To improve the lower limb surface ElectroMyoGraphic (EMG) gait recognition accuracy and real time performance, this paper deals with a pattern recognition method for optimizing the Support Vector Machine (SVM) by using the Particle Swarm Optimization (PSO) algorithm. Firstly, the values of Integrated EMG and variance are extracted as the feature samples from the de-noised EMG signals. Then, the SVM parameters of the punishment and the kernel function are optimized by PSO. Finally, the constructed SVM classifiers are trained and tested by using the EMG sample data of the gait movements. The experimental results show that for five normal walking gaits of the lower extremity, the recognition rate of the PSO-SVM classifier is significantly higher than that of the non-parameter-optimized SVM classifier, and the average recognition rate is up to 97.8%, as well as the classification accuracy and self-adaptability are also improved.