基于高阶统计量的自适应盲源分离算法

基于高阶统计量的自适应盲源分离算法:提出了一种新的自适应盲源分离算法。在无噪音实时两源两传感器的情况下, 一旦观测信号被白化, 只需要辨识一个特定的旋转矩阵就可以完成盲源分离, 并给出了能表征该旋转矩阵的角的自适应估计器。仿真结果表明, 当满足源峭度和不为零的条件时, 这种方法是一种稳定的和有效的分离算法。
关键词: 盲源分离; 概率密度函数; 高阶统计量

Bl ind Sources Separa t ion Us ing Higher Order Sta t ist ic L IU Han, L IU D ing, L IU Xiao2yan (Xi’an U niversity of Techno logy,Xi’an 710048,Ch ina)Abstract: A new learn ing algo rithm is developed fo r b lind separat ion of independen t sou rcesignals f rom their linear m ix tu res. In the no iseless real2m ix tu re two2sou rce two2sen so r scenario, once the ob servat ion s are w h itened (deco rrelated and no rm alized) , on ly a given ro tat ion m at rix rem ain s to be iden t if ied in o rder to ach ieve the sou rce separat ion. In th is paper, an adap ter est im ato r of the angle that characterizes such a ro tat ion is derived. It show s that est im ato r converges to a stab le valid separat ion so lu t ion w ith the on ly condit ion that the sum of sou rce ku rto sis be dist inct f rom zero. Sim u lat ion demon st rate the validity of the algo rithm.
Key words: b lind sou rce separat ion; p robab ility den sity funct ion; h igher o rder stat ist ic