卷积混叠信号的最小互信息量盲分离算法  

Blind Signal Seperation of Convolution Mixture Signals via Minimum Mutual Information(MMI)Method

作　　者：谭丽丽[1] 韦岗[1] 

机构地区：[1]华南理工大学无线电与自控研究所,广州510641

出　　处：《通信学报》1999年第10期49-55,共7页Journal on Communications

基　　金：国家自然科学基金;国家教委博士点基金;国家教委跨世纪人才培养基金;霍英东青年教师基金;广东省自然科学基金

摘　　要：本文提出了一种卷积混叠信号的盲分离算法。该算法利用当信号相互独立时互信息量最小的特性作为分离准则，应用随机梯度算法确定分离滤波器的系数，文中给出了详细的理论推导。理论分析及实验仿真证明了算法的有效性。This paper presents a new blind algorithm to separate convolutionary mixed signals through different LTI systems.The new algorithm ulitizes the property that the mutual information of indepent signals is minimum as the separating criterion,and estimates the coefficients of separating filters with stochastic gradient algorithm.Detailed theoritical deduction is given in the paper.The validity has been proved through theoritical analysis and experiments.This algorithm makes no special restriction on the mixing filters.

关 键 词：盲分离 熵 最小互信息量 盲解卷 信号处理 

分 类 号：TN911.23[电子电信—通信与信息系统] TN911.7[电子电信—信息与通信工程]