ESPBM

Eye Blink Detection with Prototype-Based EAR Analysis

This project presents a novel method for detecting eye blinks by establishing eye state prototypes that match blink patterns in the Eye Aspect Ratio (EAR) time series. Unlike traditional approaches that simply classify blinks as ON/OFF, our method captures blink speed, duration, and inter-eye synchronicity, offering richer insights for potential diagnostic applications.

We implement both unsupervised learned prototypes and manually defined prototypes, showing that both approaches reliably detect blink intervals. Following a “minimal working prototype” principle, our method allows even non-technical users to quickly create prototypes for specific blink patterns.

This work is published and can be cited via DOI. The source code and sample experiments are available on GitHub for anyone interested in exploring or building upon this approach: View the GitHub Repository.