by Amol Raoji Madane*, Ho Chin Kuan
ABSTRACT
Driver distraction and fatigue remain among the leading causes of road accidents worldwide, necessitating the
development of intelligent, reliable, and human-centric driver monitoring systems. This paper presents a
vision-based driver state monitoring framework for the real-time detection of inattentiveness and drowsiness
using multiple behavioral indicators, including eye closure, yawning, head lowering, and head-pose-based
attention assessment. Unlike conventional approaches that rely on a single distraction cue, the proposed
framework integrates complementary visual features to improve detection robustness and reduce false alarms
under varying driving conditions. The system employs computationally efficient facial feature extraction and
classification techniques to continuously analyze driver behavior using an in-cabin camera without requiring
intrusive physiological sensors. Furthermore, the framework utilizes interpretable behavioral indicators that
enable transparent decision-making, supporting the principles of Explainable Artificial Intelligence (XAI) in
safety-critical automotive applications. Experimental evaluation conducted on a diverse set of subjects
demonstrated high detection accuracies across multiple distraction and fatigue scenarios, validating the
effectiveness of the proposed approach for real-time deployment. The findings indicate that integrating
multiple observable behavioral cues enhances driver state recognition and contributes toward the development
of trustworthy, responsible, and intelligent Advanced Driver Assistance Systems (ADAS). Future
enhancements include explainability-driven decision visualization, fairness assessment across diverse driver
demographics, and multimodal sensor fusion for improved robustness and reliability in real-world
environments.
