Analysis of eye tracking movements using innovations generated by a Kalman filter

Abstract
A simple but efficient algorithm has been developed for computer analysis of eye tracking movements. The program separates smooth pursuit and saccadic eye movements. Separation of the two components is achieved using a two-step process of saccade detection. First, an AR model of the velocity of the smooth component is identified and used to determine a Kalman filter. Secondly the innovation sequence generated by this filter allows saccade detection. The precise beginning and end of each saccade are found using a Hinkley algorithm. Examples are given of analysis procedure for eye tracking of a random moving target. The method proved to be highly reliable and could be easily extended to other eye movements such as nystagmus.