Enhancing Facial Expression Recognition Accuracy Through Spatial Transformation and Super-Resolution Preprocessing

Abstract

Facial expression recognition (FER) has become a critical field of study, driven by advancements in artificial intelligence and the increasing ubiquity of face recognition systems. This paper proposes an innovative FER approach that integrates image preprocessing techniques with a deep learning classification network. By employing the Spatial Transformation Network (STN), the proposed method addresses challenges such as input image size, shape variance, and background fusion. Subsequently, a super-resolution (SR) algorithm is applied to enhance image quality and preserve crucial details. These preprocessing steps feed into a traditional classification network to achieve superior recognition accuracy. Experimental results demonstrate the effectiveness of this approach, showing that it outperforms existing methods in terms of accuracy and computational efficiency. The study offers valuable insights into improving the robustness and performance of FER systems through preprocessing advancements.