Advanced Real-Time Drowsiness and Distraction Detection Leveraging CMOS-IR and SoC Sensor Integration

Drowsy and distracted driving are leading causes of road accidents, resulting in approximately <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{3, 0 0 0}$</tex> deaths annually. The demand for reliable, real-time Driver Monitoring Systems (DMS) is critical to improving road safety. This project introduces a novel AI-driven DMS that addresses the limitations of existing solutions by integrating a global shutter CMOS Image Sensor (CIS) with a lightweight System-on-Chip (SoC) architecture. The proposed approach leverages infrared-based CMOS sensors-an area rarely explored-to enhance precision, operability in diverse lighting conditions, and processing efficiency. The integration of a global shutter infrared CIS ensures full-condition operability while providing superior image quality and processing speed. A neural network-based algorithm, trained on extensive datasets with several algorithmic improvements, forms the core of the DMS, effectively identifying drowsiness and distraction. The proposed system achieved promising results, including low power consumption-347 milliwatts in low frames-per-second mode and a significant accuracy improvement, with an F1 score increase of +0.103 on the validation set. This work contributes to the field by offering a high-performing, scalable, and cost-effective solution designed for implementation across various vehicle types and driving conditions, paving the way for safer roads.