Research On Blind Source Separation Problem Based On Bionic Intelligent Algorithm

Blind Source Separation(BSS)refers to a process of separating the source signals from only the observed mixed signals without any prior conditions where the source signals,transmission channel,and mixing model are unknown.In the determined BSS problems,Independent Component Analysis(ICA)is one of the most commonly used classical methods.The basic principle of ICA is to optimize the set objective function through the optimization algorithm according to the independence criterion conditions,so that the independence between the separated signals is the strongest.In the underdetermined BSS problem,the separation signals is difficult because the underdetermined mixing matrix are singular.The general solution is based on the "two-step " method,and the step is to first estimate the mixing matrix with the observed signals,and then combine the estimated mixing matrix to reconstruct the source signals using the optimization algorithm.Based on BSS theory and bionic intelligent algorithm,this paper studies the application of bionic intelligent algorithm in determined and undetermined BSS problems respectively,and the main work is as follows:Aiming at the problems of multiple solving parameters and poor separation accuracy in the application of traditional bionic intelligent algorithm in positive definite BSS,Refracted and Elite Opposition-based Learning Firefly Algorithm(REFA)was studied.First,the separation matrix is represented by Givens matrix,and taking negative entropy as the objective function.Then,the Refraction opposition-based learning is introduced into the standard Firefly Algorithm to enhance the global search ability of the original algorithm.Finally,an elite group is selected by using the elite selection strategy,which improves the convergence speed of the algorithm.The simulation results show that compared with the standard Firefly Algorithm and Particle Swarm Optimization algorithm,the proposed algorithm can effectively avoid premature convergence,and maintain outstanding separation performance in both noise-free and noisy conditions.In order to solve the problems of sensitivity to initial parameter setting and poor separation accuracy in the application of density clustering algorithm in underdetermined BSS,a hybrid algorithm combining Improved Salp Swarm Algorithm(ISSA)and Density Based Spatial Clustering of Application with Noise(DBSCAN)Algorithm was studied.First,to remove interference points and improve the performance of DBSCAN,this paper use the wavelet threshold noise reduction to de-noise the noisy observation signals.Then,this paper use the Salp Swarm Algorithm incorporating firefly perturbation strategy in order to find the domain radius of DBSCAN,solve the problem that the algorithm is sensitive to parameter settings,improve the robustness of the algorithm.Finally,the source signals is reconstructed by the L1-norm minimization algorithm.The simulation results show that,the wavelet threshold noise reduction preprocessing can effectively reduce interference points.Compared with the traditional density clustering algorithm,the proposed algorithm has better estimation of mixed matrix and better separation accuracy.