A Real-Time Embedded System for Driver Drowsiness Detection Based on Visual Analysis of the Eyes and Mouth Using...

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning
• Automotive Safety

Background:

• Increasing vehicle numbers correlate with rising accident rates.
• Driver drowsiness is a significant human factor contributing to vehicular accidents.
• Timely detection and alerts are crucial for mitigating drowsiness-related incidents.

Purpose of the Study:

• To develop and evaluate a Convolutional Neural Network (CNN)-based system for real-time driver drowsiness detection.
• To analyze eye region and Mouth Aspect Ratio (MAR) for detecting drowsiness and yawning.
• To propose a novel Driver Drowsiness Artificial Intelligence (DD-AI) architecture.

Main Methods:

• Optimized endpoint delineation for eye region of interest (ROI) extraction.
• Utilized an NVIDIA Jetson Nano device with a near-infrared (NIR) camera for real-time analysis.
• Trained and tested models using the Night-Time Yawning-Microsleep-Eyeblink-Driver Distraction (NITYMED) dataset.
• Compared the proposed DD-AI architecture against InceptionV3, VGG16, and ResNet50V2.

Main Results:

• The proposed DD-AI network achieved a high accuracy of 99.88% on the NITYMED test data.
• DD-AI demonstrated superior performance compared to InceptionV3, VGG16, and ResNet50V2.
• In real-world hardware implementation, DD-AI achieved 96.55% accuracy at an average of 14 frames per second (fps).

Conclusions:

• The DD-AI architecture offers a highly accurate and efficient solution for real-time driver drowsiness detection.
• The system's performance in real-world conditions validates its potential for enhancing road safety.
• This CNN-based approach effectively addresses the critical issue of driver fatigue.