Research On Motion Intention Recognition And Interactive Control Method Of Multi Degree Of Freedom Upper Limb Exoskeleton Rehabilitation Robot

The recovery of upper limb motor function in stroke patients is slow,so rehabilitation training based on neuroplasticity is of great significance.Compared with the traditional training,the rehabilitation robot which combines the robot technology and the rehabilitation medical technology can guarantee the intensity,effect and precision of the rehabilitation training.The rehabilitation process of patients can be divided into soft paralysis period and rehabilitation period.With the assistance of the upper limb rehabilitation robot,the patient underwent passive training and active training respectively in these two stages.Studies have shown that rehabilitation training in which the patient exercises actively with the intention to participate is more effective in the remodeling of the nervous system and the recovery of motor function,whereas EEG signal is generated by the motor center of the cerebral cortex,is the most direct embodiment of the motor intention of human body.Therefore,six movements commonly used in upper limb movement were selected in this paper to design a 6-DOF exoskeleton-type upper limb rehabilitation robot,and EEG signals were collected to distinguish active motion intentions and guide patients to conduct active and passive training.Due to the instability and complex external interference of the upper limb dynamics system,in this paper,a human-computer interactive control system based on EEG and interactive force is designed,which is suitable for active and passive training modes under the condition of Electroencephalogram(EEG)and interactive force,and Accurately Track the target’s training trajectory.The major areas of work are as follows:(1)Bioelectrical signals are used as the driving data source of the upper limb rehabilitation robot.After studying the generation mechanism and characteristics of EEG and EMG signals,a collection system is designed to synchronously collect them.Six commonly used movements of upper limbs were selected as training modes and bioelectrical signals from relevant parts are collected.EEG signals were preprocessed by band pass filtering,artifact removal and wavelet threshold denoising.In the pretreatment process of EMG signal,notch processing of power frequency interference is added.The preprocessed EEG signal is extracted by discrete wavelet transform.After preprocessing the EMG signal,the rehabilitation factor representing the degree of upper limb rehabilitation is calculated by the formula.(2)Aiming at the problem of low classification accuracy of EEG signals,Gabor transformation is introduced to transform the traditional convolutional neural network,and Gabor function is used to replace the traditional random convolutional kernel in the convolutional layer to extract richer features.On this basis,the classification model is optimized to improve the generalization ability of the model,avoid over-fitting phenomenon as far as possible,and improve the judgment level of classifier.Compared with traditional classifier,the superiority of GAB-CNN model in motion recognition is verified.(3)A human-computer interaction control system based on EEG and interaction force is designed,which is suitable for two training modes.The motion track synchronized with motion intention is taken as the tracking target of the control system.A position-based impedance control theory was used for the whole,and a fuzzy impedance controller was used for the outer ring.The fuzzy rules were related to the degree of muscle recovery.The output impedance trajectory was transmitted to an adaptive iterative learning controller for the inner ring for trajectory tracking,and the Laypunov stability of the two controllers is analyzed.(4)In order to verify the above design method,the 6-DOF upper limb rehabilitation robot virtual prototype model was established by selecting three joints of shoulder,elbow and wrist with six degrees of freedom.Then,the designed virtual prototype was imported into ADAMS,set some parameters,create a simulation experiment based on ADAMS and MATLAB,experiment the above system,and analyze the results.