20-09-2017, 12:35 PM
The detection of eye sparks is considered one of the most reliable sources of communication in modern human computer interaction (HCI) systems. This article proposes a new method for detecting eye sparks by comparing templates and measure similarity. In order to minimize false detection due to background change in the video frame, face detection is applied prior to removal of the eye template. Golden ratio concept is introduced for robust eye detection and is followed by creating eye templates for tracking. Eye tracking is performed by mapping the template between the template image and the surrounding region. The normalized correlation coefficient is calculated for successful eye tracking. The detection of eye sparks is performed based on the correlation score, as the score changes significantly each time a flicker occurs.