Edge-AI Enabled Smart Helmet: A Lightweight Drowsiness Monitoring Framework for Motorcyclists

Fatigue-induced cognitive impairment is a leading contributor to road fatalities, with motorcyclists representing a particularly vulnerable demographic due to the lack of active safety features in two-wheelers. While Deep Learning models (such as CNNs) offer high detection accuracy, their computational demands render them unsuitable for battery-powered, wearable safety devices. This paper presents the design and implementation of a resource-efficient "Smart Helmet" system for real-time vigilance monitoring. Unlike computationally intensive neural networks, the proposed framework utilizes an optimized Histogram of Oriented Gradients (HOG) and Linear SVM pipeline to extract facial features on constrained edge hardware (Raspberry Pi). By leveraging the scalar Eye Aspect Ratio (EAR) metric, the system detects microsleeps with high-frequency inference (>15 FPS) without reliance on cloud connectivity. The system integrates visual monitoring with haptic feedback, validating its potential as a standalone, low-latency safety solution for riders.