Research On Lower Limb Movement Recognition Based On Eeg And Emg Signals

The number of people with lower-limb movement disorders in our country exceeds 83.5 million,and lower limb movement disorders have seriously affected the normal life of patients.The traditional lower extremity exoskeleton walking aid robot completes the walking aid function through mechanical button control,which requires the patient to actively adapt to the exoskeleton robot,and the mechanical assist mode is likely to cause secondary injury to the patient.With the rapid development of neuroelectric signal detection technology,combining sensors and motion intention recognition algorithms to obtain human motion information can provide an efficient control strategy for lower extremity exoskeleton robots.The brain is the sender of exercise instructions,and EEG signals can represent the human movement intention with global planning;muscles are the performers of exercise actions,and the EMG signals can represent natural movement patterns.information.The EEG and EMG signals of the human are collected when standing still,walking at a constant speed,going up and down stairs,and going up and down slopes through experiments.Based on EEG signals,whether the lower limbs of the human have the intention to move is explored,and the movement types of the lower limbs are recognized based on the EMG signals.Try to fuse EEG and EMG signals to recognize the movement of lower limbs.Lower limb movement intention recognition based on the EEG signals is studied,blind source separation algorithm and wavelet threshold denoising method are used to remove interference noise in the EEG signals;extract wavelet coefficients and energy value features in the frequency band related to motion to construct EEG features vector and the features is visualized;comparing K Nearest Neighbor,Support Vector Machine,Linear Discriminant Analysis and Decision Tree classifiers to the recognition results,the Support Vector Machine has the best classification results,reaching an average recognition accuracy of 87.50%.The ensemble learning methods are used to study the movement pattern recognition of the lower limbs based on the EMG signal,the feature analysis methods of time domain,frequency domain,time-frequency domain and time sequence complexity are used to extract features of EMG signals.Feature visualization shows that the four analysis methods have their own advantages in characterizing different actions;comparing the four kinds of ensemble learning methods,including Random Forest,Adaptive Boosting,Gradient Boosting Decision Tree and Extreme Gradient Boosting,to the recognition results,and the Extreme Gradient Boosting method has the best EMG classification results,reaching an average recognition accuracy of 99.11%.In addition,an EMG classification model that recognizes the static state is constructed,five different motion modes of the lower limbs and the interference action for the interference action of the lower limbs.The recognition accuracy rate reaches 99.01%,and it has strong anti-interference performance.EEG and EMG signals are integrated to recognize the movement state of lower limbs.The fusion of EEG and EMG features to recognize the movement intention of the human lower limbs through feature-level fusion,compared with a single EEG feature,the recognition accuracy is increased by2.73%.The decision-level fusion method is used to fuse EEG and EMG signal recognition results to improve the reliability of lower limb movement pattern recognition.The use of EEG and EMG signals analysis technology to perceive and recognize human lower limb movement intentions and movement patterns as the interactive control signal of the lower limb exoskeleton rehabilitation robot can better ensure the safety of the training process,and at the same time,training based on active rehabilitation is helpful.Remodeling the neurological function of stroke patients and improving the rehabilitation effect have high application value.