The Research Of Particle Swarm Optimization Algorithm In Speech Signal Blind Source Separation Technologies

According to speech enhancement technology which is the most front key technology inthe speech signal processing.This paper used the latest blind source separation method,andintroduced particle swarm optimization algorithm into the speech signal blind sourceseparation processing, which had successful application example in artificial intelligence andpattern recognition etc. Do application research on the basis of theoretical analysis. Theconcrete research content this paper did is as follows:First,the paper Introduced the mathematical foundation applied in the blind sourceseparation algorithm,and analyzed the process of signal mixed and separation according toseveral blind source separation based models.Second,this paper analyzed particle swarm optimization algorithm and improvedalgorithm. Such as adaptive particle swarm optimization algorithm, simulated annealingparticle swarm optimization algorithm.And the paper chose the algorithm that convergencespeed which is quick and optimization performance is good.Third, the paper did a concrete analysis of several common blind source separationalgorithms and optimizations that used in the process of the objective function, and introducedtwo kinds of performance index to measure separation performance.Fourth,the paper chose appropriate blind source separation algorithm combined with thecharacteristics of speech signal,and applied several commonly used blind source separationalgorithm, particle swarm optimization algorithm and its improvement to blind sourceseparation of Three speech signal’s linear instantaneous mixture. Simulation based on matlab.Compare separation results, performance index and convergence speed.Fifth,the paper analyzed the blind source separation of signal convolution mixed in thetime domain and frequency domain, analysis the feasibility and process of speech signalconvolution mixed’s blind source separation in the frequency domain,and put forward the whole train of thought of speech signal convolution mixed’s blind source separation in thefrequency domain using particle swarm optimization algorithm and its improvement.