基于时移相关的ICA信源数目估计算法  

An Algorithm for ICA Source Number Estimation Based on Time Shift Correlation

作　　者：夏校朋 贾欢 张礼 XIA Xiao-peng;JIA Huan;ZHANG Li(Science and Techonology on Electronic Information Control Laboratory,Chengdu 610036,China)

机构地区：[1]电子信息控制重点实验室

出　　处：《电子信息对抗技术》2019年第4期11-16,共6页Electronic Information Warfare Technology

摘　　要：针对基于独立成份分析(Independent Component Analysis,ICA)的盲源分离,提出了一种基于时移相关的信源数目估计算法。通过考查ICA分离信号之间的时移相关程度,确定所分离信号是否为独立信号,进而确定真实信源数目。仿真表明本算法可以减小信号分离不彻底或弱信号等因素对信源数目估计的影响,提高信源数目估计准确度,有利于ICA算法在信号分选方面的实际应用。An algorithm for ICA source number estimation based on time shift correlation is presented.By considering the strength of time shift correlation among separated signals,it can be decided whether one separated signal is an actual signal,then the source number can be decided.Experiment results show that the proposed algorithm can increase the accuracy of source number estimation by decreasing the effect of weak signals and incomplete separation,which can be of great benefit to the application of signal sorting with ICA algorithms.

关 键 词：ICA 盲源分离 时移相关 信源数目估计 

分 类 号：TN971.1[电子电信—信号与信息处理]