| The lower limb exoskeleton robot,as a human-centered human-machine collaborative intelligent system,needs to analyze the human motion state information to enhance the human-machine collaborative compliance.Surface Electromyography(s EMG)is one of the main physiological signals of the human body.Because it contains human motion information,it is widely used in the field of human-machine collaboration.In order to improve human-machine coordination,the mainstream method is to complete mapping and decoding through s EMG signals and joint angles.At present,some achievements have been made in the angle prediction of the lower limbs of the human body,but the prediction accuracy needs to be improved,and the research on the angle control based on the prediction of the lower limbs is currently carried out less.The s EMG signal of the lower limbs contains complete motion information and is generated in advance of the motion,this paper conducts a research on the man-machine fusion control strategy based on the human lower extremity s EMG signal.The specific research includes the following aspects:(1)Design the mechanical structure of the lower limb exoskeleton to assist the robot,complete the adjustable lower limb exoskeleton robot based on ergonomics and human bionics structure design,and build a platform for the establishment of a three-layer human-machine coordination system of state perception,motion prediction,and trajectory execution.(2)In order to enable the state perception layer to obtain the s EMG signal with less redundant information and its effective characteristics,this paper first designs a data acquisition scheme,and uses a multi-channel adaptive threshold algorithm to detect the active segment of the s EMG signal to complete the online collection of human body motion state information.Then select the three most effective muscles of the human body's lower limbs according to Spearman's coefficient to reduce data redundancy.Finally,the preprocessing of the collected signal is completed,and the noise is reduced by Butterworth band-pass filtering,and then based on theoretical analysis and experimental verification,a VMD is found that is the most suitable for lower limb angle prediction feature extraction algorithm.(3)Aiming at the problem of achieving high-precision multi-joint continuous motion angle estimation in the motion prediction layer,this paper uses Spearman's correlation coefficient to perform internal analysis on various feature extraction algorithms,and select several feature extraction algorithms with better results.Then select the traditional neural network BP,ELM and time series neural network TCN,NARX and the NARX-SSA neural network proposed in this article to construct the hip and knee regression models to complete the angle prediction.The root mean square error is used as a measure of accuracy.The experimental results It shows that the NARX-SSA model proposed in this paper has the best effect in the five sports modes,and has strong generalization,and the RMSE value is within 1.(4)Aiming at the problem of how to obtain accurate control signals in the trajectory execution layer,this paper proposes a lower limb joint variable parameter predictive matrix control strategy based on the predicted output angle.First,the predicted output angle is smoothed through the Butterworth low-pass filter,and the prediction is selected Output the effective extreme point of the angle,and then use this strategy to convert the predicted angle change of the human lower limb joint into the control matrix signal of the lower limb exoskeleton controller,and finally design an online experiment to complete the three-layer fusion of the exoskeleton robot based on the human lower limb s EMG signal in different motion modes Control,proved the effectiveness of the control strategy. |
PDF Full Text Request