Research On Brain-exoskeleton Interface Pattern Based On Deep Learning And Ensemble Learning

The upper limb exoskeleton device is a wearable robot.Exoskeletons' actions and the users' "consciousnesses" should be the same.Human action is generated by the brain and performed by the muscles and bones,so the use of the human EEG signal decoding can be directly connected to the human "consciousness" and the external implementation of the device.EEG signals are generated by the human brain activity,is a specific manifestation of human consciousness activities.The application of this technique to the human body control of the upper limb exoskeleton device can effectively solve the problem of consistency between the action of the upper limb exoskeleton and the user's intention.In this paper,we propose a technique of EEG analysis based on deep learning.Using the original EEG signal directly to classify,simplifying the feature extraction process,and achieved AUC value of 0.9761 high accuracy classification effect.This paper proposes a fast training EEG feature classification method based on ensemble learning technology,which uses motion-related cortical potentials and neural oscillations of different frequencies as features,and obtains more than any one classifier accuracy rate in the same training time Of the classification algorithm.In this paper,a simulation system of brain-machine interface of upper limb exoskeleton robot is established,and a network node of Restful architecture mode is designed.The hardware,communication protocol and software components including information collection part,signal transmission part,online processing part and host computer control part are designed.The system can process the EEG signal online and convert the EEG signal into the machine instructions,realize the human body's control function to the upper limb exoskeleton and display the running status in real time through the simulation interface.