A Cauchy Potential Function Method for Underdetermined Blind Source Separation

摘要

In this paper, a novel Cauchy potential function method is proposed to estimate the mixing matrix for underdetermined blind source separation, wherein the number of mixtures is less than the number of the sources. The mixing matrix can be obtained by estimating the local maximum of the Cauchy potential function with unknown the number of the sources. In order to increase the robustness to the noise and sparsity of the sources, a global Cauchy Potential Function is employed to estimate the number of the sources and the mixing matrix. The simulation results show the validity and robustness of the algorithm.