A new method for categorizing scanpaths from eye tracking data

Haass, Michael Joseph; Matzen, Laura E.; Butler, Karin M.; Armenta, Mika · 2016 · OpenAlex-citations

DOI: 10.1145/2857491.2857503

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper introduces GazeAppraise, a novel method for categorizing eye movement scanpaths from raw data without requiring prior definitions of fixations, saccades, or regions of interest. Traditional scanpath analysis often relies on partitioning eye movements into discrete events or specifying areas of interest, which can be computationally demanding and limiting in dynamic environments. The authors adapted the Tracktable library, originally designed for geospatial trajectory analysis, to extract spatiotemporal features from eye tracking data. The goal was to demonstrate that unsupervised clustering could distinguish scanpaths based on their geometric and sequential dependencies alone. The study involved 41 participants who performed four smooth pursuit tracking tasks, where they followed a white dot moving across a screen in specific shapes: a star, an S, an O, and a swirl. Eye movements were recorded at 60 Hz using a Seeing Machines FOVIO tracker. The GazeAppraise algorithm processed the raw x and y position data by calculating median gaze positions over multiple temporal scales (whole scanpath, halves, thirds, and quarters), resulting in a 20-dimensional feature vector for each scanpath. These features were then clustered using the DBSCAN density-based algorithm, which does not require the number of clusters to be specified in advance. The results demonstrated high accuracy in categorizing scanpaths. Of the 164 recorded scanpaths, 162 were successfully assigned to one of four clusters, yielding a recall/sensitivity of 98.8%. Crucially, every categorized scanpath was grouped exclusively with other scanpaths elicited by the same stimulus shape, resulting in 100% precision. The two uncategorized scanpaths were identified as outliers. This performance indicates that the algorithm could effectively distinguish between different visual tasks based solely on the trajectory of eye movements, even without preprocessing the data into fixations and saccades. The significance of this work lies in its ability to analyze eye movement patterns in minimally processed data, offering a robust alternative to methods that require strict assumptions about eye movement components. GazeAppraise is particularly advantageous for dynamic environments or cluttered visual scenes where defining regions of interest is difficult. The authors suggest that this approach could be valuable for visual search tasks requiring systematic strategies and for characterizing eye movements related to mind-wandering or visual imagery. Future work aims to validate the method with more complex, cognitively guided viewing tasks and to explore scaling algorithms that account for variations in scanpath speed and size.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• eye movements scanning