Enhancing Road Safety: An Intelligent Drowsiness Detection System Based on Deep Neural Networks

The aim of this research is to create an intelligent vehicle system that can effectively mitigate the risks associated with drowsy driving, thereby preventing road accidents. Various factors like driver exhaustion, challenging road conditions, and adverse weather contribute to motor vehicle crashes, leading to over 1.35 million deaths annually, as estimated by entities such as the World Health Organization (WHO). To achieve this, the development of a robust warning system is crucial. In this study, we proposed an alert system to detect drowsy drivers. It utilizes a video stream process (VSP) to analyze the gaze's eye-aspect ratio (EAR) and geometric distance through an innate reflex model. Additionally, a facial landmark-based algorithm is employed for accurate eye recognition. When driver drowsiness is detected, the system issues a warning message, along with crash impact and location information, and activates the voice surveillance system. The suggested technique establishes a non-invasive approach for evaluating driver weariness in connection with accident severity. The camera, the fundamental element of this system, continuously acquires facial landmarks recognized by designated locations and analyzes them. Drivers who are distracted and have poor lighting may show signs of sleepiness.