Research On Signal Processing And Attention Identification Based On Single Channel EEG

Measuring attention level has many benefits in medicine,monitoring,detection,driving,gaming and other aspects.Attention levels can be used to identify conditions such as attention deficit hyperactivity disorder(ADHD)in children to evaluate the reliability of neural feedback therapy,which could provide a treatment for some brain disorders.Attention level can also be used to evaluate driver fatigue,and the corresponding arousal feedback mechanism can effectively avoid the occurrence of accidents caused by driver fatigue.Therefore,it is of great significance to establish an attention model.It is a popular research direction in brain computer interface(BCI)system to use EEG signal to study the physiological mechanism of attention,measure its level and detect its cognitive function.How to accurately and effectively improve the classification accuracy of Electroencephalogram(EEG)related to attention is the key problem of this study.Firstly,in order to ensure the reliability of data,this study uses a single channel EEG signal device to collect the data set,and designs the experimental paradigm related to attention.This scheme focuses on the attention level in the process of three different tasks,and collects data from multiple subjects for analysis and verification.Secondly,there is a large interference of Electrooculogram(EOG)in the original EEG signals,which will affect the classification of subsequent attention tasks.Therefore,this thesis proposes an ocular artifact removal algorithm based on single channel.In this scheme,first of all,original EEG signal is decomposed into the EEG components and EOG components by separation technology.Secondly,the EOG components were further separated by wavelet threshold denoising technology.Finally,all EEG components are reconstructed.Compared with other algorithms,this algorithm has remarkable performance and retains EEG signals to the maximum extent.Then,Feature extraction was carried out on EEG signals after the removal of ocular artifacts.In this thesis,the rhythm signals are obtained by wavelet packet transform,and then the statistical characteristics such as energy,mean value and power spectrum characteristics of the rhythm signals are obtained,and the sample entropy of EEG signals is calculated.Finally,a sparse autoencoder network model is used to classify the attention related EEG signals.Compared with other EEG signal classification models of single channel,simulation results show that the network model can effectively improve the accuracy of classification,and the accuracy can reach 97.9%.In conclusion,the research result are helpful to better analyze different attention states through EEG signals,and provide theoretical support for practical applications such as fatigued driving and electronic teaching.