Driver Drowsiness Detection

Abstract

The increase in road accidents caused by driver fatigue has necessitated the development of real-time driver drowsiness detection systems. This paper presents a hybrid solution for detecting driver drowsiness using facial landmark analysis and alternative vehicle behavior monitoring techniques. The system utilizes OpenCV and MediaPipe for real-time video processing to analyze Eye Aspect Ratio (EAR) and Mouth Aspect Ratio (MAR) for detecting prolonged eye closure and yawning, respectively. In scenarios where facial detection fails due to poor lighting or obstructions, the system monitors sudden vehicle speed changes and harsh braking as alternative indicators of drowsiness. The proposed solution triggers timely audio-visual alerts to prevent accidents and logs driver behavior for future analysis. This paper discusses the implementation, testing, and performance analysis of the system, highlighting its accuracy, reliability, and real-world applicability.