Facial Emotion Recognition Using Geometric and HOG Features
Songhyun Chu, Beom Kwon · Journal of Korea Multimedia Society · 2025
In this study, we propose a face image-based emotion recognition technique that uses both geometric features and histogram of oriented gradients (HOG) features. Geometric features are designed based on the fact that the shapes of the eyebrows, mouth, and eyes vary depending on the emotional state in human facial expressions. For the HOG features, the shape of the wrinkles in the eye area changes according to the emotional state, and texture information-the shape of the wrinkles-is extracted by applying HOG. To demonstrate the effectiveness of geometric and HOG features in facial image-based emotion recognition, we conducted an ablation study using four machine learning (ML) models and the extended Cohn-Kanade (CK+) dataset. The experimental results show that the highest emotion classification accuracy is achieved for all four ML models when the two features are combined and used as input, rather than using each feature alone. Among the four ML models, logistic regression outperformed the others, achieving an accuracy of approximately 97.93%.