Natural Grasping Movement Recognition And Force Estimation Using EMG

Human-computer interaction(HCI)based on electromyography(EMG)signal is a hot research topic at present.Collection and research of surface electromyography(sEMG)signal of arm muscle,which can help the disabled with motion assistance and limb rehabilitation.However,most of the current research about EMG focuses on gesture recognition,and few of them use EMG to decode natural grasping movements of hands.At the same time,most of the researches on natural grasping movements do not involve force information,and the decoding performance is limited.The object of research for this paper is four natural hand grasping movements : pinch,palmar,twist,and plug grasp.The core content of this research includes:(1)For natural grasping movements,the EMG experimental platform was designed,and data acquisition system was built.The subjects were recruited to carry out the experiment under different experimental paradigms for classification and force estimation of grasping movements.(2)The characteristics of EMG signal were analyzed,and the butterworth filter was used to selection and combination of ten characteristic of EMG signal.Meanwhile,the models were established for classification and regression.(3)Based on EMG of classification and force estimation of natural grasping movements,the experimental results were analyzed and discussed.For classification,the recognition accuracy for the four natural grasping movements involving force information was over 90%.In the research of motion parameter decoding for grasping movements,the recognition accuracy of grasping speed was more than 92.07%,and the recognition accuracy of grasping force class was more than 88.76%.For the continuous classification of grasping movements,the recognition accuracy could reach more than 91.76%.For force estimation,the regression determination coefficients of the four natural grasping movements were over 0.8044 for pinch grasp,over 0.8324 for palmar grasp,over 0.809 for twist grasp,and over 0.8621 for plug grasp,respectively.The influence of grasping speed on the regression performance was found,and slower grasping speed was beneficial to force estimation.The innovation of this paper is that,in contrast to the common EMG-based gesture classification,four natural hand grasping movements based on EMG were successfully decoded and the possibility of decoding these actions involving force was explored.The study on the force estimation of the four natural grasping actions was completed,confirming the feasibility of using EMG for grasping force prediction.The research in this paper has important implications for the natural control of myoelectric prostheses and provides new ideas for the implementation of force control based on myoelectric signals.The research on the force estimation of four natural grasping movements was completed and the feasibility of using EMG to predict force was confirmed.The research in this paper is of great significance for the natural control of EMG prosthesis and provides a new idea for the realization of force control based on EMG signal.