Attention Mechanisms-Based Convolutional Neural Networks for Facial Emotion Recognition in Human-Computer Interaction
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Understanding user emotions is crucial for building adaptive and intuitive systems. Facial emotion recognition (FER) enables machines to capture subtle emotional cues from facial expressions, which are universal indicators of human affective states. While convolutional neural networks (CNNs) have achieved very good performance in related FER tasks, their ability to focus on important facial regions is still limited. To address this problem, this paper aims to propose a lightweight but powerful FER framework using MobileNet, improved by two typical attention mechanisms called SENet and CBAM. We train and evaluate our framework on the FER-2013 dataset and use various preprocessing and data augmentation strategies to improve generalization. Experimental results demonstrate that the inclusion of SENet and CBAM can improve classification accuracy, precision, recall, and area under the curve (AUC) over the baseline MobileNet, while keeping computational efficiency suitable for real-time applications. These findings can prove the potential of the attention-enhanced MobileNet architecture for improving practical emotion-aware HCI systems.
Fields of Study & Subjects
Emotion and Mood Recognition
Face recognition and analysis
Face and Expression Recognition
Publication details
- Journal / source
- IEEE International Conference on Power, Electronics and Green Energy (ICPEGE)
- Pages
- 606–610
- Publication date
- 28 January 2026
- Language
- EN