Aiming at the problems of premature convergence, poor ability in global optimization and low accuracy in local optimization for support vector machine ( SVM) based on parameter optimization, an adaptive particle swarm optimization / cuckoo ( APSO/ CS ) parameter optimization method is proposed, in which the optimization of kernel function parameters and penalty factors in SVM model is realized. The optimization performance of APSO/ CS, APSO and CS is compared and analyzed by test functions, which shows that APSO/ CS can accelerate the convergence speed of local and global optimization. The gesture recognition methods based on surface electromyography signal ( sEMG) by APSO/ CS, APSO and CS are compared. The experiment results show that the SVM method optimized by APSO/ CS can realize the highest recognition accuracy, which is about 94. 50%. The proposed method can provide a new way for the recognition and classification algorithm.