Detection and prediction of drowsiness by reflexive eye movements

Abstract

A reliable predictor of drowsiness using objective measures is desirable for machine and vehicle operations in which human errors may cause fatal accidents. We have evaluated the Vestibulo-Ocular Reflex (VOR) as a possible predictor of drowsiness. The VOR is a compensatory eye movement that stabilizes retinal image during head motion, and is inevitably induced by vibration in a car running on the road. We employed an uneventful driving simulation (DS) featuring vibration stimulation to induce both drowsiness and VOR in healthy human subjects. VOR performance was characterized by its gain and variability, and evaluated in relation to the subjects' drowsiness. A significant decrease in VOR gain and increase in variability accompanied subjective sleepiness, with the changes occurring before subjects became aware of sleepiness. From this finding, we developed a reliable method (88.9% accuracy) to predict oncoming sleepiness using changes in VOR performance as a cue.