Underdetermined mixing matrix estimation algorithm based on data field clustering and cloud model partitioning
Ruan Guoqing, Guo Qiang, Gong Xue · 2017
This paper analyses the estimation problem of the underdetermined mixing matrix for the blind signal separation. A novel algorithm involving data field clustering (DFC) and cloud model partitioning (CMP) is proposed. Firstly, single source points (SSPs) are identified by argument detection method. Then, the DFC algorithm is innovatively introduced to select the high potential SSPs to determine initial cluster centers. Finally, the membership degree of the adjacent clustering centers is obtained by combining the cloud model method to correct the clustering centers, and the estimation of the mixing matrix is obtained. Experiment results indicate that the estimation problem of the adjacent cluster centers can be effectively solved by using the proposed algorithm, and the estimated mixing matrix has higher precision.