Unlocking Emotions Through Heat: Facial Emotion Recognition via Thermal Imaging

Abstract

Facial emotion recognition involves the analysis of facial expressions to identify and classify human emotions. As there is a relationship between facial temperature changes and emotions, this study focuses on facial emotion recognition using thermal images. This work aims to improve the efficiency, robustness, and accuracy of emotion recognition systems to improve human-computer interaction, mental health monitoring, personalized education, marketing effectiveness, and security. The proposed Convolution Neural Network (CNN) architecture is designed to recognize human emotions by classifying facial images into six distinct categories (happiness, sadness, anger, fear, disgust, and surprise). With an average detection accuracy of 0.9857, the proposed model showcases its effectiveness in emotion prediction. The findings indicate that our proposed system outperforms YOLOv5 and YOLOv5-NMS models across all emotions, showcasing superior results.