Discrimination of fixations and smooth pursuit movements in high-speed eye-tracking data

A novel three-stage algorithm for detection of fixations and smooth pursuit movements in high-speed eye-tracking data is proposed. In the first stage, a segmentation based on the directionality of the data is performed. In the second stage, four spatial features are computed from the data in each segment. Finally, data are classified into fixations and smooth pursuit movements based on a combination of the spatial features and the properties of neighboring segments. The algorithm is evaluated under the assumption that the intersaccadic intervals represent fixations in data recorded when viewing images, and mainly smooth pursuit movements in data recorded when viewing moving dots. The results show that the algorithm is able to detect 94.3% of the fixations for image stimuli, compared to a previous algorithm with 80.4% detected fixations. For moving dot stimuli the proposed algorithm detects 86.7% smooth pursuit movements compared to 68.0% for the previous algorithm.