一种非平稳卷积混合声音信号的盲分离算法  

A Blind Source Separation Algorithm of Non-stationary Convolved Sounds Mixtures

作　　者：熊英[1] 

机构地区：[1]电子科技大学电子工程学院,成都610054

出　　处：《实验科学与技术》2009年第1期34-36,共3页Experiment Science and Technology

摘　　要：讨论一种非平稳卷积混合声音信号盲分离算法。该算法基于多时延解相关准则建立代价函数,用Frobenius范数最小化代价函数得分离矩阵。算法计算量低,适用于自适应在线运算,而且不要求时延相关矩阵是正定的。对非平稳卷积混合声音信号进行仿真实验,验证了算法的有效性。This paper discusses a blind separation algorithm of non-stationary convolved sounds mixtures. A cost function is built on multiple timedelay decorrelation rifle, the separation algorithm is obtained by minimizing the cost function , using the Frobenius norm. The algorithm has low computing cost, more adapt to on-line blind separation and the time delayed correlation matrix is not necessarily positively definite. Simulation confirms its validity for non - stationary convolved sounds mixtures.

关 键 词：盲信号分离 卷积混合 非平稳信号 代价函数 

分 类 号：TN911.72[电子电信—通信与信息系统]