Analysis Of EMG Signal During Upper Limb Movement

Following the continuous socio-economic development of China,the increasing quality of life of the people and the ageing of our society,the number of patients with functional impairment of the upper limbs due to factors such as stroke,traumatic brain injury and spinal cord injury is rapidly increasing.Functional impairment of the upper limbs affects the independence of the patient’s daily life to a great extent and even causes lifelong disability.According to neuroplasticity,patients can reshape the connection between the injured central nerve and the limb through post-operative rehabilitation training,gradually achieving effective control of the limb and reducing the rate and degree of disability.However,traditional rehabilitation training has problems such as long rehabilitation cycles,limited efficiency and high staff consumption.The design and development of exoskeletal rehabilitation robots can reduce the public health burden,reduce healthcare costs,improve rehabilitation quality and increase rehabilitation efficiency.The surface electromyographic(sEMG)signals are of great research value for control studies of upper limb assisted rehabilitation robots due to their noninvasive acquisition,simple processing,wireless transmission and provision of motion intent50-100 ms prior to actual movement.This paper addresses the design and development of an upper limb exoskeleton rehabilitation robot by conducting research on sEMG signals during upper limb motion: upper limb motion analysis provides a basis for the motion trajectory of the upper limb exoskeleton rehabilitation robot;sEMG signal research enables upper limb motion intent recognition.The main work of this paper is as follows.At first,upper limb motor EMG experiments and hand motor EMG experiments were designed using the Vicon motion capture system and the Noraxon Ultium wirelessEMG device.For the upper limb motion experiments,a whole-body model of the Vicon system was utilised in 20 healthy subjects(10 males,10 females),while eight right upper limb motion-related muscles were selected for upper limb motion capture of five daily upper limb movements and8-channel sEMG signal acquisition.For the hand motion experiment,20 healthy subjects(10males,10 females)were used to capture 6 daily hand movements and 6 channels of sEMG signal acquisition using the thumb and index finger models built in the Vicon system,while 6right small arm and hand muscles were selected.Simultaneous kinematic and sEMG signal data of upper limb movements and hand movements were obtained.In the second,upper limb movement analyses were performed.Based on the kinematic data obtained from the upper limb movement experiment,a movement analysis of the right upper limb joint was performed for five upper limb movements(drinking,raising the arm,touching the back pocket,touching the head and touching the opposite shoulder).The subjects’ kinematic data were intercepted and standardized to the effective motion interval.The angles and angular velocities of the shoulder,elbow and wrist joints of the right upper limb during different motion cycles were analyzed to provide joint motion trajectories for the design of the upper limb exoskeleton rehabilitation robot.One-way ANOVA was performed to analyse and extract the three feature values of Integral Electromyographic(i EMG),Median Frequency(MF)and Mean Frequency(MNF)of the sEMG signal in relation to different movements of the upper limb.The i EMG feature values of the eight upper limb movement-related muscles were found to be significantly different in each upper limb movement cycle,providing a basis for the prediction of subsequent upper limb movement patterns.In the third,sEMG signal analysis of hand motion was carried out,and hand gesture recognition based on sEMG signal was implemented.Based on the sEMG signal data obtained from hand motion experiments,the time-domain feature values of the sEMG signal were extracted,and the pattern recognition of six hand gestures,namely,rolling mouse,thumb inward,thumb-index finger pinch,four-finger bend,cup grip and five-finger pinch,was carried out by comparing Artificial Neural Network(ANN),Adaptive Boosting(Adaboost),K-Nearest neighbor(KNN),Support Vector Machine(SVM),Decision Tree(DT),Random Forest(RF),Gradient Boosting Decision Tree(GBDT)and e Xtreme Gradient Boosting(XGBoost)multiple classifiers.Based on the kinematic data obtained from hand motion experiments,the angles of the five joints of the thumb and index finger were calculated for the six gestures.The ANN classifier had the best prediction result with 97.9% prediction accuracy and 0.975 kappa coefficient for the test set.Using the small arm and hand sEMG signals for gesture recognition,it was possible to achieve classification prediction results with almost identical results,demonstrating the feasibility of sEMG signal gesture recognition applied to upper limb exoskeleton hand control.Finally,the effect of sex differences on the sEMG signal-based hand gesture recognition algorithm was considered.In this paper,a one-way ANOVA for the sex factor of the sEMG signal was conducted,demonstrating a significant sex difference(p<0.05)for the same muscle pair in the right hand.Further,two algorithms considering sex differences in hand gesture recognition(differentiating the sex dataset and adding a sex label)were proposed for application to three machine learning algorithms: KNN,SVM and ANN.This was complemented by t-tests and 5-fold cross-validation to verify the impact of considering sex differences on improving classification performance.The results have shown that considering sex differences can significantly improve the classification performance of hand gesture recognition,with the ANN algorithm with the addition of the sex label achieving the highest hand gesture recognition prediction accuracy of 98.4%.It is demonstrated that hand motion recognition algorithms that consider sex differences can be applied to control systems for prosthetic hands or exoskeletons.In conclusion,this paper conducted a study on upper limb motion,analyzed the upper limb kinematic change patterns and sEMG signal eigenvalue differences during five daily upper limb motion cycles;conducted kinematic data analysis of six hand gestures and gesture recognition based on sEMG signal,and demonstrated that the sEMG signal gesture recognition algorithm considering gender differences can significantly improve the gesture recognition prediction accuracy.This study provides a kinematic basis for the design of the upper limb rehabilitation robot and demonstrates the feasibility of applying sEMG signals to the control system of the upper to rehabilitation robot.