关键词: 说话人识别, 粒子群算法, 支持向量机, 早熟收敛

Abstract: Aiming at the premature convergence problem of Particle Swarm Optimization（PSO），an improved PSO algorithm is proposed.When the current particles fall into local optimum，according to average distance of particles，this algorithm makes part of particles variate with a certain probability so that it can expand the global search ability.This improved algorithm is used to train Support Vector Machine（SVM），then applied to speaker recognition system.The experiment shows that it can achieve higher convergence speed and higher recognition accuracy.

Key words: speaker recognition, particle swarm optimization, support vector machine, premature convergence