System Calibration

Abstract

Currently, most video-based eye trackers require calibration. This is usually a sequence of simple stimuli displayed sequentially at far extents of the viewing region. The eye tracker calculates the Point Of Regard (POR) by measuring the relative observed position of the pupil and corneal reflection at these locations, and then (most likely) interpolates the POR value at intermediate eye positions. The individual stimuli used for this purpose are simple white dots or cross-hairs on a black background. In cases where the eye tracker is used in an outside setting (e.g., for use during driving or while walking outside the lab), calibration marks may be made from simple targets such as tape or other visible markers fixed to objects in the environment. The purpose of calibration is to present a sequence of visible points at fairly extreme viewing angle ranges (e.g., upper-left, upper-right, lower-left, lower-right). These extrema points should be chosen to provide a sufficiently large enough coordinate range to allow the eye tracker to interpolate the viewer’s POR between extrema points. Most (video-based) eye trackers provide built-in calibration techniques where a number of such extrema points (e.g., 3, 5, or 9 typically) are presented in order.