A Review of Intelligent Cockpit Driver Behavior Recognition and Analysis

Driver Behavior Recognition (DBR) in intelligent cockpits is essential for enhancing active safety and human-computer interaction through real-time monitoring and classification of driver actions. While unimodal approaches have made progress, they struggle with challenges such as occlusion and low-light conditions. Multimodal DBR overcomes these limitations by incorporating a variety of data sources. The integration improves the system’s robustness and precision, especially in practical scenarios. This paper reviews the evolution of DBR technologies, covering CNN-, RNN-, Transformer-, GCN-, and HGNN-based methods, and analyzes five benchmark datasets. Future research directions include lightweight multimodal modeling, cross-modal self-supervised learning, edge computing for real-time deployment, and explainable AI to bridge AI perception with automotive safety, advancing next-generation intelligent vehicles.