The University of Sheffield
Department of Computer Science

Andrew Savill Undergraduate Dissertation 2005/06

"Blink rate estimation"

Supervised by Dr JP Barker

Abstract

Blinking is a subconscious action with its main purpose to keep the surface of our eyes moist but the average blink rate of an individual has been linked to many different fields of research such as estimating levels of fatigue, stress and deception. This project created a semi-autonomous blink detection system that is robust and user independent. It is able to detect the user's blinks within the time domain with an area under the ROC curve of up to 93%. This project investigates various techniques for blink detection. Among these techniques will be algorithms for eye detection such as using a skin tone threshold, eye tracking using a bordered variance algorithm and eye analysis using differenced eye image frames. This project has uncovered further areas of research such as blink duration detection and improved multi-cue blink detection algorithms.