Date of Award
5-2025
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Biomedical Engineering and Sciences
First Advisor
Peshala Thibbotuwawa Gamage
Second Advisor
Mehmet Kaya
Third Advisor
Kenia Pedrosa Nunes
Fourth Advisor
Linxia Gu
Abstract
This research investigates multimodal physiological signal analysis for real-time stress classification, with a focus on identifying biomarkers suitable for wearable implementation. A standardized experimental protocol was developed involving rest, stress, guided meditation, and recovery phases. Participants completed cognitive stress tasks such as the Stroop test and mental arithmetic under controlled conditions while Electroencephalography (EEG), Electrocardiography (ECG), and Electromyography (EMG) signals were simultaneously recorded. Sessions were designed to isolate the physiological effects of acute stress and meditation across repeated trials. Advanced signal processing techniques were applied to extract relevant features for classification, and statistical analysis was conducted to assess significance.
The study evaluates HRV and trapezius muscle EMG as practical alternatives to EEG. Using both subject-specific and generalized machine learning models, HRV and EMG showed comparable accuracy to EEG, supporting their potential for continuous stress monitoring.
Stress responses were further analyzed in emergency room healthcare providers versus non-healthcare controls. While both groups followed similar stress response trends, ER providers exhibited stronger and faster physiological responses and quicker recovery, underscoring the importance of population-specific stress monitoring.
Additionally, guided meditation was examined as a stress-reduction strategy. Preliminary results from a small sample showed improvements in reaction time and accuracy, suggesting potential benefits that warrant further validation in larger cohorts.
Novel EMG features based on frequency bands and optimized sensor placement were introduced, demonstrating strong potential for stress classification.
The findings advance personalized, signal-based approaches for practical stress monitoring and lay the groundwork for developing efficient, accessible, and minimally obtrusive systems suitable for real-world deployment.
Recommended Citation
Ahmed, Mohammad Haroon, "Advancing Wearable Stress Monitoring: Multimodal Analysis of EEG, HRV, and EMG for Real-Time and Personalized Stress Detection" (2025). Theses and Dissertations. 1558.
https://repository.fit.edu/etd/1558
Included in
Bioelectrical and Neuroengineering Commons, Biomedical Devices and Instrumentation Commons, Cognitive Neuroscience Commons, Signal Processing Commons
