EEG-based vigilance analysis by using fisher score and PCA algorithm

Driving fatigue is the most dangerous killer on the highway. Supervising mental vigilance is able to warn the driver and avoid some disasters. The current study mainly focuses on the power spectrum. The electroencephalography (EEG) activities in the δ(0-4 Hz), θ(4-8 Hz), α(8-13 Hz) and β(13-35Hz) bands, reflect the change of the physiological vigilance. The ratios of (θ + α)/β, α/β, (θ + α)/(α + β), and θ/β, are also used for assessing the vigilance. We make use of PCA algorithm and fisher score to remove background noise and select the significant discriminative features. After that, the sleepy and wakeful selected data are trained and tested by SVM classifier to evaluate the vigilance levels. Compared with the result obtained from non-PCA algorithm, the classification result from PCA algorithm, achieves higher accuracy in the α and β bands, as well as at the ratios of (θ + α)/β and α/β. The significant difference in the cerebral cortex appears at the δ, α and β bands, as well as at the ratios of α/β, (θ + α)/(α + β) and θ/β. These results suggest that estimating vigilance levels is feasible. Measuring the vigilance precisely is helpful to keep drivers away from some dangerous driving behaviors in the proposed system.