Ocular-Behavioral Metrics for Driver State Classification for Indian Driving Contexts: KSS-Based Evaluation, Conformance and implementation challenges

With the increasing integration of Advanced Driver Assistance Systems (ADAS) in Indian automotive markets, robust Driver Monitoring Systems (DMS) have become critical for road safety. This study focuses on the development and validation of a Driver Distraction and Attention Warning System (DDAWS), tailored to the unique challenges of Indian driving conditions. The DDAWS module, a core component of DMS, leverages a driver-facing high-resolution camera to continuously evaluate driver behaviour. The system architecture is bifurcated into two primary subsystems: the Driver Attention Monitoring module and the Drowsiness Detection module. DMS primarily is divided into two parts, first being the Driver Attention monitoring system, which ensures the driver maintain a proper focus on the road by not being distracted either by the inner and the external environment of the vehicle. This is generally achieved using the head pose monitoring of the driver, post processing which results in an identification of abnormal positions that may indicate distraction or inattention. Other parts of the DMS includes the driver drowsiness system, classically analysis includes perception of the driver using a camera and analysing measures like PERCOS and EAR with yawn monitoring over face landmarks and relevant threshold. These types of analysis give a situational analysis of the driver drowsiness and because of inaccuracies at time is generally referred to as driver fatigue detection system. Generally, Karolinska sleepiness scale (KSS) is defined as a measure of the drowsiness detection and is scaled from 1-9, where 1 represent extremely alert and 9 represent extremely sleepy. Overall being a clinically self-assessed scale, it becomes non-trivial to rate the drowsiness of a driver on this, as while driving due to environmental agents and other factors this might over-rate or under-rate. The paper contributes the analysis of automotive standards like AIS184, EU 2023/2590 and EURO-NCAP for DDAWS and proposed methodologies for analysis of driver drowsiness as per the KSS scale. A comparative study is conducted between algorithm-generated KSS scores and self-reported KSS ratings to evaluate model fidelity. Furthermore, a multi-level redundancy architecture is proposed to enhance robustness in state prediction by fusing inputs across attention and drowsiness detection layers. This paper contributes a comprehensive framework for DDAWS, optimized for Indian driving conditions, with an emphasis on scalability, regulatory compliance, and predictive reliability