Classification And Recognition Of Motor Imagery EEG Signals Of Brain Controlled Robot

In recent years,due to the aging population and increasing accidents,the number of patients with motor dysfunction of lower limbs caused by stroke,hemiplegia and autism is also increasing year by year.In order to avoid secondary injury to the patients' muscles and nervous system and realize the patients' motor thinking intention,it is urgent to establish a new way of interaction between the human brain and the outside world.The brain controlled robot uses the EEG signal of human brain in thinking to control the movement,assist the human body to complete the action of human imagination,and realize the interaction and control between brain and external intelligent devices.Therefore,the research of brain controlled robot has important social significance and broad application prospects.It is the key technology and difficulty to analyze the detected EEG signal and identify the human motion intention in brain controlled robot.In this paper,the spontaneous left-hand and right-hand motor imagery EEG signal is taken as the research object,and the Hilbert Huang transform and common space pattern are used to extract its features,and then the random forest classifier is used for pattern recognition.Finally,a robot is programmed to realize the control of the robot arm motion by motor imagery EEG signal.The main contents of this paper are as follows.1.EEG signal characteristic analysis and filtering.By analyzing the characteristics of human EEG,the EEG signals related to motor imagery are mainly concentrated in 8-30 hz.In order to filter the interference of other frequency signals when detecting EEG,an FIR band-pass filter is designed.2.Multimodal feature extraction.Compared with the single feature extraction method,this paper proposes a feature extraction method based on Hilbert Huang transform(HHT)and common spatial pattern(CSP)to extract multiple features in timefrequency spatial domain from filtered EEG signals.Firstly,multiple narrow-band signals IMF with physical meaning are decomposed by empirical mode decomposition,and the IMF components are analyzed by Hilbert spectrum to obtain the characteristics in time domain and frequency domain;then spatial filter is constructed to filter EEG signals in spatial domain,and then spatial features are extracted by CSP decomposition;finally,the features in time domain,frequency domain and spatial domain are fused to form a new feature matrix.3.Classification and comparative analysis.In this paper,random forest is used for pattern classification and recognition of EEG signals.Because the decision tree generated by the general random forest algorithm has many leaves and large size,this paper also uses an improved decision tree generation algorithm,which makes the decision tree leaves significantly less.This paper classifies the feature vectors extracted by HHT,CSP and time-frequency-space fusion.Finally,the classification results show that,compared with the other two methods,the feature extraction method proposed in this paper improves 10% and 7% respectively,and can significantly improve the classification and recognition rate of left and right hand motor imagery EEG signals.4.Design of EEG control robot control system.The NAO humanoid robot is used as the control object to build the experimental platform.The source code of control program such as signal processing is edited by chorographe software and python language,and the EEG signal is input to realize the control of the robot arm action.The experimental results show that this method can finally realize the motion control of the robot arm by using the movement imagination EEG signal.