization algorithms with appropriately designed auxiliary functions. Experimental results for a simple two-source two-microphone case are given to illustrate the characteristics of these five methods.

DOI:10.1017/ATSIP.2019.5