基于二阶统计量的盲辨识方法研究  

Blind Identification Approach Research Based on Second Order Statistics

作　　者：张颖 ZHANG Ying(Electronic Information Engineering College,Henan Institute of Technology,Xinxiang 453003,China)

机构地区：[1]河南工学院电子信息工程学院,河南新乡453003

出　　处：《河南工学院学报》2020年第2期9-14,共6页Journal of Henan Institute of Technology

摘　　要：在信号的预处理阶段,可简化盲源分离或独立分量分析方法,采用基于二阶统计量的盲辨识方法,分离不同功率谱的有色源信号,以改善自适应系统的收敛特性,消除冗余或减少噪声污染。提出了一种基于特征值分解的二阶统计量的盲辨识方法,可以较好地识别混合矩阵,对有色源信号进行盲源分离。通过不同时滞τ的比较,以及不同算法的比较,证明了该辨识方法对不同的时滞在τ=1时的分离效果最好,且该方法比其他方法的分离效果更好。In the signal preprocessing stage,the blind source separation or independent component analysis method can be simplified,and the blind identification method based on second-order statistics can be used to separate the colored source signals with different power spectra,so as to improve the convergence characteristics of the adaptive system,eliminate redundancy or reduce noise pollution.A kind of second-order statistics based on eigenvalue decomposition of blind identification method is proposed,which can identify mixed matrix very well and make blind source separation of colored source signals.It is proved that the separation effect o f this identification algorithm is best when time delay is equal to 1 by the comparison of different time delay and different algorithms,and the separation effect of this algorithm is better than other algorithms.

关 键 词：二阶统计量 特征值分解 盲辨识 时滞 

分 类 号：TN911[电子电信—通信与信息系统]