Human stress detection using EEG signals

Abstract

Stress is a problem that is widespread throughout the world, dubbed by the World Health Organization as the Health Epidemic of the 21st Century. Although it is natural to have a certain level of stress due to the challenges we face on a daily basis, stress is not only caused by factors external to the individual, in many cases it is related to the way we interact with the environment and the internal processes involved. In Europe alone, more than 50% of workers and students suffer from stress. This Master Thesis aimed to detect stress and relaxation states from EEG data acquired with the Neuroelectrics Enobio device and contrast them with other physiological signals, in this case the electrocardiogram (ECG) and galvanic skin response (GSR) acquired with the Biopac MP36 system. While previous studies with EEG focus on predicting stress with this signal alone, this project aims to contrast prediction results with EEG data and other physiological data (ECG and EDA), acquired simultaneously in the same experiment, to assess whether this combination provides a significant benefit that the signals do not provide separately. At the same time, the performance of different Machine Learning and Deep Learning techniques were investigated to corroborate which one is the most suitable to achieve the proposed goals. The models used were LightGBM, a 16-layer CNN, a KNN with Grid Search for Hyperparameter Tuning and a non-linear SVM