Identification Of EEG Signals Based On Motor Imaginary

Electroencephalography(EEG)is often used in modern Brain Computer Interface(BCI)systems.BCI technology is a technology that connects the human brain with some electronic devices,so that the brain can directly communicate with electronic devices Human-Computer Interaction Technology.In general,for patients with irreversible nerve damage,motor imaging brain-computer interface technology can enable them to control equipment such as prosthetics.Although the patient's body does not have real activities,they can rely on their brain to imagine themselves Of his body is moving,that is,moving through his imagination.In fact,the measurement of EEG signals is more complicated,because the equipment for collecting EEG signals is composed of multiple electrodes,so different EEG signals are measured in different electrode channels,and then they are integrated together.Therefore,the collection of EEG signals is often disturbed,such as objects or noise in the surrounding environment.Therefore,for the acquisition of motor imaging EEG signals,there may be other signals mixed in it,then it is accurately classified It is very important and can help more patients with nerve damage.This paper is divided into three parts.The first part is to introduce the characteristics of EEG signals,using the original EEG data set from BCI competition III.The four kinds of EEG signals collected in this data set are left and right hand motion EEG,tongue and foot motion EEG.In this paper,each kind of EEG in EEG data set is labeled and stored,then preprocessed,and the original EEG is reconstructed.In this paper,EEG is preprocessed and reconstructed based on wavelet transform,and EEG signals are analyzed by time domain analysis,frequency domain analysis and spatial domain analysis.In the second part,based on the common space pattern method(CSP)and its derivative algorithm,the filter bank common space pattern method(FBCSP)is used to extract the features of the reconstructed EEG signals.The wavelet packet is introduced to improve the fbcsp,and the wavelet packet based common space pattern method(WPCSP)is designed to extract the features of EEG,and the three feature extraction methods CSP,FBCSP and WPCSP are used to extract the features of EEG respectively Feature extraction and feature calculation.The last part is to identify the four classifications of the four types of motion imaging EEG after feature extraction.The classifier used in this paper is an improved decision tree support vector machine(SVM).Because traditional SVM is only suitable for two classifications,this This article studies the problem of four classifications,so SVM needs to be improved,and the traditional SVM is improved to a classifier that can perform four classifications by using the nature of decision trees.Finally,the control variable method is used to identify three different feature extraction algorithms based on the above,namely CSP,FBCSP,and WPCSP.The decision tree SVM classifier is used to identify the Kappa coefficient and the accuracy rate Po after analysis.