Research On Channel Selection And Multi-class Feature Fusion Of Eeg Signal For Mental Fatigue Detection

With the rapid development of modern social science and technology,people are facing increasing pressure in their daily work and study.Prolonged or high-intensity use of the brain can easily cause mental fatigue,which has become a sub-healthy problem that seriously affects people's normal lives.Electroencephalogram(EEG)signals directly reflect changes in the central nervous system.Using EEG signals to evaluate mental fatigue is a hot topic in related neighborhood research.At present,most of the researches on mental fatigue detection based on EEG signals are based on32 and 64 channel analysis.The amount of data is large,the calculation is complicated and the processing time is long,which is not conducive to online detection and is not conducive to the portability of device.In addition,most studies use a feature which the effective information is single or simply concatenate multiple features for classification.The former affects the effect of fatigue detection,and the latter increases the feature dimension and computational complexity.Aiming at the above problems,based on the research of the Relief channel selection algorithm and the sparse representation feature fusion algorithm,this paper designs an on-line mental fatigue detection system based on EEG signals for motor imagery tasks.The main research work is as follows:(1)Selection of Common EEG Channels in Fatigue Based on Relief and Single-channel Performance Analysis.Aiming at the problems of large amount of multi-channel data,high computational complexity,and long processing time,the weight value of each channel under a single feature is calculated by the Relief algorithm,and the classification accuracy rate of each channel under the corresponding feature is analyzed,combined with each single channel the weighted sum of the accuracy of each channel is used to obtain the weight value of each channel,and the highest ranked channel is selected as the common channel according to the weight value ranking.A common channel selection method based on Relief algorithm and single channel performance analysis is proposed.Experimental results show that the method can effectively reduce the amount of data and the time of data processing on the basis of ensuring the accuracy of classification.(2)Multi-feature fusion based on statistical difference analysis and sparse representation.Aiming at the problem that the selected feature caused the loss of effective information and the multi-feature direct stitching caused the high feature dimension.Based on the extraction of multiple types of features in the frequency and time domains,the difference level of each feature under different fatigue states is obtained through statistically significant difference analysis.After selecting the features with significant differences,multi-class feature fusion is performed by sparse representation method to obtain sparse multi-class fusion features.Experimental results show that this method effectively reduces the feature dimension and computational complexity and improves the classification accuracy of fatigue EEG signals.(3)Build an online brain fatigue detection system based on EEG signals.This system receives the EEG signal of the subject's motion imaging task collected by the UE-16 B EEG amplifier through UDP communication,and performs feature extraction and classification processing on the EEG signal through MATLAB software.The subject's brain fatigue status was dynamically fed back to the subject and the system.A motion imagination game that dynamically adjusted the difficulty based on the results of brain fatigue detection was designed and completed to verify the effectiveness of the channel selection and feature fusion methods proposed in this paper.