一种具有自动聚类检测功能的欠定型盲信号混叠矩阵估计算法

A New Matrix Clustering Algorithm with Auto Detection for Underdetermined Sparse Component Analysis

针对源信号个数未知情况下的欠定稀疏分量分析模型,提出一种具有自动聚类检测功能的混叠矩阵估计算法。提出实现源信号个数的判定的观测信号自动检测聚类方法,同时利用主成分分析对超直线进行估计,从而实现混叠矩阵的精确估计。仿真实验结果表明,该算法适用范围广,是一种快速精确且稳健的混叠矩阵估计算法。

To the underdetermined sparse component analysis （SCA） model with unknown sources number, a new robust clustering algorithm with auto detect function for mixture matrix estimation is addressed. This approach consists of two parts： signal clustering and mixing matrix estimation. In the first step, a probability criterion is pro- posed for sources number detection, which stems from deduction by using a fit mathematical statistics model. To the second stage, principal component analysis （PCA） is introduced to the mixing matrix estimation. Experiment simu- lations illustrate the effectiveness of the new clustering algorithm.