Advances in Facial Emotion Recognition: A Comprehensive Review

Facial emotion recognition (FER) refers to the process of identifying human emotions and feelings through facial expressions. It consists of several types of emotions which are classified into six elementary categories (happy, angry, sad, disgust, fear, and surprise) along with other additional complex emotional states. Emotion recognition can be applied to several applications such as security, health, education, and entertainment. Although there have been several previous studies for facial emotion recognition, there is still a need to explore and enhance the feasibility and results of these techniques. Recent studies that incorporate transformers and pre-trained models have demonstrated promising results for enhancing the precision of FER systems. In this paper, we aim to present a comprehensive comparison of facial emotion recognition techniques through image datasets using deep learning and neural networks, traditional and CNN-based methods, transfer learning, transformer and attention mechanisms, transfer learning and pre-trained models, and object detection models which aim to provide an improvement in recognition accuracy, making FER systems more effective for real-world applications.