Running Gait Recognition And Knee Prosthesis Control For Lower Limb Amputees

Running is one of the most basic human activities.For prosthetic wearers,running exercise helps to restore body functions for sports,which can effectively prevent residual limb muscle from atrophying.Prosthesis with running function can serve the overall development of amputees and meet the needs of amputees for the diversification and intelligentialize of prosthetic function.Therefore,it is of great significance to develop the knee prosthesis with running function.This thesis mainly includes three aspects: data acquisition,running state and gait phase recognition,control strategy of prosthesis knee joint movement.First of all,the running pattern of human body is analyzed,which realizes the detailed division of running gait cycle.According to the characteristics of lower limb running movement,the system lower limb movement information acquisition is built,which realizes the collection of lower limb prostheses movement information.The combination of wavelet packet hard and soft threshold denoising method and Chebyshev I filter method is used to filter the information of different lower limb running motion,which provides an effective data basis for gait recognition of prosthetic wearers.Secondly,the method of correlation analysis is used to judge the state of running motions.A multi-classification model based on the support vector machines(SVM)binary tree is designed.The processed data is composed of multi-feature vector as the input of SVM,which solves the problem of the gait recognition for lower limb amputees.To further improve the recognition rate of running gait,PSO was used to optimize classification model parameters.The running gait recognition accuracy rate is improved to 92.78%.The local optimization of SVM kernel parameter selection is eliminated effectively.In this thesis,the algorithm of PSO-SVM is compared with the typical multi-classification model based on BP neural network.Experiment results show that the recognition correct rate is 92.78%,which is higher than that of SVM,traditional BP neural networks and PSO-BP neural networks.Finally,the four-link prosthetic knee joint mechanism is acted as a controlled object on the based of the recognition of running gait.The finite state machine(FSM)is applied to the control of prosthetic knee joint,and a motion control model of the knee joint was developed.The key points of the running gait is taken as the transfer conditions to change the running state.And the keen joint running action of the healthy people is taken as the target intention to control the scheduled turning of the prosthetic knee joint.This thesis provides a certain basis for the dynamic lower limb prostheses to coordinate with the human body to achieve running movement.