Speaker Recognition Based on APSO-K-means Clustering Algorithm

In this paper, combined with ant colony algorithm, particle swarm optimization algorithm, K-means clustering algorithm, we propose an APSO-K-means clustering algorithm applied to speaker recognition. The algorithm utilizes the strong ability of the ant colony algorithm to process local extremum to avoid the sensitivity to local optimization of the PSO algorithm (APSO). Meanwhile, it utilizes APSO to guide the initialization of the cluster centers to improve the deficiency of the K-means clustering algorithm which depends on the initial value, which makes it easy to converge toward global optimality. The experiments in speaker recognition show that the new approach is better than the traditional method and effectively reduces the error recognition rate.