Integrated physiological signal-based biomarkers for automatic stress detection

Abstract

Numerous mental intentions and life rates are responsible for dispensing mental stress. It’s an essential purpose behind delivering numerous cardiovascular ailments. Identifying and addressing the effect of stress on the creation of automatic identification of different levels of mental stress thus provides a crucial path for progressive research to tackle stress. This paper presents an investigation on mental stress identification with the guide of preparing the Electrocardiogram (ECG), Galvanic Skin Response (GSR) chronicles utilizing Hjorth Parameters, Autoregressive, Shannon entropy and few other features that were used in finding best features using Wrapper and BorutaShap function. The primary reason for this experiment was to evoke particular affective states in the participants. ECG, GSR recordings of 40 people while watching short videos was used from AMIGOS Dataset. Random Forest Classifier(RF), Logistic Regression(LR) and K-Nearest Neighbor Classifier(KNN) are used to detect stress and have achieved an accuracy of 82.23%, 79.84%, 78.48% respectively. These results can be used to make a device capable of identifying and measuring the stress levels experienced by individuals, so that stress can be better managed as short-term or long-term stress still poses a risk of harm to physiological and mental health.