Research On Lower Limb With Normal Gait Joint Toque Calculation Based On EMG Fusion Model

The calculation of joint torque in the process of human motion is an important part of the research field of human biomechanics.Obtaining accurate and reliable joint torque is of great significance for the development of medical diagnosis,rehabilitation training,human-computer interaction and other fields.Flexible exoskeleton is an important branch of man-machine interactive in recent years.Its main goal is to provide effective power for human motion and reduce energy consumption in motion.Its assistance to human body is torque assistance in essence,so it is a very challenging research topic to obtain the torque curve of lower limb joint through human-computer interaction device as a reference for assistance strategy in the research of flexible wearable assistance device.According to this application background,this project is expected to provide a stable,accurate and online optimization method of lower limb joint torque in non laboratory environment.The main contents of this paper can be summarized as follows.1)This paper analyzes the advantages and disadvantages of the existing methods for calculating the joint torque of the lower limbs,and discusses the main components of the neuromusculoskeletal model.According to the physiological basis,the basic hypothesis is added to the model,the simplified neuromusculoskeletal model of human body under normal gait is established,the mapping relationship between s EMG signal and muscle force is established,and the main influencing parameters are discussed.2)A decoding method of single electrode s EMG signal for normal gait phase division task is proposed.This paper discusses the composition of surface electromyography(s EMG)signal and muscle synergy,and uses filtering and non negative matrix decomposition to decode s EMG signal to obtain spinal nerve unit activation which controls muscle synergy.The spatial and temporal information of spinal nerve activation was extracted,and the classification model was selected to divide the gait phase.3)The joint torque calculation method of non landing state inverse dynamics is introduced to establish the fusion physiological model.The LSTM network was established to estimate the physiological parameters of the model.4)Bayesian optimization network is used to identify the parameters and input coefficients of the fusion model,improve the accuracy of gait phase division,and adjust the super parameters of LSTM network.The online updating optimization method is proposed,and the accuracy of the model in single gait,multi gait,online walking and other states is verified.The main contribution of this paper is to solve the problem that the neuromusculoskeletal model can not be updated online.A practical joint torque calculation method is proposed,which improves the accuracy of multi gait estimation of lower limb joint torque in normal gait.A new decoding method of s EMG signal mapping to central nervous system command is proposed,which improves the accuracy of phase division of lower limb gait using s EMG signal.It provides a more effective reference for the power strategy of flexible power equipment.