The Research Of Blind Source Separation And Extraction Based On Particle Swarm Algorithm

The technolegy of blind source separation and extraction is a new signal processing method that has come out recent years. It is very useful in many parts. As the technolegy is originality, complicated and useful, more and more researchers have been working on it. Many algorithms are produced, such as the Fixed Point Algorithm, the Grad Algorithm and the Genetic Algorithm, and so on. The paper works on the sereral parts as fowllows according to the disadvantages of traditional algorithms:Systematic analysis and summary is made in the main course of the development, theory and research of the blind source separation and extraction field at home and abroad with an emphasis on fixed-point algorithm, Informax algorithm and gradient algorithm, and the writer finds a number of defects in the algorithms. Gradient method is necessary to introduce a non-linear function according to the signal characteristics and the algorithm performance is affected. Fixed-point algorithm is easy to converge to a local extremum points if it does not have enough variety, and the genetic algorithm has a disadvantage in the computation time because of its complexity.The paper analies the principle and procedure of particle swarm optimization algorithm, and introduced it to the blind source separation and extraction, and provides a new research methods and ideas for blind signal processing. And we take the linear instantaneous mixtures as an example, and put it into realty in general signals, sound signals and image signals based on kurtosis. Algorithm directly optimize the fitness fuction that is base on kurtosis, without non-linear function, and the algorithm performance is not affected by the signal characteristics. The simulation examples of the algorithms, genetic algorithms and the standard gradient altorithm show that the particle swarm algorithm is better than the genetic algorithm in convergence speed, and higher than the standard gradient in accuracy, proves the validity and superiority of the algorithm.