Reseauch And Application Of Underdetetmined Blind Souparation

Blind source separation(Source Separation Blind,BSS)technology,has increasingly become the focus of attention in the field of signal processing issues.The concept of "blind source separation" is originally proposed,which aims at solving a system to restore the source signal in the case that the source signal and the number of signals are unknown,and the mixture matrix is unknown but only the observed signal is known.In this paper.we mainly study the problem of blind source separation,that is,the number of observed signal is smaller than the number of source signals.Based on the sparse component analysis(Component Analysis Sparse,SCA)method,the estimation of mixing matrix and source signal are discussed in two stages,and a new method for estimating the mixing matrix and restoring the source signal is proposed respectively.The main contents of this paper include:The two step method based on SCA is discussed".In the mixing matrix estimation stage of three kinds of estimation methods,namely k-means algorithm.Hough transform and the potential function method;the principle of the algorithm is analyzed,and through the simulation experiment to achieve and verify the feasibility of the algorithm.In the estimation of the source signal.the most commonly used method of the shortest path is studied.An improved ant colony clustering algorithm based on ant foraging theory is proposed to estimate the mixing matrix.and the cluster centers are modified by the grid density method.Firstly the sparsity of the signal source,of observed signals were standardized treatment to form spherical reactor:and then to determine the initial pheromone matrix according to the Euclidean distance between the observed signal.obtained initial clustering center;then according to the traditional ant colony clustering method to cluster the data;then and grid density method is used to extract each kind of maximum density of grid,the center of the grid as the cluster center.Finally.the output of each cluster center as the column vectors of the mixing matrix.A weighted minimum norm method based on signal source for recovery,compared with the traditional norm method to find a set of optimal solution.improved norm method will other possible decomposition according to the weight added,so that the restoration of the signal are much closer to the original signal vector is proposed in this paper.When there are two observed signals.according to the decomposition of the angle difference between the observed signal and the size of the weighted value;when there are two or more observeds of the signal,the norm of each feasible solution as a weighted value of the mean.