Facial Expression Recognition In Low Light And Occluded Environment Using Machine Learning Approaches

Abstract

Facial Expression Recognition (FER) is important in the context of intelligent systems to recognize human emotions, and has been applied across human-computer interaction, healthcare monitoring, surveillance, and driver assistance systems. Although considerable improvements have been made in recent years, performance of FER systems significantly decreases in the real-life situation, especially when using low-light scenarios and face obfuscation. These influences decrease the visibility of features, distort facial information and cause unfinished or unreliable representation, which decreases recognition accuracy. This review is a narrative synthesis of machine learning and deep learning methods to FER in such a challenging environment. It discusses methods of enhancement, occlusion-sensitive models as well as state-of-the-art architectures, such as attention systems and transformer frameworks. It is seen that the joint enhancement-recognition models can be more effective in restoring the semantically meaningful features when the light is low, and attention and transformer-based models are more efficient in coping with the occlusion through focusing on informative areas of the faces. Nevertheless, one of the main drawbacks of the current studies is the absence of cohesive datasets and benchmarks that can cover both the low-light and occlusion conditions at the same time. The review adds a hierarchical taxonomy of the current methodologies and indicates a way forward in future research to create robust and practical systems of FER.