Disclaimer: The work showcased in the repository is a miniaturized version of the original. The sole purpose of the repo is to prove the credibility of the research work carried out.
Stress plays an integral role in influencing one’s decision making capability, attention span, learning, and problemsolving capacity. Stress detection and modeling has been an active area of research in the fields of psychology and computer science in recent times. Psychologists quantify stress using affective states, which is the experience of feeling the underlying emotional state. Most of the work in classifying human stress was achieved using user-dependent models, incapable of generalizing to a new user. This causes a new user to spend a significant amount of their time in training the model to predict their affective states. In this paper, the authors propose a User-Independent classification model for human stress detection where a new user requires no prerequisite calibration of their affective state. The classification of affective states were carried out on the publicly available dataset WESAD to confirm our hypotheses. The authors of this paper demonstrate User-Independent models for Bi-affective state classification (Stress vs Non-Stress) and Tri-affective state classification (Stress vs Amusement vs Baseline) cases in addition to providing a novel classification model for the Multi-affective state classification case (Stress vs Amusement vs Baseline vs Meditation). The models achieved a classification accuracy of 95% in the Bi-affective state, 85% in Tri-affective state and 83% in the Multi-affective state cases.
The Paper has been published in IEEE Conference "Intelligent Systems'2020"IEE'IS 20. Please cite us if you use them. (https://ieeexplore.ieee.org/document/9199928)
Authors: Ramanathan Murugappan,Joish J Bosco Vineeth Vijayaraghavan