A Personalized Feature Extraction Method for Hand Gesture Classification of sEMG Signals

To realize artificial limb control, it is an important method to use surface electromyography (sEMG) signals to recognize gestures. The sEMG signals are electromyographic signals collected from the surface layer of the skin, which can reflect the neuromuscular activities to a certain extent. Due to its random and non-smooth characteristics, the use of sEMG signals to control artificial intelligent prosthetic limbs has never been able to achieve the desired results, especially reflected in the influence of individual differences such as gender. In this paper, an energy-based gesture recognition method for surface EMG signals is proposed and validated on the Ninapro dataset through gender difference analysis. The experimental results show that the recognition accuracy of the method is significantly improved in all four classifiers, with an accuracy of 97.06%. Compared with other methods, the classification accuracy is effectively improved.