Recognition of Psychological Stress Levels Using Wearable Biosensors

抄録

Remote work has spread rapidly since the pandemic of COVID-19, making it more convenient to work from home for a large number of people. However, excessive work concentration can lead to serious health damage such as mental stress if not monitored. In this paper, we propose a deep learning-based scheme aimed at identifying the different stress levels caused by concentration while working on a PC. We perform a set of experimental tests designed to provoke different levels of stress, namely low, medium, high, and no stress. Dedicated wearable devices are used to collect Heart Rate Variability (HRV), Galvanic Skin Reaction (GSR), and Breath Chest Force (BCF) to train and test a CNN-based stress classifier. The result shows the ability of the proposed CNN model to correctly classify stress levels achieving an accuracy of 99.08%.