Lightweight Neural Network with Attention Mechanism for Enhanced Facial Expression Recognition

Facial expression recognition plays a pivotal role in various domains, including human-computer interaction and healthcare. In healthcare, recognizing pain and smile expressions aids in assessing patients' pain status and recovery progress. However, challenges such as inaccurate data labeling and complex model training hinder real-world deployment. This study introduces a novel lightweight neural network model (A-C model) incorporating an attention mechanism (D-SE-BAM) and transfer learning to recognize spontaneous expressions with limited datasets. The model demonstrated excellent performance, achieving 93.1% accuracy on the GENKI-4K dataset, 88.8% on the UNBC database, and 89.02% in real scenarios. These findings provide valuable insights for advancing facial expression recognition technologies.