摘要
独立分量分析算法是一种多维统计方法。该算法的研究对象是多元随机信号,其研究目的是将这些多元随机信号转化成包含统计上相互独立的多个分量的信号。简要介绍了各种独立分量分析算法,包括基于二阶统计量的二阶盲辨识算法和多未知信源分离算法,以及基于高阶统计量的信息极大化法、改进的信息极大化法、快速固定点独立分量分析和特征矩阵联合近似对角化算法;比较了各种方法的运行性能并展望其在生物医学工程中的应用前景。
Independent component algorithm (ICA) is a method of higher-order statistics(HOS) with the study objects of multivariate random signals that are mutual independent. It aim is to transform multivariate random signal into the signal having components that are mutually independent in complete statistical sense. This article briefly introduce series of the ICA algorisms including second order blind identification, multiple unknown source extraction algorithm based on second-order statistics, as well as Informax, modified Informax, fast fixedpoint ICA and joint approximative diagonalization of eigenmatrix (JADE) algorithm that are based on HOS. At the end of the article, the performance of each algorithm is compared and its application prospect is forecasted.
出处
《国际生物医学工程杂志》
CAS
北大核心
2011年第4期249-252,I0003,共5页
International Journal of Biomedical Engineering
基金
北京师范大学认知神经科学与学习国家重点实验室开放课题资助项目
关键词
二阶统计量
高阶统计量
独立分量分析
生物医学信号
特征矩阵联合近似对角化算法
Second-order statistics
Higher-order statistics
Independent component analysis
Biomedical signal
Joint approximative aiagonalization of eigenmatrix