Algorithm And Application Of Blind Signal Separation

BSS (Blind signal separation) technology is a hot research in the feild of modern digital signal processing. BSS algorithm requires very little prior information about the source signals and hybrid system, only based on some statistical properties of the source signals or some structural features of hybrid system, BSS can recover the source signals from the observed signals directly.This algorithm has very important theoretical value, it is widely used in the feild of voice, communication, biological medicine, feature extraction etc.Many researchers have done a lot of work in this area and blind source separation algorithm has been applied in more fields.The paper first introduces the basic knowledge of blind source separation.Thougth the analysis of ICA(Independent Component Analysis) and JDA(joint diagonalization algorithm),some appropriate improvements have been made on the original algorithms.For the ICA,nature gradient which is widely used to optimize criterion function has low convergence.So we introduce conjugate gradient to optimize EASI algorithm,some simulations have been done to verify the performance of new algotithm and results show that the algorithm speed up the convergence.For the JDA,to deal with the defects of easily immerging in degenerate solution frequently, constraint is imposed on updated matrix. Minimize the condition number of matrix while minimize the error function, the improved JDA avoids degradation solution and expands the range of applications. Simulations have been done to compare the performance of the new algotithm,AMUSE and SOBI. In the end, BSS algorithm is applied to multi-user detection, and our algorithm is compared with other three blind adaptive multi-user detection algorithms which are LMS,RLS and Kalman.Thougth experiments and simulations by matlab,the results show that it can improve the interference free and near-far performance of CDMA system.