摘要
独立向量分析根据信源统计独立特性对观测信号进行分离运算,扩展Informax算法既能分离超高斯信号,也能分离亚高斯信号,得到广泛的应用。本文基于扩展Info-max算法特点,提出了一种自适应的学习算法,该算法使得学习步长根据信号的代价函数变化而变化,克服了扩展Infomax算法在稳态步长调整过程中的不足,仿真结果证实了该算法的有效性。
Independent component analysis did signal separation operation based on independences of the observed signal. Extended Informax could separate Super - Gaussion signal and Sub - Ganssion signal, and got widely used. An improved self-adaptive learning algorithm was introduced in the paper. The algorithm made learning step change according to the cost function of signal changes, and overcame the disadvantages of extended informax algorithm in the process of step size change of adaptive steady state. The simulations had verified its validity.
出处
《九江学院学报》
2008年第6期20-23,共4页
JOurnal of Jiujiang University :Social Science Edition
关键词
独立向量分析
扩展Infomax
超高斯
亚高斯
independent component analysis
extended information - maximization
super - Gaussion
sub - Gaussion
