摘要
提出了一种"逐次替代"卷积混叠盲信号分离方法。针对观测信号数目多于源信号数目的混叠情形,基于最小均方误差准则,将问题转化为求解一个关于信道参数和源信号的优化问题。通过代价函数对未知参数求导数,分别得到关于信道参数和源信号的两个表达式,通过对两个表达式的相互逐次替代来寻求源信号。由此给出了"逐次替代"卷积混叠盲分离方法。逐次替代盲解卷算法无需设置迭代步长,容易编程实现。仿真表明,该算法对于FIR SIMO情形能得到较好的效果。
An SS (successive substitution) convolutive BSS algorithm is developed for convolutive BSS problems where the number of mixtures is larger than the number of sources. And based on the least square error criterion, this problem can be transformed to solve an optimization problem with respect to the channel parameters and the sources. By taking the derivatives and setting the derivatives to be zeros, two expressions with respect to the channel parameters and source signals can be obtained respectively. By alternatively update the channel parameters and sources using these two expressions, the algorithm will be convergent ultimately. Then the estimations of the sources is obtained and it is easy to implement the proposed algo- rithm. Simulation results show that the proposed algorithm have a great performance in the ease of FIR SIMO.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2007年第9期1436-1440,共5页
Systems Engineering and Electronics
基金
国家杰出青年科学基金(60325310)
国家自然科学基金重点项目(U0635001)
国家自然科学基金(60505005)
广东省自然科学团队研究项目(04205783)
科技部重大基础前期研究专项(2005CCA04100)
广东省自然科学基金(05103553)资助课题
关键词
单输入多输出
卷积混叠
解卷
盲信号分离
single input multiple output
eonvolutive mixing
deeonvolution
blind signal separation