一种针对复值信号的独立分量分析方法  

A method for ICA for complex-valued sources

作　　者：李小军[1] 楼顺天[1] 张贤达[2] 

机构地区：[1]西安电子科技大学雷达信号处理国家重点实验室,陕西西安710071 [2]清华大学自动化系,北京100084

出　　处：《西安电子科技大学学报》2005年第3期447-451,共5页Journal of Xidian University

基　　金：国家自然科学基金资助项目(60072043)

摘　　要：利用独立分量分析方法对相互独立的复值信号进行盲提取,即从统计特性出发,通过变换观测信号,能够获得尽可能相互独立的信号.并在广义EHA准则基础上进行了变化,通过直接挑选非线性函数的方法,实现盲复值信号的提取.利用梯度下降法实现算法学习,仿真实验验证了算法的有效性.ICA is used to get the complex-valued signals. ICA is a statistical method for transforming an observed random vector into components that are as mutually independent as possible. By changing the extended Hebbian learning and directly using the nonline function, blind separation of complex-valued signals will be obtained. By using this method, the algorithm based on the gradient steepest ascent is proposed, with its computational efficiency shown by simulations.

关 键 词：独立分量分析 广义EHA准则 梯度下降法 

分 类 号：TN911.23[电子电信—通信与信息系统]