Simple and robust eye tracking with a webcam

Abstract

Existing eye tracking schemes typically require invasive or expensive hardware in order to get the accurate results. This paper describes a non-invasive remote eye tracking method that only requires a webcam. In this work we employed several methods that are consisted of three processes, such as face detection, eye detection, and pupil tracking. For face and eye detection, we used Haar feature-based cascade classifier. For pupil tracking, we modified a method so-called means of gradients by implementing preprocessing and post-processing techniques. Finally, we integrated these methods for eye tracking. The experimental results demonstrate that the proposed scheme using a built-in webcam does support eye tracking under various human conditions such as: low contrast, dominant eyelids/eyelashes, any types of skin colors, and reflection of glasses; and environmental situations such as indoor and outdoor. ？ 2016 International Information Institute.