Assessment of Driver's Stress using Multimodal Biosignals and Regularized Deep Kernel Learning.

Area of Science:

• Physiological computing
• Machine learning for driver monitoring
• Human-computer interaction

Background:

• Driver stress impacts road safety and requires objective monitoring.
• Physiological signals offer a promising avenue for assessing driver stress levels.
• Current methods for stress detection in drivers have limitations.

Purpose of the Study:

• To classify driver stress states using multimodal physiological signals.
• To evaluate the effectiveness of regularized deep kernel learning for stress detection.
• To identify the most reliable physiological signals for stress classification in driving scenarios.

Main Methods:

• Acquired electrocardiography (ECG), electrodermal activity (EDA), photoplethysmography (PPG), and respiration rate (RESP) from 10 drivers in a simulated environment.
• Extracted time and frequency features from physiological signals.
• Employed a fusion framework with regularized deep kernel learning, including intermediate fusion and subsequent classification using Support Vector Machine (SVM) and Random Forest (RF).

Main Results:

• The proposed multimodal approach successfully discriminated between different driver stress states.
• The combination of PPG and ECG signals with a Random Forest classifier achieved the highest F1-score of 0.97.
• Individual signal analysis showed PPG alone with RF yielded a maximum F1-score of 0.90, indicating the importance of signal selection.

Conclusions:

• Multimodal physiological signals, particularly ECG and PPG, are reliable indicators for classifying driver stress.
• The developed regularized deep kernel learning framework demonstrates potential for real-time stress assessment in driving.
• Subject-specific cross-validation enhances the performance of stress classification models.