在线增强型复值混合信号盲分离算法研究  

Online Blind Separation of Enforced Mixed Complex-Value Signal Sources

作　　者：丛丰裕[1] 雷菊阳[1] 许海翔[1] 周士弘[2] 杜栓平[2] 史习智[1] 

机构地区：[1]上海交通大学振动冲击噪声国家重点实验室,上海200240 [2]杭州应用声学研究所海洋声学国家级重点实验室,杭州310006

出　　处：《西安交通大学学报》2006年第9期1070-1073,共4页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(60372075);国家重点实验室基金资助项目(VSN-2006-04)

摘　　要：基于二阶统计量,对在线分离复值混合信号法进行了研究.假设源信号是独立的且非常态,信号的伪协方差矩阵能增加约束条件,从而可证明二阶统计量能够完全分离复值混合信号,而且对信号是否平稳不作要求.结合非常态信号的独立性,构造出代价函数,利用梯度下降法推导出在线盲分离算法.通过盲分离算法的仿真试验,发现所提出的盲分离算法能充分利用复值非常态信号的二阶统计量,算法具有鲁棒性好、运算速度快和可在线实现等优点.Based on the second order statistics, an algorithm is proposed to separate mixed complex-value signals online. Since a pseudo-covariance matrix of the observed signals supply additional constraints if the sources are assumed to be independent and improper, it can be proved that the second order statistics is sufficient to separate the mixed complex-value signals for both stationary and non-stationary sources. A cost function is explored by considering the independence of the improper sources, and the online separation algorithm is then deduced with descending gradient. A simulation comparatively demonstrates the robustness and effectiveness of the new online algorithm due to sufficient utilization of the second order complex statistics.

关 键 词：盲分离 复值混合信号 二阶统计量 伪协方差矩阵 在线 

分 类 号：TN911.72[电子电信—通信与信息系统]