An analysis method for eye motion and eye blink detection from colour images around ocular region

Abstract

This paper examines an analysis and measurement method that to aim to involve incorporating eye motion and eye blink as functions of an eye-based human-computer interface. Proposed method has been achieved by analysing the visible-light images captured without using special equipment, in order to make the eye-controlled input interface independent of gaze-position measurement. Specifically, we propose two algorithms, one for detecting eye motion using optical flow and the other for classifying voluntary eye blinks using changes of eye aperture areas. In addition, we tried evaluating experiment both detection algorithms simultaneously assuming application to input interface. The two algorithms were examined for applicability in an eye-based input interface. As a result of experiments, real-time detection of each eye motion and eye blink could be performed with an accuracy of about 80% on average.