Driver drowsiness is an important public safety issues. The difficulty in monitoring is that the driver's individual differences make the error rate of fatigue judgments of different individuals larger. Nine steering wheel variables and three reaction time variables were measured as participants drove a fixed road course in a high fidelity motion-based driving simulator. Drowsiness levels of participants were judged by the expert scoring method combined with the driver's self-describing method using the Stanford Sleepiness Scale. For each driver, a Hidden Markov Model (HMM) was trained and recognition of a given observation sequence was performed to score each drowsiness driving model. Ellipse, NMRHOLD, PNS, SW_Range_2, RT1 and RT2 were selected through One-way repeated ANOVA as the characteristic of drowsiness. Sixteen HMMs were identified by data sequences of 16 drivers and the accuracy rate was 87.5%. It is recommended that HMM could be used to detect driver drowsiness and address the inaccuracies caused by individual differences.

CICTP