Multi-object intergroup gesture recognition combined with fusion feature and KNN algorithm

SEMG signal is a bioelectrical signal produced by the contraction of human surface muscles. Human-computer interaction based on SEMG signal is of great significance in the field of rehabilitation robots. In this study, a feature extraction method of SEMG signal based on activated muscle regionis proposed, which is based on the study of activated muscle regionin human forearm and hand movement. At the same time, the main research object of this study is the multi-object intergroup SEMG signal which is closer to the practical application environment. The new feature extracted is fused with the sample entropy feature and the wavelength feature to obtain better signal features. After combining the fusion feature with KNN algorithm, the hand motion pattern recognition and classification between multi-object groups is carried out. The combination of the fusion feature and KNN classification algorithm can achieve 91.05% in the multi-object intergroup hand motion classification. This method has lower computational cost without expensive hardware support, and improves the robustness of hand motion recognition based on EMG signals.