Research On Blind Source Separation Methods Based On Improved Meta-heuristic Algorithm

Blind Source Separation(BSS)refers to the recovery method of extracting unknown source signals from observed signals only when the prior information such as original signals and mixing mode is unknown.With the development of recent years,BSS has gradually evolved into a hot topic in the field of signal processing and has potential application prospects in many fields.Independent Component Analysis(ICA)is a conventional method to solve linear mixed BSS problems.When most ICA algorithms solve problems,the quality of gradient information determines the speed of convergence,and the algorithm is easy to fall into the local extreme value,but also involves the selection of a nonlinear activation function.The meta-heuristic algorithm is one of the effective methods to optimize an objective function,which is derived from the computational intelligence mechanism to analyze complex problems and then obtain the optimal or satisfactory solution.Therefore,applying the Meta-heuristic algorithm to the linear BSS problems can make up for the deficiency of the traditional BSS algorithm.Based on the research of Meta-heuristic algorithm and BSS technology theory and algorithm,this paper mainly completes the following work:(1)To overcome the disadvantages of slow convergence speed and poor separation accuracy of traditional Meta-heuristic algorithms for Blind Source Separation problems,a BSS method based on the Improved Elephant Herding Optimization(IEHO)algorithm was proposed.This method combines kurtosis and mutual information of separated signals to construct an objective function according to the principle of independence.In the algorithm updating stage,the diversity of search methods is improved by improving the scale factor and adding neighborhood search terms.In the separation stage,quantum particle swarm optimization is introduced to improve the global search capability of the algorithm.To verify the effectiveness of the improved algorithm,four test functions are selected to test the optimization effect of the algorithm.Compared with the traditional image swarm optimization algorithm and particle swarm optimization algorithm,the IEHO algorithm has a better optimization effect and achieves the blind separation of the image signal and speech signal.By comparing the similarity coefficient,signal-to-dry ratio,and PI performance evaluation criteria,the superiority of the proposed algorithm in separation accuracy and convergence speed is verified.(2)To improve the traditional fuzzy clustering algorithm,which is sensitive to initial values,has poor robustness,and is easy to fall into local extremum.A hybrid clustering algorithm combining Improved Shuffled Frog Leaping Algorithm(ISFLA)and Possibilistic Fuzzy C-means(PFCM)was proposed.The method is applied to the problem of underdetermined blind source separation.In this algorithm,the iterative process of the gradient descent method in PFCM was replaced by the optimization process of the hybrid leapfrog algorithm.ISFLA initialized the population through the current optimal reverse learning mechanism and added the Gaussian random walk strategy after the local search of subgroup,which effectively improved the optimization ability of the algorithm.The simulation results show that the ISFLA algorithm is better than the traditional hybrid leapfrog algorithm and particle swarm optimization algorithm.At the same time,the fusion algorithm improves the robustness,clustering accuracy,and searching ability of the fuzzy clustering algorithm,and successfully realizes the signal reconstruction for the under-determined blind source separation problem.The algorithm has high estimation accuracy and good stability.