基于ICA的快速定点算法

详细信息 本馆镜像全文    |  推荐本文 | | 获取馆网全文

英文篇名：A Fast Fixed-Point Algorithm Based on ICA 中文刊名：电声技术 英文刊名：Audio Engineering 作者：唐江波 ; 曾庆宁 ; 郭威 英文作者：TANG Jiang-bo ; ZENG Qing-ning ; GUO Wei (Department of Communication and Information Engineering ; Guilin University of Electronic Technology ; Guilin Guangxi 541004 ; China) 中文关键词：负熵 ; 独立成分分析 ; 定点ICA ; 快速定点算法 ; 对照函数 英文关键词：negentropy ; independent component analysis ; FastICA ; fast fixed-point algorithm ; contrast function 出版日期：2007-08-17 机构：桂林电子科技大学通信与信息工程系,桂林电子科技大学通信与信息工程系,桂林电子科技大学通信与信息工程系 广西桂林541004,广西桂林541004,广西桂林541004 年：2007 期：08 出版单位：电声技术

摘要

介绍了独立分量分析(ICA)的模型定义、数学原理等基本问题。在分析ICA基础上引入了固定点算法(FastICA)。FastICA算法收敛速度快,迭代次数由传统算法的2000次减少到3～10次。实验表明,FsatICA具有良好的盲源分离性能和鲁棒性。
The fundamental problems of ICA(Independent Component Analysis) about the model definition and mathematical principles are given. Based on analysis of the existing algorithms, the FastICA is introduced. The convergence speed of FastICA is fast and the times of iterations is reduced from about 2 000 to 3~10 using FastICA. Finally, the experimental results show that the FastICA has good performance for blind source separation and robustness.

引文

[1]JUTTER C,H ERAULT J.Blind separation of sources,PartⅠ:an adaptive algorithm based on neurom im atic architecture[J].Signal Processing,1991,24(1):1-10.
    [2]COM ON P.Independent com ponent analysis,a new con鄄cept-[J].Signal Processing,1994,36(3):287-314.
    [3]COM ON P.Blind separation of sources,PartⅡ:problem statem ent[J].Signal Processing,1991,24(1):11-20.
    [4]BELL A J.An inform ation-m axim ization approach to blind separation and blind deconvolution[J].Neural Com-putation,1995(7):1129-1159.
    [5]LEE T W.Independent com ponent analysis using an ex-tended infom ax algorithm for sub-Gauss and super-Gauss source[J].Neural Com putation,1999,11(2):409-433.
    [6]AM ARI S I.New learning in structural param eter space-natural Riem annian gradient[J].Advance in Neural Infor-m ation Processing System s,1997(9):127-133.
    [7]AM ARI S I.Natural gradient works efficiently in learn-ing[J].Neural Com putation,1998,10(2):251-276.
    [8]H YVARINEN A.Fast and robust fixed-point algorithm for independent com ponent analysis[J].IEEE Trans.on Neural Network,1999,10(3):626-634.
    [9]LEE T W,GIROLAM I M,BELL A J,et al.A unifying fram ework for independent com ponent analysis[J].Interna-tional Journal of Com puters and M athem atics with Appli-cations,2000,39:1-21.
    [10]刘喜武,刘洪,李幼铭.独立分量分析及其在地震信息处理中初探[J].地球物理学进展2003,3:90-96.