Study Of Estimating Mixing Matrix Of Undetermined Blind Source Separation

Currently,Blind Source Separation(BSS)is a popular research direction of signal processing,which the source signals can be estimated effectively without the need to know the source signals and the channel characteristics.This method has broad application prospects in communication,speech signal processing and biomedical signal processing,etc.Underdetermined Blind Source Separation(UBSS)is a challenging and hot topic in BSS,combining with the current research results,researches about the improvement of the underdetermined blind source separation is conducting in this paper:(1)Introduces other clustering algorithm to estimate the mixing matrix and analyzes the advantages and disadvantages of each algorithm.(2)Based on the window width of potential function method is difficult to be determined and K-means clustering algorithm relies on the initial clustering center and other shortcomings,a new method based on classifying the observation signals is proposed.Then compares the new algorithm to the potential function and K-means clustering with the simulation experiment.(3)Proposes the improved K-means method.The improved method can effectively reduce the influence of extreme values to clustering centers compared to the original K-means method and also has lower time complexity than the K-medoids method.(4)Applys underdetermined blind source separation algorithm to solve the problem of chaotic signals and speech signals aliasing,the experiment shows that the underdetermined blind source separation algorithms are efficient to the chaotic signals and speech signals aliasing problems,and validates the properties of the algorithms during estimating the mixing matrix.