Research On Lower Limb Rehabilitation Training Path Planning And Action Classification Based On Body Coordination Movement

In recent years,with the continuous aging of the population in China,the number of patients with movement disorders caused by stroke has been increasing,and the rehabilitation treatment of stroke patients has received widespread attention in the current society.Traditional stroke rehabilitation treatment faces problems such as insufficient rehabilitation physicians,tedious rehabilitation training process,insufficient consideration of patients’ subjective intentions,incomplete and subjective rehabilitation evaluation indicators.Rehabilitation exoskeleton robots,with their convenience and high efficiency,have been widely used to assist traditional rehabilitation treatments in stroke rehabilitation.Patients can improve their motor function recovery and neural recovery compensation by wearing rehabilitation exoskeleton robots and following them to complete corresponding rehabilitation movements.The movement path planning of lower limb rehabilitation exoskeleton robots based on limb coordination will have an important impact on the rehabilitation effect and experience of stroke patients.In addition,lower limb rehabilitation movement classification for stroke patients is particularly important for recognizing patients’ subjective intentions.In this paper,the path planning of human lower limb rehabilitation training and the movement classification method based on surface electromyography(sEMG)of lower limbs are researched,as follows:(1)For the problem of rehabilitation training path planning,a normalized output function of human lower limb characteristics is obtained by Fourier fitting based on the human motion characteristics output of the subject completing lower limb rehabilitation movements.By analyzing the consistency between the normalized output function and the joint angle error of the human body,it is demonstrated that the normalized output function is highly consistent with the actual rehabilitation movements of the human body.(2)For the problem of path not conforming to human ergonomics,the ergonomics of the human body is introduced,and the comfort index and complexity index are used as comprehensive rehabilitation indicators and constraints.The Golden Section algorithm is used to optimize the original path to obtain a rehabilitation planning path that conforms to human ergonomics.By analyzing the comprehensive rehabilitation indicators and joint angle errors before and after optimization,it is shown that the optimized path is more consistent with human ergonomics and still highly consistent with actual rehabilitation movements.(3)For the problem of lower limb rehabilitation movement classification,a dataset of six-channel sEMG signals for different subjects completing lower limb rehabilitation movements is constructed.The multi-layer perceptron(MLP)network and AlexNet network are utilized for same-subject and cross-subject lower limb movement classification experiments.By analyzing the accuracy of movement classification on the test set of the two neural networks,it is verified that the AlexNet network has a good effect on lower limb movement classification.(4)For the problem of high complexity of the classification network,knowledge distillation is used to compress the AlexNet network to achieve lightweight of the movement classification model.By analyzing the complexity of the neural network and the accuracy of movement classification on the test set before and after lightweight,it is shown that knowledge distillation can reduce the complexity of the network model,and the lightweight network model still has a good effect on movement classification.