Proposal of an Image Processing Method for Enhancing Robustness in Stress Coping Style Classification Using TFI

Abstract

Human stress coping styles, classified as active or passive, can be distinguished through hemodynamic measurements. However, conventional cuff-based methods are impractical for daily use. Previous studies have explored thermal facial imaging (TFI) as a non-contact alternative and developed classification models for stress coping styles. However, challenges such as facial misalignment and noise from muscle activity have led to a decrease in classification accuracy. This study proposes an image processing method to reduce these sources of noise and enhance model robustness. As a result, our approach improved classification accuracy by approximately 10% compared to conventional methods. © 2025 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.