Assessment of Driver Fatigue and Drowsiness Based on Eye Blink Rate

Abstract

Drowsy and fatigued driving is a major factor in many traffic accidents. The slow onset of drowsiness or extreme fatigue in a driver can be detected, although it is more difficult to do so than it is to detect closed eyes. We propose a novel yet simple single camera-based real-time computer vision technique for detecting drowsiness and fatigue levels that solely relies on the eye blinking rate estimated from the eye aspect ratio and moving average calculation over a period of 30 s which is updated every 10 s. An alert is generated to stop the user from going into microsleep or caution the user in case of extreme fatigue if the rate of eye blinking falls below a level or is too high respectively that has been scientifically validated in the literature. The existing methods use facial expression-based detection, blink-based detection using Electrooculogram or simple eye aspect ratio-based methods or calculation of blink rate in pixels/seconds or combination of these which only results in detection of drowsiness, while the proposed method uses single camera-based detection calculating blink rate to estimate both—the drowsiness and fatigue levels in blinks/minute; thus, the proposed method is much less complicated in terms of hardware and computation, and results are repeatable over different ambient illuminance conditions.