Multi-Modal Acute Stress Recognition Using Off-the-Shelf Wearable Devices.

Area of Science:

• Physiological monitoring
• Machine learning applications
• Wearable technology

Background:

• Stress and emotion monitoring are crucial for applications like high-risk missions and mental health.
• Off-the-shelf wearable sensors offer a promising avenue for non-invasive physiological data acquisition.
• Existing methods for stress detection often rely on specialized equipment or controlled environments.

Purpose of the Study:

• To evaluate the feasibility of monitoring stress and emotions using readily available wearable sensors.
• To propose and validate a multi-modal machine learning approach for detecting acute stress episodes.
• To investigate the individual contributions of different wearable sensors to stress detection accuracy.

Main Methods:

• Acquisition of physiological signals using Shimmer3 ECG Unit and Empatica E4 wristband.
• Development of a multi-modal machine learning technique fusing data from multiple biosignals and sensors.
• Experimental evaluation of the proposed technique on an unseen test set for acute stress recognition.

Main Results:

• The multi-modal machine learning approach achieved an accuracy of 84.13% in detecting acute stress episodes.
• Sensor-fusion techniques significantly improved the accuracy of stress detection compared to individual sensors.
• Demonstrated the practical possibility of recognizing acute stress using common wearable devices.

Conclusions:

• Wearable sensors, when combined with multi-modal machine learning and sensor fusion, can effectively detect acute stress.
• This approach holds significant potential for real-time stress monitoring in various practical applications.
• Further research can explore the integration of more diverse biosignals and advanced machine learning algorithms.