Driver State Monitoring Systems: A Comprehensive Review Bridging Academic Research and Industrial Deployment

Abstract A substantial gap remains between academic promise and industrial reality for driver state monitoring (DSM) systems. This comprehensive review systematically examines 171 publications on DSM systems, establishing a three-tier framework encompassing monitoring, assessment, and intervention methodologies. Although laboratory-based physiological signal approaches achieve over 96% accuracy and behavioral monitoring achieves 95% recognition rates, practical deployment encounters significant obstacles. Our analysis reveals four fundamental constraints limiting the real-world implementation of DSM systems: (1) reactive paradigms that detect events rather than predict emerging risks, (2) modular architectures that hinder human–machine synergy, (3) universal models that do not adequately address individual driving variability, and (4) the disparity between controlled research environments and industrial realities, including computational constraints and privacy regulations. Contemporary commercial systems emphasize robustness over algorithmic sophistication, favoring proven infrared camera technology over advanced multimodal solutions, largely because of regulatory requirements (EU GSR 2024/2026) and safety certification standards. To address these challenges, we outline four strategic research directions: proactive risk prediction through probabilistic modeling, integrated human-in-the-loop architectures for adaptive intervention, personalized digital twin systems for individualized monitoring, and industry-aligned methodologies that incorporate real-world deployment constraints. This framework outlines a potential pathway for advancing DSM technology toward more proactive and personalized intelligent assistance, while acknowledging that realizing this transition will require sustained interdisciplinary effort.