Robust pupil detection for gaze-based user interface

Abstract

Eye tracker has the potential to play an assistive role in human computer interaction. However, the operation is easily hindered by excessive corneal reflection and noise which prevent correct detection and precise localization of the pupil. In this paper we describe a robust method for pupil detection in the presence of severe interference. The algorithm works by analyzing the intensity distribution of the pupil region and iteratively searching for the pupil center with an effective noise removal scheme. We demonstrated several gaze-based user interfaces, including web-browsing and photo-viewing, with a wearable eye tracker constructed using the proposed method.