Term of Award

Fall 2023

Degree Name

Master of Science, Mechanical Engineering

Document Type and Release Option

Thesis (restricted to Georgia Southern)

Copyright Statement / License for Reuse

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.


Department of Mechanical Engineering

Committee Chair

Biswanath Samanta

Committee Member 1

Jinki Kim

Committee Member 2

Junghun Choi


This study aims to present a self-supervised contrastive learning (SSCL) model that will detect cognitive stress using physiological signals acquired from non-invasive wearable sensors. This study involved data collection using wearable sensors, feature extraction, normalization, and combination of data signals into a usable dataset, and the use of data augmentations, different contrastive machine learning techniques and algorithms to analyze the dataset and determine the best overall combination for assessment of mental stress from physiological data. A handful of volunteers participated in the study by completing different mental activities while wearing the Empatica E4 wristband to record their biometric signals. The key physiological signals recorded during this study were heart rate (HR), galvanic skin response (GSR), and blood volume pulse (BVP). The publicly accessible dataset Wearable Stress and Affect Detection (WESAD) was also used in this study. Both datasets were pre-processed and analyzed using a contrastive learning model to determine the combinations of the best data augmentation and SSCL models. These models achieved 97.90-99.95% accuracy. Results using the WESAD data indicated the potential for using data augmentation and SSCL to differentiate multiple emotional states. The study showed the potential of SSCL for the assessment of mental stress using physiological signals from wearable sensors.

OCLC Number


Research Data and Supplementary Material