Abstract:

The characteristics of electroencephalograph (EEG) parameters among drowsy drivers were investigated to define physiologic index of driving fatigue. Quantitative EEG (qEEG) analysis techniques were used among 22 subjects using a driving simulator to extract δ band power and 90% spectral edge frequency (SEF90) and to analyze the parameters' changes during drowsy driving and the differences from various recording locations by using binary logistic regression, discriminant classification, and receiver operating characteristic (ROC) curve. The results show that 60 minutes’ mono- environment driving lets to driving drowsiness with increase of relative δ band power and decrease of SEF90. The EEG quantitative changes recorded at the temporal regions are more than that recorded at the frontal regions. Therefore, the increase of normalized δ band power can be used as an indicator to discriminate the alert from drowsy state during driving. Decrease of SEM90 can be another indicator to determine the drowsiness during driving.

Key words: automotive safety, quantitative electroencephalography (qEEG), drowsy driving, binary logistic regression, receiver operating characteristic (ROC) curve, 90% spectral edge frequency