Early Detection of Driver Drowsiness Detection using Automated Deep Artificial Intelligence Learning (ADAI)

Abstract

Early detection of driver drowsiness significantly contributes to road accidents worldwide. This study presents a novel Driver Drowsiness Detector (DDD) system powered by Automated Deep Artificial Intelligence Learning (ADAI) to address this issue and improve road safety. Created by merging open pose and Gaze Tracking Networks (GTN), the Driver Drowsiness Detector system provides a comprehensive method for drowsiness identification. Sensor-based modules collect data from various vehicle sensors, such as steering wheel movements and accelerator/brake pedal dynamics. In contrast, vision-based modules use a front-facing camera-based monitoring system for the driver's facial expression, different types of eye movements, and head posture. The pre-trained model, such as VGG-19, can be fine-tuned for specific drowsiness detection tasks to leverage the knowledge learned from large-scale datasets. Diverse driving scenarios encompass lighting conditions, weather conditions, and driver demographics. The dataset is labeled with alert and drowsy states to enable supervised learning. The proposed DDD model learns to extract meaningful features from visual and sensor data, allowing it to detect drowsiness in real time. When signs of drowsiness are detected, the DDD system can provide real-time alerts to drivers, assisting them in remaining alert and focused on the road. This system contributes to a safer road environment and a reduction in road accidents by reducing the occurrence of drowsy driving incidents.