融合SSA-GA算法优化XGBoost的sEMG手势识别方法OA

This paper proposes a hybrid sparrow search algorithm and genetic algorithm-optimized XGBoost model for gesture recognition to address the low accuracy of surface electromyography(sEMG)gesture recognition.First,sEMG signals are preprocessed using notch filtering and Butterworth filtering.Then,an optimal subset of features is selected by SSA,and the classifier hyperparameters are optimized by GA,enabling rapid localization of high-quality parameter regions followed by fine-tuning.The optimized XGBoost model is subsequently applied for gesture classification.Experimental results show the prediction performance of the SSA-GA-optimized XGBoost model is markedly superior to other methods,achieving an average recognition accuracy of 94.37%,up by 3.2%,1.8%,and 2.1%compared with that of the conventional XGBoost,SSA-XGBoost,and GA-XGBoost methods.The proposed approach possesses enhanced anti-overfitting ability,stronger generalization capability,and improved robustness.