Study On The Classification Of EEG Signals Based On LM Algorithm

Brain Computer Interface(BCI)is a communication system that does not depend on external nerves and muscles,which can directly transforms EEG signals into control commands to control the operation of external devices.In the world,a lot of people lose their exercise ability due to all kinds of accidents.Most disabled people have limited activity space.They usually only have activities in beds and wheelchairs.The birth of the BCI provides alternative means for them to communicate with the outside.In addition,the BCI is widely used in the fields of military,brain science and so on.In recent years,it has become a hot spot of artificial intelligence.At present,the main research direction of brain computer interface is steady visual evoked potential,P300 potential,slow cortex potential and motion imagination,among which the first three kinds of brain signals are evoked by external stimuli,such as visual stimulation and somatosensory stimulation which induce the computer to produce certain regular signals,then deal with the evoked signals.And classification.Sports imagination is the signal that the human body spontaneously produces the intention of action,which is of practical significance,in which the research team of the University of Graz science and technology in Austria is the pioneer in this field.Today,BCI system based on sports thinking mode has been applied to life,such as brain control racing and brain control games.This paper is the research direction of motion imagination.This paper focuses on the brain machine interface system based on the model of motion imagination and analyzes the classification of the EEG signals of the four kinds of motion imaginary tasks.The research focuses on EEG signal acquisition,eigenvalue extraction,pattern classification and so on.First,taking the 2008 BCI competition signal acquisition mode as the experimental paradigm,using Emotive Epoc+to collect EEG signals and filtering the EEG signal,then using the principal component analysis to extract the eigenvalues of the EEG signal,based on the shortcomings and limitations of the standard BP algorithm,the modified algorithm Levenberg Marquardt(LM)method is used respectively.The adaptive LR gradient method and the momentum gradient descent method are used to classify and compare the extracted signal eigenvalues in three ways.Finally,design algorithm verification scheme,design interactive interface based on MATLAB GUI platform,write callback function,verify the feasibility of the algorithm through the Bluetooth serial port and the Arduino intelligent vehicle link verification algorithm.For signal classification,the average error of the adaptive LR gradient method and the momentum gradient descent method for the EEG signal classification is 0.0207 and 0.0328,and the classification accuracy is 52%and 49%respectively.The average error of LM algorithm is 5.6306×10~-77 and the accuracy of classification is 86%.The experimental results show that the LM algorithm has the best classification effect,the communication between the designed serial communication interface and Arduino car is good,which lays the foundation for further research of online brain computer interface system.