独立分量分析算法的改进研究  

Researches on Modifity of Independent Component Analysis Method

作　　者：王小敏[1] 曾生根[1] 夏德深[1] 

机构地区：[1]南京理工大学计算机科学系,南京210094

出　　处：《电子器件》2008年第5期1681-1684,共4页Chinese Journal of Electron Devices

摘　　要：独立分量分析是信号处理技术的新发展,本文对独立分量分析快速算法进行了研究,在分析FastICA算法的核心迭代过程的基础上,提出了改进算法M-FastICA,以改善了算法的收敛性能,减少算法的迭代次数。针对M-FastICA算法的收敛依赖于初始权值的问题,在算法过程中加入松弛因子,提出LM-FastICA算法,改善了算法对初始权值的依赖性。Independent Component Analysis （ICA） was a new development of signal processing. ICA algorithm and its fast algorithm （FastICA） were introduced. Based on analyzing kernel iteration course of the FastICA algorithm M-FastICA was advanced. M-FastICA improved convergence performance and reduced iterations process. Aimed at the convergent speed of M-FastICA was dependent on initial weights, LM-Fas- tICA was advanced by imported looseness agent,and reduced the dependence on initial weights.

关 键 词：独立分量分析 FASTICA M-FastICA LM-FastICA 

分 类 号：TN912.3[电子电信—通信与信息系统]