Fast algorithm to estimate mixing matrix based on binary state sparse MoG model

Based on a binary state Gaussian sparse MoG model,a new fast method for identifying the mixing matrix in multiple dominant component sparse component analysis(SCA) is proposed.Based on the SCA models’ geometrical meaning,the observed datas’ distribution is discussed and it is proved that in some specific conditions,the mixing matrix A can be estimated by a simple density de-tection method without knowing the concentration hyperplanes generated from columns of A.A density detection estimation method which can greatly improve the computing efficiency of the algorithm is presented to estimate the mixing matrix.The simulation results prove the effectiveness of the proposed algorithm.