Driver Drowsiness Detection with an Alarm System using a Webcam
Keywords:Microsleep, Blinking, Yawning, 68 key points face landmark, Eye Aspect Ratio, Mouth Aspect Ratio
Microsleep has contributed to 100,000 crashes, 71,000 injuries, and 1,550 fatalities on average every year. Blinking slowly and yawning frequently are the common symptoms of a drowsy driver, and they happen mostly when lighting conditions are at their worst between midnight and early morning. The aim of this proposed project is to design a driver drowsiness detection system based on eye and mouth behaviour that works with various light ranges. The system is integrated with an internal web camera and speaker as input and output tools, respectively. The drowsiness state can be determined by Eye Aspect Ratio (EAR) and Mouth Aspect Ratio (MAR) algorithms that run in real-time using OpenCV and dlib 68 key points for facial landmarks. The project flow for this system is face analysis, followed by eye blinking and yawning detections that run concurrently after the web camera live streams the driver’s face. An alarm sound will be played once the driver is detected as being drowsy. This system can detect the driver's drowsiness with or without the presence of eyeglasses in both ideal and poor lighting conditions. Experiments carried out throughout the study discovered that the proposed system has an accuracy range of 85% to 95%.