Real-Time Drowsiness Detection

Driver drowsiness is a major cause of road acci- dents, especially during long-distance and night-time driving. Fatigue severely degrades reaction time, decision-making ability, and situational awareness, leading to life-threatening incidents. Existing detection techniques such as physiological sensing and vehicle behavior monitoring often suffer from intrusiveness or environmental sensitivity. This paper presents a real-time hybrid drowsiness detection system combining the speed of Haar Cascade classifiers with the accuracy of a Convolutional Neural Network (CNN) for open/closed eye-state classification. The system continuously monitors eye regions and evaluates prolonged eye closure using temporal thresholding to determine the onset of drowsiness. The lightweight design ensures real-time performance at 24–28 FPS on standard CPU hardware, achieving an accuracy of 96%. The proposed system effectively addresses limitations in classical EAR-based systems identified in previous research, including issues with illumination, spectacles reflection, and head movement. Experimental results demonstrate strong robustness and make the approach suitable for integration into low-cost ADAS systems.