Feature Augmented Hybrid CNN for Stress Recognition Using Wrist-based Photoplethysmography Sensor

Area of Science:

• Biomedical Engineering
• Machine Learning
• Wearable Technology

Background:

• Stress negatively impacts mental and physical health, with increased prevalence during the COVID-19 pandemic.
• Continuous stress monitoring is crucial for proactive health management.
• Wrist-worn smartwatches with photoplethysmography (PPG) sensors offer a convenient platform for physiological signal monitoring.

Approach:

• This paper proposes a novel hybrid Convolutional Neural Network (H-CNN) classifier for stress detection using Blood Volume Pulse (BVP) signals from wrist-based PPG sensors.
• The H-CNN integrates both hand-crafted features and features automatically extracted by CNNs.

Key Points:

• The H-CNN model demonstrated superior performance on the WESAD dataset compared to traditional classifiers and standard CNNs.
• For 3-class classification (Baseline vs. Stress vs. Amusement), H-CNN achieved higher accuracy and macro F1 scores.
• For 2-class classification (Stress vs. Non-stress), H-CNN also showed significant improvements in accuracy and macro F1 scores.

Conclusions:

• The proposed H-CNN classifier effectively detects stress using PPG signals from wrist-worn devices.
• Hybrid feature extraction enhances the accuracy of stress detection models.
• This approach shows promise for application in consumer-grade smartwatches for stress monitoring.