GMR-118 Stress Detection by Wearable Devices: Integrating Multimodal Physiological Signals and Machine Learning Techniques

GMR-118 Stress Detection by Wearable Devices: Integrating Multimodal Physiological Signals and Machine Learning Techniques

https://www.kennesaw.edu/ccse/events/computing-showcase/sp24-cday-program.php

In the medical field, the most common complaint of patients is “stress”. Stress can cause severe effects on the human body. For example, prolonged mental stress can cause serious health issues in long term such as hypertension, cardiovascular diseases, increased susceptibility to infections, and depression. These health issues can be prevented by early detection of stress and by taking preventive measures. The most common detecting stress was determined from the questionnaires or from the interactive sessions conducted to assess the people's affective state. However, this method is not highly reliable and can be biased depending on the person who is conducting the session. To address this issue, detecting stress from physiological values such as electrocardiogram, blood volume pulse and body temperature can be an effective solution. When a person is under stress, the sympathetic nervous system (SNS) of the person triggers a physiological response that leads to a change in the heart rate, and muscle tension. Hence, understanding physiological values can be an effective way to detect the state of mind. To understand the effect of physiological values on the state of the art, we implemented a stress detection method by using the publicly available “Wearable Stress and Affect Detection” (WESAD) dataset, which has physiological data collected from 15 subjects. This data is collected from the wrist-worn and chest-worn sensors. These physiological values include blood volume pulse, electrocardiogram, electrodermal activity, electromyogram, respiration, body temperature and three-axis acceleration. In this research, we used wrist-worn sensor data to identify the state of mind. To achieve this task, we implemented classification machine learning models with the help of Logistic Regression, Decision Tree and Random Forest machine learning algorithms. Further, we implemented Stacking Ensemble Learning (SEL).