A State of Art Survey on Deep Learning Techniques for Facial Expression Recognition

Facial Expression Recognition (FER) has become an important factor towards human-computer interaction, mental health care, and assistive technologies. This paper provides a detailed literature survey and comparative analysis of ten competitive research works in FER dealing with CNN-based architectures, hybrid approaches, transformer-based methodologies, and domain-specific models. The reviewed studies show improvements in terms of accuracy, interpretability and clinical relevance, notably in the context of recognizing expressions in case of individuals with cognitive disabilities or cerebral palsy. However, large discrepancies are still present in real-time adaptability, multimodal data fusion and inclusive dataset accessibility. Thematically categorizing, critically assessing and identifying research gaps in the existing literature, this review exposes current trends and trends that should be followed in future FER research emphasizing on fairness, explainability, and efficiency for deployment in the real world.