Related Experiment Video
Updated: Jul 12, 2026
05:51
Exploring the Use of Isolated Expressions and Film Clips to Evaluate Emotion Recognition by People with Traumatic Brain Injury
Published on: May 15, 2016
8.9K
An enhanced GhostNet model for emotion recognition: leveraging efficient feature extraction and attention mechanisms
Jie Sun1, Tianwen Xu1, Yao Yao2
1Psychological Development Guidance Center, School of Educational Sciences, Quanzhou Normal College, Quanzhou, China.
Frontiers in Psychology
|April 24, 2025
Summary
This study introduces an Enhanced GhostNet with Transformer Encoder (EGT) model for accurate facial emotion recognition. The EGT model demonstrates superior performance in complex environments, improving human-computer interaction.
Area of Science:
- Computer Science
- Artificial Intelligence
- Cognitive Science
Background:
- Emotion recognition is vital for understanding decision-making, but current systems struggle with real-world complexity and data limitations.
- Facial expressions are key indicators of emotional states, yet robust recognition remains challenging.
Purpose of the Study:
- To develop a deep learning model for enhanced facial emotion recognition.
- To address limitations in existing systems regarding environmental complexity and data scarcity.
Main Methods:
- Proposed an Enhanced GhostNet with Transformer Encoder (EGT) model.
- Integrated GhostNet for efficient feature extraction and Transformer for global context.
- Employed a dual attention mechanism to focus on critical facial features.
Main Results:
- Achieved 89.3% accuracy on the RAF-DB dataset and 85.7% on the AffectNet dataset.
- Outperformed state-of-the-art lightweight models in emotion recognition.
- Demonstrated robust performance in challenging and noisy environments.
Conclusions:
- The EGT model offers improved accuracy and robustness for facial emotion recognition.
- Potential applications include enhancing human-computer interaction, personalized recommendations, and mental health monitoring.
- Advanced deep learning techniques significantly advance emotion recognition capabilities.
