We propose a novel algorithm to estimate the similarity between a pair of eye movement sequences. The proposed algorithm relies on a straight-forward geometric representation of eye movement data. The algorithm is considerably simpler to implement and apply than existing similarity measures, and is particularly suited for exploratory analyses. To validate the algorithm, we conducted a benchmark experiment using realistic artificial eye movement data. Based on similarity ratings obtained from the proposed algorithm, we defined two clusters in an unlabelled set of eye movement sequences. As a measure of the al - gorithm's sensitivity, we quantified the extent to which these data-driven clusters matched two pre-defined groups (i.e., the 'real' clusters). The same analysis was performed using two other, commonly used similarity measures. The results show that the proposed algorithm is a viable similarity measure.
|Number of pages||15|
|Journal||Journal of eye movement research|
|Publication status||Published - 1 Jan 2012|
- Cognitive Science
- Visual Perception
- Eye movements