摘要
针对现有算法基于最大信噪比的盲源分离算法在低信噪比情况下分离效果降低甚至失效的问题,提出了一种利用小波包改进的最大信噪比的盲源分离算法,提高了现有算法在较低信噪比下的分离效果。具体地,先利用小波对预处理后的观测信号进行滤波,用此信号代替现有算法中估计信号的滑动平均,然后根据信噪比建立的代价函数进行广义特征值分解计算得到分离矩阵。仿真结果表明,改进算法的分离效果优于现有算法的分离效果。
The existing blind source separation algorithm based on the maximum SNR will reduce or even invalidate in the case of low SNR.Aiming at this problem,a promoted algorithm called promoted the maximum SNR(PMSNR)by wavelet packet is proposed.The algorithm improves the separation of existing algorithms at lower SNR.Firstly,the preprocessed observation signal is filtered by wavelet.This signal is then used to replace the sliding average of the estimated signal in the existing algorithm.Finally,based on the cost function established by the SNR,the generalized eigenvalue decomposition is calculated to obtain the separation matrix.The simulation results indicate that the separation effect of the improved algorithm is better than that of the existing algorithm.
作者
申晨宇
刘增力
SHEN Chen-yu;LIU Zeng-li(Faculty of Information Engineering and Automation,Kunming University of Science and Technology,Kunming Yunnan 650500,China)
出处
《通信技术》
2018年第12期2812-2818,共7页
Communications Technology
基金
国家自然科学基金(No.61271007
No.60872157)~~
关键词
盲源分离
小波包分解与重构
最大信噪比
评价标准
BSS
wavelet packet decomposition and reconstruction
maximum SNR
evaluation standard
