Facial Emotion Recognition Methods, Applications and Future Challenges

Abstract

Facial Emotion Recognition (FER) has become a vital research area at the intersection of computer vision, psychology, and artificial intelligence, enabling machines to understand and respond to human affective states. With applications in healthcare, automotive safety, human–computer interaction, and security, FER systems are increasingly integrated into daily life. This review paper provides a comprehensive overview of the evolution of FER, covering traditional handcrafted feature-based methods, deep learning techniques, and emerging transformer-based models. Publicly available datasets such as JAFFE, CK+, FER2013, AffectNet, and RAF-DB are discussed, highlighting their contributions and limitations. Furthermore, recent progress in multimodal approaches that integrate facial, speech, and physiological cues is examined. Despite remarkable advances, FER still faces challenges related to dataset imbalance, cross-cultural variability, illumination and pose variations, and ethical concerns such as privacy and algorithmic bias. This review also identifies key gaps between laboratory performance and real-world deployment, emphasizing the need for robust, interpretable, and ethically responsible systems. Finally, future research directions are outlined, including self-supervised learning, domain adaptation, on-device FER for edge computing, and explainable AI. The paper aims to serve as a reference point for researchers and practitioners, offering insights into both current methodologies and potential pathways for innovation in emotion-aware intelligent systems.