Framework for Stress Detection Using Thermal Signature

Abstract

Autonomic nervous system (ANS) activity requires usage of contact sensors with patients' body. Computational psychophysiology based on thermal imaging is suggested as an alternative. It is a non-invasive and non-contact method that can be used for medical applications such as stress detection, human psychology, geriatric medicine, autonomic nervous activity, medical diagnostics and psychophysiology. It is free from pain and radiations. Very few works are reported to identify stress states at individual level. This work presents a framework to detect stress based on heart rate variability (HRV). Methods were proposed for extracting thermal signatures such as cardiac pulse, breath rate, sudomotor response, and stress response from various regions. Psychophysiological disorders are categorized as bradycardia, tachycardia, stress, and no stress. The system enables monitoring of thermal features at four facial areas such as forehead, neck, periorbital, and nose. The proposed system is tested on bench mark datasets and proved with high confidence w.r.t existing works and ground truth values.