In order to overcome the shortcomings of the particle swarm optimization(PSO), an improved particle swarm optimization based on simultaneous perturbation stochastic approximation(SPSA) method is proposed. It embeds SPSA into PSO as a local search operator in the proper time, and makes use of the computing resources available in the optimization process. Numerical experiments for benchmark functions have been done, The results indicate that the proposed algorithm performs better than the existing ones in terms of efficiency, accuracy and stability.