A Novel Approach for Underdetermined Blind Sources Separation in Frequency Domain

Abstract

In this paper, we discussed the separation of n sources from m linear mixture when the underlying system is underdetermined, that is, when m < n. The underdetermined blind sources separation has two steps. In matrix-recovery step, we defined a characteristic of the signals as the durative-sparsity and proposed a novel approach called as a searching-and-averaging-based method in frequency domain. This approach tells us how to search some data points that are very close to the basis lines along the direction of basis vectors a ^j and how to use them to estimate the mixing matrix. In source-recovery step, we used Bofill and Zibulevsky’s shortest-path algorithm. Finally, the separation results were obtained using their short-time Fourier transforms.

The work is supported by the Guang Dong Province Science Foundation for Program of Research Team (grant 04205783), the National Natural Science Foundation of China (Grant 60274006), the Natural Science Key Fund of Guang Dong Province, China (Grant 020826), the National Natural Science Foundation of China for Excellent Youth (Grant 60325310) and the Trans-Century Training Program, the Foundation for the Talents by the State Education Commission of China, and the SRF for ROCS, SEM.