Evaluation of driver drowsiness based on respiratory metrics

Area of Science:

• Physiology
• Biomedical Engineering
• Human-Computer Interaction

Background:

• The transition from alertness to drowsiness significantly alters respiratory patterns.
• These respiratory changes present a potential avenue for detecting driver drowsiness.

Purpose of the Study:

• To identify specific respiratory features indicative of driver drowsiness.
• To develop a classification system for drowsiness and alertness levels using these features.

Main Methods:

• Eight respiratory features (e.g., respiration rate, inspiration/expiration durations) were extracted from abdominal motion signals using a force-sensing belt.
• Data was collected from 20 male students using a driving simulator.
• Statistical significance was assessed at the 0.05 level.

Main Results:

• All eight extracted respiratory features were statistically significant.
• Driver drowsiness was detected with high performance metrics: 88% accuracy, 82% precision, 86% recall, and 90% F1 score.

Conclusions:

• Respiratory signal analysis offers a promising, less intrusive method for driver drowsiness detection.
• Findings support the development of driver monitoring systems, especially for automated driving applications.