Real-Time Machine Learning-Based Driver Drowsiness Detection Using Visual Features

Area of Science:

• Computer Science
• Artificial Intelligence
• Road Safety

Background:

• Drowsiness-related car accidents pose a significant threat to road safety.
• Early detection and alerting of drowsy drivers can prevent numerous accidents.

Purpose of the Study:

• To develop a non-invasive system for real-time driver drowsiness detection.
• To utilize visual features extracted from facial landmarks and head pose for accurate detection.

Main Methods:

• A system employing facial landmarks and face mesh detectors to extract visual features.
• Extraction of mouth aspect ratio, eye aspect ratio, and head pose features.
• Classification using random forest, sequential neural network, and linear support vector machine models.

Main Results:

• The proposed system achieved high accuracy in detecting drowsy drivers.
• Evaluations on the National Tsing Hua University dataset demonstrated up to 99% accuracy.
• Successful real-time detection and alarming of drowsy drivers.

Conclusions:

• The developed system offers an effective solution for real-time driver drowsiness detection.
• Visual feature-based analysis combined with machine learning classifiers can significantly improve road safety.
• The non-invasive approach provides a practical method for driver monitoring.