Blind Source Separation in Noisy and Reverberating Environment Using Genetic Algorithm
Masato Katou, Kaoru Arakawa · Hokkaido University Collection of Scholarly and Academic Papers (Hokkaido University) · 2009
A method of blind source separation is proposed using genetic algorithm (GA) for separating mixed voices in noisy and reverberating environment. Generally, the performance of the blind source separation becomes degraded when random noise is added to the mixed voices. Moreover, the reverberating environment must be considered in separating the mixed voices actually obtained. In order to solve the problem of blind source separation in such circumstances, a method to utilize GA is proposed, in which the system parameters are represented as chromosomes and the correlation between the output voices is minimized using GA. Computer simulations show its high performance in separating voices influenced with additive noise and reverberation.