Using the Mean Shift Algorithm to Make Post Hoc Improvements to the Accuracy of Eye Tracking Data Based on Probable Fixation Locations

摘要

In eye tracking research, there is almost always a disparity between a participant's actual gaze location and the location recorded by the eye tracker. In this paper, we propose a mean shift error correction method that can reliably reduce the systematic error-which tends to stay constant over time- and restore the fixations to their true locations. We show that the method is reliable when the visual objects of the experiment are arranged in an irregular manner, such as not on a grid in which all fixations could be shifted to adjacent locations using the same directional adjustment. Using the mean shift method, the disparity between fixations and their nearest objects are calculated and plotted on a graph in terms of their x and y deviations. The highest density point in this graph, calculated using the mean shift algorithm, is shown to correctly capture the magnitude and direction of the systematic error. This paper presents the method, an extended demonstration, and a validation of the efficacy of the error correction technique.