论文部分内容阅读
基于互信息最小化的独立性测度对各分离信号间的非线性相关度度量没有归一化的问题,提出一种基于广义相关系数的盲信号分离(BSS)算法.首先选取后非线性混叠模型(PNL)分析基于广义相关系数的独立性测度;然后采用Gram-Charlier扩展形式估计输出参数并获取评价几率函数,结合最陡下降法求得分离矩阵和参数化可逆非线性映射的算法迭代公式.仿真结果表明,采用所提出的算法能够定量分析各分离信号间的非线性相关程度,有效分离后非线性混叠信号.
Based on the independence measure of mutual information minimization and the non-linear correlation measure of each separated signal, we propose a blind signal separation (BSS) algorithm based on generalized correlation coefficient.Firstly, The model (PNL) analysis is based on the independence measure of generalized correlation coefficient. Then the output parameters are estimated by Gram-Charlier extended form and the probability function of evaluation is obtained. The algorithm iterative formula of separation matrix and parameterized reversible nonlinear mapping is obtained by steepest descent method The simulation results show that the proposed algorithm can quantitatively analyze the nonlinear correlation between the separated signals and effectively separate the non - linear aliasing signal.