Microexpression Classification Based on Vision Transformer and Fusion Super-Resolution Technology

With the advancement of artificial intelligence technology, microexpression recognition has garnered significant attention in the fields of public services and lie detection. It is an essential tool for understanding human emotions and non-verbal communication. However, due to the brevity and subtlety of microexpressions, research in this area faces numerous challenges, including limited image datasets, low image resolution, diverse categories, and insufficient model generalization capabilities. To address these issues, a microexpression image classification method based on Vision Transformer (ViT) and fusion super-resolution technology is proposed. This method leverages the strengths of super-resolution technology and the ViT model. The fusion dataset method, which utilizes a diffusion model for super-resolution, can effectively expand the microexpression dataset. Furthermore, the ViT model can effectively enhance the feature extraction ability for microexpression images. After applying this method to extract image features, the validation accuracy of the experimental dataset increased from 66.99% to 92.59%, representing a 25.6% improvement. The experimental results demonstrate that the proposed method of integrating ViT with fusion super-resolution effectively enhances the accuracy of microexpression image classification, providing a valuable reference for the application and development of microexpression recognition and classification technology in areas such as teaching activities, psychological research, and the legal field.