Study of Codebook Design Based on Particle Swarm Optimization

Because the standard particle swarm optimization(PSO)easily occurs a local optimization problem in codebook design,a new improved PSO codebook design algorithm by adopting simulated annealing(SA)is proposed to improve the global extremum updating rule.The improved updating rule adopts the form of random probability disturbance,which can randomly accept the better search results and the worse ones.The aim is to increase the search ability in the global optimum area,and avoid the premature phenomena of the particles.By applying the new codebook design algorithm to speech vector quantization,the experimental results show that the reconstructed speech of the standard PSO has a light faintness than that of the improved PSO,and the reconstructed speech of the improved PSO is more clear and natural than that of the standard PSO.