稀疏混合数据的矢量聚类分析与应用

对稀疏混合数据进行分析,发现该类数据具有方向性聚集分布的特点。首先证明了可以采用方向性聚类方法对稀疏混合数据进行处理分离出原数据。即用方向性聚类算法对稀疏混合数据进行聚类分析可以估计出混和矩阵。然后证明采用方向性聚类算法分离出来的数据和原数据之间具有确定的尺度和次序变化关系。最后针对多通道混合数据的盲分离提出了基于中心矢量聚类的稀疏混合数据分离算法SMDDCVC(sparse mixing data decomposition based on center vector clustering),并将该算法用于稀疏混合图像的盲分离。实验结果表明基于SMDDCVC算法的稀疏混合数据盲分离算法是有效的。

Vector Clustering Analysis for Sparse Mixing Data and Its Application

It was found that the sparse mixing data had a character of directional centralization distribution. The feasibility to decompose the mixing data into source by directional clustering was demonstrated. The decomposition was implemented by the estimation of the mixing matrix using the directional clustering algorithm. The relationship of the estimated source data and the real source data was demonstrated. A sparse mixing data decomposition algorithm based on directional clustering algorithm named SMDDCVC (sparse mixing data decomposition based on center vector clustering) was proposed to deal with the decomposition of multi-channel sparse mixing data. And this SMDDCVC algorithm was used to the decomposition of sparse mixing image. The experiment result shows that the proposed vector clustering algorithm and the proposed SMDDCVC algorithm is valid.