矢量聚类及其在稀疏分量分析中的应用

针对传统聚类分析不能有效处理矢量数据聚类的问题，提出矢量聚类算法。该算法以点到矢量的距离最小化为分类依据，所得类簇中心为一矢量。根据稀疏信号的分布特性，用矢量聚类方法估计系统的混合矩阵，再利用估计的混合矩阵分离混合信号，从而得到稀疏信源的估计，简化了传统的混合信号分离过程。实验结果表明该矢量聚类方法能比传统的标量聚类方法更有效地估计矢量数据的中心，能在稀疏的处理域中很好地分离出稀疏信源。

Vector Clustering and Its Application in Sparse Component Analysis

A vector clustering algorithm is proposed to cope with the inefficacy of traditional clustering algorithms to vector data. The algorithm classifies data into clusters by minimizeing the distance of a datum to a vector. The clustered centers are vectors. According to the distributing character of the sparse signal, a sparse signal separation algorithm is proposed which estimates the mixture matrix based on the vector clustering algorithm, and separates the source signal using the estimated mixture matrix. The algorithm is simple in computation comparing with traditional separation algorithms. Experimental results show that the algorithm is effective in vector data clustering and sparse signal separation.