| Abstract |
| To improve the safety of take-over requests (TOR) in Level 3 autonomous driving, this study introduces a non-contact advanced driver monitoring system that avoids reactive, behavior-based approaches, in which an in-vehicle RGBW camera measures the driver’s heart rate via remote photoplethysmography (rPPG). The system then analyzes the driver’s heart rate variability to classify their real-time arousal state (drowsy, stressed, or normal) and employs a deep learning model for face detection, advanced signal processing, and a pattern recognition algorithm for state classification. This system was validated in over 105 hours of real-world driving with 28 participants, achieving 85.14% heart rate accuracy compared to an ECG and 90.81% state classification accuracy. This study is expected to enhance TOR safety by confirming that physiological monitoring provides reliable metrics to assess driver readiness. |
|
|
| Key Words |
| Driver Monitoring, Drowsiness Warning, Takeover Request, Facial Recognition, Heart Rate Measurement, Heart Rate Variability(HRV), 모니터링, 졸음경고, 운전이양권, 얼굴인식, 심박수 측정, 심박변이도 |
|
|
 |
|
