학술논문
Blind Separation of Mutually Correlated Sources Using Precoders
Document Type
Periodical
Author
Source
IEEE Transactions on Neural Networks IEEE Trans. Neural Netw. Neural Networks, IEEE Transactions on. 21(1):82-90 Jan, 2010
Subject
Language
ISSN
1045-9227
1941-0093
1941-0093
Abstract
This paper studies the problem of blind source separation (BSS) from instantaneous mixtures with the assumption that the source signals are mutually correlated. We propose a novel approach to BSS by using precoders in transmitters. We show that if the precoders are properly designed, some cross-correlation coefficients of the coded signals can be forced to be zero at certain time lags. Then, the unique correlation properties of the coded signals can be exploited in receiver to achieve source separation. Based on the proposed precoders, a subspace-based algorithm is derived for the blind separation of mutually correlated sources. The effectiveness of the algorithm is illustrated by simulation examples.