The assessment of driver's arousal states from the classification of eye-blink patterns

Abstract

To realize the real-time assessment of driver's arousal states, we propose the assessment method based on the analysis of eye-blink characteristics form image sequences. The driver's arousal level while driving is not monotonous falling from high to low. We proposed the two-dimensional arousal states transition model which was taken into account the fact that a driver usually held out against sleepiness. The eye-blink pattern categories were classified from image sequence using HMM (Hidden Markov Model), then the driver's arousal states were finally assessed using HMM by histogram distribution of those typical eye-blink categories. The arousal assessment results are also verified against the rating results by trained raters. © 2009 Springer Berlin Heidelberg.