Mixing matrix identification for underdetermined blind signal separation: Using hough transform and fuzzy K-means clustering
Tsung-Ying Sun, Ling-Erh Lan, Chan-Cheng Liu, Chih-Li Huo · 2009
This paper focuses on the underdetermined blind signal separation problem with sparse representation. The algorithm is proposed to identify the parameters of mixing model which are unknown. The distribution of mixtures are mapping to a new histogram domain by Hough transform which converts the Cartesian image space to the normal parameterization. And then, fuzzy k-means clustering is employed to seek the cluster centers, i.e. parameters of mixing model, on the histogram. Obtaining accurate estimates, the sources can be recovered clearly. The proposed algorithm and three existing algorithms are tested in the simulations. By the simulation results, our algorithm is able to perform a nice accuracy of estimation through a very low computational consumption.