Real-Time Blind Source Separation of Speech Signals with Adaptive Particle Swarm Optimization

To solve the problems of slow convergence and low computational precision of blind source separation(BSS) based on traditional particle swarm optimization(PSO), a novel approach-based adaptive particle swarm optimization for real-time blind source separation is proposed, in which the observations are linear convolutive mixtures of statistically independent speech sources. It combines the independent component analysis algorithm and the adaptive particle swarm optimization, takes the negentropy of mixtures as a target function, and adaptively adjusts the inertia weight factor according to the negentropy of separated signals, and optimizes the separation matrix so as to makes each signal component independent. The simulation and experimental results on mixed speech signal showed an improved separation and a superior convergence rate of the proposed algorithm over that of PSO. Moreover, this algorithm can converge to a much lower cost value than that of PSO.