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.