An elaborate algorithm for automatic processing of eye movement data and identifying fixations in eye-tracking experiments

Abstract

High sampling frequency of eye-trackers introduces noise in raw eye movement data. Furthermore, the unstable sampling frequency of devices generates fluctuating time interval between samples, which has negative influence on the quality of raw eye movement data. Therefore, the article aims to propose a valid and universal procedure on raw eye movement data processing. The procedure will be suitable for eye movement data collected by most types of eye-trackers. Characteristics of raw eye movement data were analyzed, and a comprehensive data processing procedure was proposed. Steps in this procedure include preliminary inspection of raw data, the checking and correction of actual sampling frequency, the selection of eye movement data, small gaps filled-in, data filtering, and fixation identification. By this procedure, raw eye movement data collected by eye-trackers can be used to identify types of eye movements and obtain fixation points in eye-tracking experiment. In the end, a paradigm test was designed to examine the correctness of data processing flow, and the application of eye movement data was also discussed.