Stress Detection Through Wrist-Based Electrodermal Activity Monitoring and Machine Learning.

Area of Science:

• Biomedical Engineering
• Psychophysiology
• Machine Learning in Healthcare

Background:

• Stress significantly impacts modern life and mental health.
• Effective stress management requires accessible monitoring and support.
• Wearable technology offers novel solutions for physiological signal monitoring.

Purpose of the Study:

• To assess the feasibility of using wrist-based electrodermal activity (EDA) signals for stress prediction.
• To identify factors influencing stress classification accuracy using wearable data.
• To explore machine learning models for stress detection.

Main Methods:

• Collected wrist-worn electrodermal activity (EDA) data for stress and non-stress classification.
• Examined five machine learning classifiers, including Support Vector Machine (SVM).
• Evaluated classification performance across four EDA databases with varying feature selections.

Main Results:

• Support Vector Machine (SVM) demonstrated superior performance with 92.9% accuracy in stress prediction.
• Gender information influenced classification accuracy, showing significant differences between males and females.
• Multimodal approaches were explored for enhanced stress classification.

Conclusions:

• Wrist-based EDA signals from wearable devices show high potential for accurate stress monitoring.
• Machine learning, particularly SVM, is effective for classifying stress levels.
• Wearable technology can provide valuable insights for proactive mental health care.