欠定盲源分离的稀疏分量分析方法

Sparse Component Analysis for Underdetermined Blind Source Separation

分析了解决欠定盲源分离问题的稀疏分量分析方法,通过聚类算法得到混合矩阵的估计,通过优化算法恢复源信号。总结了各种实现"聚类"和"优化"的具体方法,论述了欠定盲源分离和稀疏分量分析、压缩传感之间的紧密联系,对欠定盲源分离进行了展望。

The main method, namely, sparse component analysis （SCA）, which is used to solve underdetermined BSS, is analyzed. This class of methods consists of two stages： first clustering and then optimization, all of which firstly estimate the mixing matrix and then estimate the source signals given the estimated mixing matrix by means of optimization. Various algorithms for clustering and optimization in underdetermined BSS are summarized, and the close relationship between underdetermined BSS, SCA and compressed sensing is discussed. Finally prospect for underdetermined BSS is made.