Potential Function Agglomeration Clustering Algorithm for Sparse Component Analysis

In this paper, the Potential Function Agglomeration Clustering (PFAC) algorithm has been proposed for estimating the mixing matrix in underdetermined Sparse Component Analysis (SCA), wherein the number of mixtures is less than the number of the sources. In contrast to many existing SCA methods, the PFAC algorithm can accurate estimate the number of sources and the mixing matrix. The algorithm also exhibits two robust characteristics: (1) robust to the additive noise and outliers; (2) robust to the source signals are insufficient sparsity. The simulation results show the validity of the algorithm.