Driver's Fatigue Recognition Algorithm Based on EEG and Its Validity Verification

In order to recognize the driver's fatigue state effectively, a method of driving fatigue state identification was constructed based on electroencephalogram (EEG). Firstly, combined with the driver's subjective indicators, driving behavior performance was taken as an objective evaluation index, to verify the rationality of the different levels of fatigue. Then, three bands of average amplitude and five synthetic indicators were chosen as characteristic indexes after the EEG signals were analyzed by fast Fourier transform (FFT).Meanwhile constructing fatigue recognition bispectral index through kernel principal component analysis (KPCA), a driver fatigue state recognition model was proposed with the support vector machine (SVM). Finally, 2 hours continual driving EEG data was collected from 30 drivers to test the model. Result show that the recognition accuracy rate is between 79.17%-92.03%, and the average accuracy rate is 84.62%, which proves the validity of the model.