基于稀疏特性的欠定盲信号分离算法

Underdetermined blind source separation algorithm based on sparse representation

在源信号在非充分稀疏条件下,提出了一种改进的两步法欠定盲源分离算法.与现有的大多数稀疏分量分析算法法都是假设源信号是充分稀疏不同,该算法放宽了源信号的稀疏性.与此同时,该算法能够估计出聚类空间的个数,能够克服源信号个数未知的情况.模糊划分矩阵的应用更加有利于源信号的分离.仿真结果表明了该算法的有效性.

Based on the insufficient sparsity assumption of the source signals,a new algorithm is presented for underdetermined blind source separation.Different with existing sparse component analysis algorithms which assume that source signals are strictly sparse,the proposed algorithm is able to solve sparse component analysis problem in non-strictly condition;meanwhile,the proposed algorithm is also able to detect the clustering spaces number when the sources number is unknown previously.The fuzzy clustering method is also helpful in the second stage.Simulations are given to demonstrate the effectiveness of the proposed algorithm.