Computational approaches to apply the String Edit Algorithm to create accurate visual scan paths.

Area of Science:

• Ophthalmology
• Computer Science
• Human-Computer Interaction

Background:

• Eye movement detection algorithms like I-VT necessitate accurate threshold selection for identifying fixations and saccades.
• Inaccurate thresholds can lead to misclassification of eye movements and affect visual scan path analysis.
• Current threshold evaluation methods are often labor-intensive or require unavailable data on expected scan paths.

Purpose of the Study:

• To introduce and evaluate two novel computational approaches for determining optimal thresholds in eye movement detection.
• To address the limitations of existing manual or data-dependent threshold selection techniques.
• To enhance the accuracy of identifying eye fixations, saccades, and subsequent visual scan paths.

Main Methods:

• Proposed two computational methods: 'between-participants comparisons' and 'within-participants comparisons'.
• Evaluated these methods using the Gazebase dataset with a bullseye-target tracking task.
• Compared the accuracy of identified thresholds against the known trajectory of the bullseye-target.

Main Results:

• The proposed methods identified optimal thresholds of 220°/s and 210°/s.
• These thresholds achieved 83% similarity to the expected visual scan path.
• This significantly outperformed a benchmark threshold of 30°/s, which yielded only 41.5% similarity.

Conclusions:

• The developed computational approaches offer an effective means for researchers to identify accurate thresholds for the I-VT algorithm.
• These methods show promise for improving the analysis of eye movement data across various research fields.
• The findings suggest a more objective and efficient way to process gaze data for reliable visual scan path reconstruction.