DriveMind : A Vision and Physiological Sensor-based Driver Drowsiness Detection System using Machine Learning

Abstract

Road accidents mostly occur because of driver drowsiness. This paper presents DriveMind, a driver monitoring system that combines MediaPipe Face Mesh-based Eye Aspect Ratio (EAR) analysis with physiological sensors to provide an approximate continuous driver safety score. A multimodal dataset was used to train a regression model, and the trained model is deployed on a Raspberry Pi platform for real-time inference using live sensor data. K-Nearest Neighbors (KNN) regression model is used to provide the driver safety score using multimodal data. The experimental results indicate good performance with MAE of 1.31, RMSE of 1.68 and R2 of 0.913. Raspberry Pi has been selected to deploy the system in real-time inference and cloud-based monitoring, which proves to be applicable in low-cost embedded driver safety applications.