File(s) under permanent embargo

Underdetermined blind source separation based on relaxed sparsity condition of sources

journal contribution
posted on 2009-02-01, 00:00 authored by D Peng, Yong XiangYong Xiang
Recently, Aissa-El-Bey et al. have proposed two subspacebased methods for underdetermined blind source separation (UBSS) in time-frequency (TF) domain. These methods allow multiple active sources at TF points so long as the number of active sources at any TF point is strictly less than the number of sensors, and the column vectors of the mixing matrix are pairwise linearly independent. In this correspondence, we first show that the subspace-based methods must also satisfy the condition that any M × M submatrix of the mixing matrix is of full rank. Then we present a new UBSS approach which only requires that the number of active sources at any TF point does not exceed that of sensors. An algorithm is proposed to perform the UBSS.

History

Journal

IEEE transactions on signal processing

Volume

57

Issue

2

Pagination

809 - 814

Publisher

IEEE

Location

Piscataway, N.J.

ISSN

1053-587X

eISSN

1941-0476

Language

eng

Publication classification

C1 Refereed article in a scholarly journal; C Journal article

Copyright notice

2008, IEEE