The Design and Implementation of a Drowsiness Detection and Alarm System

Drowsy driving is a growing cause of traffic accidents with serious human and economic consequences, particularly for professional drivers. This paper presents the design and implementation of a real-time drowsiness detection and alarm system that uses a camera module and Google’s MediaPipe FaceLandmark model to monitor facial features. The system performs image preprocessing via histogram equalization and applies an algorithm based on Eye Aspect Ratio (EAR) and head pose estimation to detect four drowsy states: frequent blinking, prolonged eye closure, frequent head nodding, and head drooping. When drowsiness is detected, a buzzer connected to Raspberry Pi GPIO pins issues an audible warning, while avoiding alarms for brief, non-drowsy movements. Experimental implementation in a vehicle environment shows reliable detection of the predefined drowsy states and successful alarm activation. Beyond driving, the system is applicable to industrial safety, educational and office settings, healthcare monitoring, and home safety. Identified limitations include environment-dependent accuracy, possible false alarms, and user discomfort, which future work aims to mitigate through algorithm refinement, broader environment testing, and incorporation of user feedback.