Research On Blind Sources Separation Issue Based On Momentum Technology And Auxiliary Separation System

Blind Source Separation(BSS)is intended to separate or recover a set of unobservable source from another set of mixtures,without having access to the mixing matrix coefficients.Natural gradient algorithm is one of the most important algorithms for blind source separation,which has attracted more and more attention.At the same time,because of the natural gradient algorithm has low calculation,fast convergence speed,good separation effect,highlighting its good application prospect.And in the field of speech signal processing,image processing,wireless communication has been widely used.This paper introduces the blind source separation of the background and basic theory:The basic model of the blind source separation,discusses the characteristics and pretreatment methods of blind source separation;From the point of view of optimization algorithm,this paper introduces several kinds of cost function.And extract the classical blind source separation algorithm--natural gradient algorithm.And further to the natural gradient algorithm with a non-holonomic constraint,non-holonomic natural gradient algorithm.Aiming at the defects of the natural gradient algorithm for blind source separation,in order to update,join the renewal a moment ago through the tap coefficients of adaptive iterative algorithm of the rules in the present moment.In order to improve the system convergence and tracking speed,improve the output performance of the algorithm,so we put forward the momentum of blind source separation algorithm.The accuracy of the parameters and the rationality of the momentum factor will greatly influence the convergence performance of the algorithm.When the momentum factor is set to a fixed value,the algorithm will still exist the convergence rate and the steady-state error cannot compromise problem,this paper proposes a variable momentum factor to optimize the separation algorithm.In addition,this paper proposes a blind source separation algorithm called non-holonomic natural gradient algorithm in optimization strategy based on step.The paper leads into a dual system to use auxiliary separation performance index value of PI system update step size adaptively.This can effectively avoid the contradiction between convergence speed and steady-state error.The simulation results show that,the proposed algorithm has better separation effect.