Enhancing Facial Expression Recognition in Education with Hybrid Attention-Driven Feature Clustering

Abstract

Facial Expression Recognition (FER) is increasingly being used in education to analyze student engagement and emotional responses, especially in online learning settings. By identifying emotions like interest, confusion, or frustration, FER provides educators with insights to refine their teaching methods and adapt to student needs. This paper reviews the current FER techniques applied in educational environments, emphasizing recent technological progress that has enhanced the accuracy and efficiency of these systems. Advances in computer vision and deep learning have significantly improved emotion detection, enabling real-time feedback and a more personalized learning experience. Despite these developments, challenges persist, such as high computational requirements and privacy issues related to students' emotional data. To tackle these problems, we suggest creating lightweight algorithms and privacy-focused solutions to make FER more applicable in classrooms. Additionally, we introduce a novel model, the Hybrid Attention-Driven Feature Clustering Network (HAFNet), which combines three components: the Feature Clustering Network (FCN), Multi-Head Attention Network (MAN), and Attention Fusion Network (AFN). The FCN enhances class separation using an affinity loss function, while the MAN captures detailed attention from different facial regions. The AFN integrates these attention maps to improve emotion classification accuracy, potentially enhancing educational outcomes through better FER performance.