Research On Gait Recognition And Prediction Of Exoskeleton Robot Based On DTW Template Matching

Although the research of exoskeleton robots has lasted for more than 50 years,it is still in the stage of research and development.As we all know,exoskeleton robots have been widely used in military,medical and industrial fields.Exoskeleton robot technology is a comprehensive technology that integrates sensing,control,information,fusion and mobile computing.It provides a wearable mechanical mechanism that can follow human behavior steadily for the operator.Therefore,in order to better realize the follow-up control of exoskeleton robots,it is necessary to accurately recognize the human gait phase and match the human motion intention in real time.At the same time,due to the delay in the process of obtaining gait information by sensors,in order to achieve the real-time follow-up control,the gait should be accurately predicted.The gait of a person changes periodically during walking,and the information of the plantar pressure and the angle of the joint of the lower limb are very important in analyzing the gait of the human body.Firstly,the thesis designs a set of gait data measurement system,which communicates with each sensor board through RS485 to collect real-time gait information of human body in the process of movement.Then,through the comparison of the plantar pressure information and the joint angle information,we find that they have a certain correlation,and ultimately determine that the plantar pressure information is the main information for gait analysis and research.By comparing the four gait recognition schemes of Euclidean distance method,threshold method,support vector machine and proportional method,the thesis uses the support vector machine to adopt one-to-one method to identify the gait of six models.In the thesis,the human motion intentions are mainly divided into five gait states: stand,upstair,downstair,uphill,downhill,and the DTW algorithm for speech isolated word recognition is used to recognize human motion intention here.Then the DTW algorithm is improved to complete the matching of human motion intention.Finally,the two algorithms of interpolation and BP neural network are designed to predict the gait data.The prediction of gait can offset the delay of the exoskeleton robot caused by the delay of the system,so that the exoskeleton robot can follow the human real-time follow-up and control it in real time.In the end,the recognition of gait phase,the matching of human motion intention and the prediction of gait data are realized.The experimental results verify that the exoskeleton robot can accurately achieve the above identification,matching and prediction.