期刊名称：International Journal of Computer Science and Network Security
印刷版ISSN：1738-7906
出版年度：2006
卷号：6
期号：3A
页码：224-229
出版社：International Journal of Computer Science and Network Security
摘要：This paper introduces new source separation techniques exploiting the cyclostationarity property of the source signals. Two cyclostationary-based reference contrast functions are presented for convolutive mixtures and cyclostationary signals. Contrary to the major contrasts, any para-unitarity constraint is made on the mixing system. Their maximization enjoys identifiability properties, and aims at delivering outputs satisfying specific properties, such as statistical independence or a cyclostationarity character. Simulation examples are presented to illustrate the effectiveness of this approach in terms of the estimated source variances.
