Driving Fatigue State Detection Based On EEG And EOG

Fatigue driving is one of the important causes of traffic accidents,which poses a serious threat to social public safety and people’s lives and property.Therefore,it is of great significance to study fatigue detection methods to curb fatigue driving and improve the traffic environment in China.For a single detection standard,it is easy to be disturbed by changes in the external environment.At the same time,in order to further improve the accuracy and generalization of the model.Based on the fusion of EEG and EEG signals,machine learning methods such as deep learning and transfer learning are used to detect the driver’s fatigue state,and the complementary information of the two signal characteristics is fully explored.The research focuses on a series of core issues such as the fusion signal feature extraction method,EEG feature selection in different brain regions,and EEG feature selection.The main research contents of this article include:(1)EEG and EEG signal preprocessing: In order to reduce the complexity of the data,the original signal is down-sampled to 200Hz;the fast independent component analysis(FASTICA)algorithm is used to separate the forehead EEG signal from the forehead electrode signal;the EEG signal through a 0 to 75 Hz band-pass filter is used to remove artifacts and noise;the forehead electrooculogram signal is filtered through a 0 to 30 Hz band-pass filter.(2)EEG and EOG signal feature extraction: Using short-time Fourier transform(STFT)to extract 25-band and 5-band power spectral density(PSD)and differential entropy(DE)features,respectively,and the smoothed operation is performed on the extracted features by a linear dynamic system smoothing algorithm.The subtraction rule is used to separate the horizontal electrooculogram signal and the independent component analysis is used to separate the vertical electrooculogram signal.Using continuous wavelet transform and peak detection method to extract eye movement features such as blinking,saccade and gaze from horizontal and vertical electrooculogram signals.(3)Driving fatigue detection network transfer learning model: In order to improve the accuracy and generalization of driving fatigue detection methods,a onedimensional fully convolutional neural network is constructed for the effects of individual differences in the training of fatigue state detection models based on EEG and EEG signals.The network(1DCCNN)transfer learning model selects several relatively balanced sample of each state(wake,fatigue,drowsiness)to train a pretrained model,and then micro-calls the feature extraction part of the pretrained model to the training and testing of other individuals.Experimental verification was performed using the SEED-VIG fatigue alert estimation data set of the Shanghai Jiao Tong University Research Center for BrainFocused Computing and Machine Intelligence.Experiments were performed on the characteristics of EEG signals,EOG signals,and fusion EEG and EOG signals in different regions.The experimental results show that the one-dimensional fully convolutional neural network transfer learning model proposed in this paper has better classification ability.