CNN-based Drowsiness Detection with Alarm System to Prevent Microsleep

According to data from the World Health Organization, approximately 1.25 million people die every year from road traffic accidents. One of the underlying factors that often cause traffic accidents is drowsiness. It has a significant impact on performance, alertness, memory, concentration, and reaction speed which is crucial when driving. Drowsiness could further lead to microsleep which is very dangerous when driving. This study proposes a convolution neural network (CNN)-based system to monitor the driver's eyes and mouth separately to identify drowsiness. An alarm will be sounded if drowsiness is detected and therefore prevent microsleep. The result shows that the proposed system can predict closed eyes with 76% accuracy and 74% precision. In comparison, it achieved 62% accuracy and 65% precision for yawn detection.