Related Experiment Video
Updated: Jan 11, 2026
06:37
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
Published on: December 15, 2023
5.2K
Drowsiness Classification in Young Drivers Based on Facial Near-Infrared Images Using a Convolutional Neural Network:
Ayaka Nomura1, Atsushi Yoshida1, Takumi Torii1
1Department of Electrical Engineering and Electronics, Aoyama Gakuin University, Sagamihara 252-5258, Japan.
Sensors (Basel, Switzerland)
|November 13, 2025
Summary
Near-infrared imaging shows promise for detecting drowsy driving. This technology uses facial images to identify drowsiness, potentially improving road safety by detecting alertness declines earlier than current methods.
Area of Science:
- Biomedical Engineering
- Computer Vision
- Transportation Safety
Background:
- Drowsy driving is a significant global cause of traffic accidents.
- Current driver monitoring systems (DMS) rely on behavioral cues, which manifest after alertness has already decreased.
Purpose of the Study:
- To investigate facial near-infrared (NIR) imaging as a physiological indicator for early drowsiness detection.
- To assess the feasibility of using 940 nm NIR facial images for drowsiness classification.
Main Methods:
- A pilot study used young adult participants and single-frame NIR facial images at 940 nm.
- A convolutional neural network (CNN) was trained to classify drowsiness levels.
- Gradient-weighted Class Activation Mapping (Grad-CAM) was used for region interpretation.
Main Results:
- Drowsiness classification using 940 nm NIR images achieved approximately 90% accuracy in binary classification.
- Grad-CAM identified the nasal dorsum as a key region for drowsiness detection, aligning with physiological signs.
Conclusions:
- Facial NIR imaging at 940 nm is a feasible method for classifying drowsiness in young drivers.
- This approach offers a potential advancement for driver monitoring systems and road safety.
