AI Safety Guard: Design, Prototype Implementation, and Validation Roadmap for a Privacy-Preserving Multi-Sensory Edge-AI Driver Drowsiness System
The paper frames driver drowsiness as a persistent road-safety problem that is hard to measure because incidents are intermittent and often unreported. It presents a low-cost edge-AI prototype that pairs non-contact facial-landmark analysis — prolonged eye closure, mouth opening and yawning, and head-pose deviation — with temporal decision fusion to trigger an audible alert and an optional atomised scent pulse. Unlike cloud-based monitoring, all camera processing happens locally, retaining no video and recording only minimal local event data.