Continuous Motion Estimation Of Lower Limb Joint Angle Based On SEMG

Surface electromyography(sEMG)contains a lot of physiological information of the human body.The use of surface electromyography to identify human body movement intentions has been widely used in recent years.This thesis presents a regression model prediction method for continuous movement of the human lower limb joints under different motion modes.The collected surface electromyographic signals are used to predict the angles of the human knee and ankle joints.The design of a patient rehabilitation training control system is of great significance.In this experiment,eight leg muscles related to human lower limb joint movements were selected for experiments.First,the surface electromyographic signals of the selected muscles and the joint angle signals of the knee and ankle joints were collected simultaneously,and the denoised signals were extracted.The time domain characteristics of wavelet low-frequency coefficients,then three different neural network models are established for training and prediction,and the root-mean-square error(RMSE)is used to evaluate the prediction effect.The main research contents of this article are as follows:(1)The EMG signal and joint angle signal were collected synchronously,and the experiment action was designed and standardized.For knee joint and ankle joint,four surface electromyography signals related to their movements are collected as signal sources,and the joint movement is divided into three different movement modes according to whether there is load and speed.(2)A wavelet denoising method is designed to denoise the original EMG signal.Symlets7 is selected as the wavelet basis function,and a good denoising effect is achieved.In this paper,a feature extraction method of sEMG based on RMS combination of wavelet low-frequency coefficients is proposed.SEMG is decomposed into three layers of wavelet,and the RMS of the original signal,the RMS of each layer of wavelet low-frequency coefficients and the RMS combination of three layers of wavelet coefficients are compared.(3)A joint motion estimation method based on muscle cooperative analysis is proposed,which combines muscle cooperative model with NMF algorithm,and uses neural network model based on muscle cooperative analysis to predict and analyze the data.According to the selected features,BP neural network and GA-BP neural network are designed to predict the joint angle in the process of knee joint extension and flexion,as well as ankle joint back extension and plantar flexion.(4)The experimental results show that the GA-BP neural network prediction model based on muscle cooperative analysis and the three-layer wavelet coefficient RMS combination have the best prediction accuracy in the low-speed and no load motion mode.