Fine-Tuned CNNs with Self-Attention Mechanism for Enhanced Facial Expression Recognition

The growing need for facial emotion recognition in various domains, particularly in online education, has driven advancements in Artificial Intelligence (AI) and computer vision. Facial expressions are a vital source of nonverbal communication as they convey a wide range of emotions through subtle changes in facial features. Recent developments in Deep Learning (DL) and Convolutional Neural Networks (CNNs) have opened new avenues for analyzing and interpreting human emotions. This study proposes a novel CNN-based real-time facial expression recognition (FER) framework tailored for online education systems. The framework incorporates dynamic region attention and self-attention mechanisms, enabling the model to focus on key facial regions that vary in importance depending on emotional context. The proposed model is fine-tuned to enhance its capability to identify facial expressions in various situations by integrating these methods with transfer learning. Experimental results demonstrate that the model achieves an accuracy of 83% using FER 2013, surpassing traditional static image-based techniques. This study proposes to bridge the gap in facial expression observation in online education, facilitating educators with valuable visions into pupil sentiments to advance learning consequences.