A Review of Advancements in Driver Emotion Detection: Deep Learning Approaches and Dataset Analysis

This review paper delves into the rapidly evolving field of driver emotion detection, with a specific focus on the contributions of deep learning methodologies and the diverse datasets that facilitate this research. Facial emotion detection and recognition is a highly dynamic and challenging field in Machine Learning (ML) and Artificial Intelligence (AI). It has gained attraction for several decades, but it is extremely challenging due to the intrinsic complexities associated with understanding and interpreting human emotions. Understanding and responding to a driver's emotions is increasingly important when autonomous or assisted driving becomes common. To ensure safety, comfort, and optimal interaction between the driver and the car's systems, predicting the emotional state of the driver is essential, also of great significance in practical applications. We give a comprehensive survey of the state-of-the-art driver emotion recognition works that can effectively make use of the recent deep-learning approaches to identify complex emotional cues. Moreover, we explain a variety of datasets that play a vital role in flourishing this field along with the analysis of their effect, like AffectNet, CK+, and EMOTIC. Via this survey we try to investigate the challenges authors are faced with involving this field, e.g., concerns about data privacy, real-time processing demands, and the need for interdisciplinary collaboration. More importantly, the potential of these technologies to improve the driving experience and road safety has been highlighted. We hope this survey can benefit researchers and practitioners to have more insights and provide directions for advancing drivers’ emotions.