关键词: 萤火虫算法, 变步长, 盲源分离

Abstract:

Traditional natural gradient algorithm of blind source separation for speech signals is contradictions, which suffers errors between convergence speed and steadystate. A novel variable stepsize algorithm is proposed to solve this problem. The stepsize of traditional natural gradient algorithm is adjusted by the firefly algorithm to yield variable signals with the separation degree. Computer simulations verify the feasibility of this new algorithm, which offers better separation results with faster convergence and less steadystate errors.

Key words: firefly algorithm, variable step size, blind source separation