针对模糊神经网络训练采用BP算法比较依赖于网络的初始条件,训练时间较长,容易陷入局部极值的缺点,利用粒子群优化算法(PSO)的全局搜索性能,将PSO用于模糊神经网络的训练过程.由于基本PSO算法存在一定的早熟收敛问题,引入一种自适应动态改变惯性因子的PSO算法,使算法具有较强的全局搜索能力.将此算法训练的模糊神经网络应用于语音识别中,结果表明,与BP算法相比,粒子群优化的模糊神经网络具有较高的收敛速度和识别率. In order to overcome the shortcomings of BP neural network training using BP algorithm, the PSO is used in fuzzy neural network Training process.Because of the premature convergence problem of the basic PSO algorithm, a PSO algorithm adaptively changing the inertia factor is introduced to make the algorithm have a strong global search ability.Fuzzy neural network trained by this algorithm is applied to speech recognition The results show that compared with BP algorithm, PSO-based fuzzy neural network has higher convergence rate and recognition rate.