Emotional stress detection using facial thermal imaging

Abstract Thermal imaging is a noninvasive method for assessing psychological states. This paper examines the features of physiological responses to emotional stress and their detection by thermal imaging. The correlations between thermal patterns and stress markers (cortical level and heart rate) are weak and not statistically significant, suggesting limited direct association. However, facial thermal patterns show sensitivity to emotional stress and could allow real-time stress detection. Immediate temperature rises occur during startle responses, but experimental results indicate that the periorbital region does not rise immediately during the Trier Social Stress Test. Instead, it shows a delayed increase after stress begins. Further, periorbital thermal signal-based heart rate extraction is developed using a novel region of interest to estimate heart rate in real-world contexts. Diverse energy-Phase Space Attention Magnification Network is proposed to amplify and extract stress-induced signals. The encouraging result further demonstrates the accuracy of emotional stress detection is 97.1%. Overall, these experimental results exhibit that facial thermal imaging is a valuable tool for estimating emotional stress.