| The number of lower limb amputations in China has exceeded one million.At present,the commercial lower limb prostheses put into use on a large scale in the domestic market are still passive lower limb prostheses.Passive prostheses will bring a series of problems such as more energy consumption and asymmetrical gait patterns.With the development of human-computer interaction technology,the focus of lower limb prosthesis research has shifted from passive lower limb prosthesis to intelligent dynamic lower limb prosthesis.However,the vast majority of intelligent powered lower limb prostheses currently on the market still have a series of problems in the field of how to automatically and accurately identify human lower limb movement intentions,such as high development cost and low intention recognition rate,which reduces the coordination between lower limb prostheses and wearers and human-computer interaction experience.In order to reduce the development cost and realize the coordinated movement of prosthetic limb and amputee,this paper developed a low-cost and high-stability human intention recognition sensor system with lower limb prosthetic limb as the research carrier,and proposed an improved human intention recognition algorithm.Through theoretical analysis,algorithm structure design and experimental verification,the reliability of the designed sensor system and the proposed human intention algorithm is verified.The main research contents include:Firstly,by comparing the advantages and disadvantages of current mainstream motion recognition information sources,analyzing the motion mechanism and characteristics of human lower limbs during motion,a motion intention recognition sensor system for lower limb prostheses was designed based on this theoretical support.The sensor system includes an inertial measurement unit,an angle sensor,and a pressure sensor;In order to verify the effectiveness and safety of the system,a set of motion intention recognition sensor system for healthy people was built as a pre experiment and a comparative experiment.Secondly,six healthy volunteers and five patients with knee amputations were recruited to conduct a motion information collection experiment.The experiment collected motion data from nine typical motion modes,and processed the motion data,including data noise reduction,windowing processing,data feature value extraction,and normalization processing.Then,machine learning algorithm is used to recognize human intention.LDA,SVM,ELM and CNN algorithms were used to model train and identify the data of pre-processed healthy volunteers,and the results of four different algorithms were compared.After comprehensive consideration,The ELM algorithm which is more in line with the actual project is selected as the key research object.Finally,the validity of the motion intention recognition sensor system is verified using the data of patients with knee amputation,and a NELM human intention recognition algorithm is proposed.Taking the motion data of the knee amputee as the root node and the ELM algorithm as the evaluation standard of the decision node,a decision tree model is designed to recognize 9 kinds of motion patterns hierarchically.At the same time,the Sparrow Search Algorithm optimization algorithm is used to optimize the parameters of the local optimal solution caused by the arbitrary generation of ELM input weight and hidden layer bias.The experimental results show that the average recognition accuracy of the NELM algorithm for lower limb movement intention recognition is 98.6%,and the Macro-F1 value is 0.986,which greatly improves the effect of human intention recognition and improves the human-computer interaction experience between the prosthesis and the knee amputee.The algorithm proposed in this study can more accurately identify the movement intention of human lower limbs.The research in this paper can provide theoretical basis and technical support for the product development of intelligent lower limb dynamic prostheses in China. |
PDF Full Text Request