Algorithms for the automated correction of vertical drift in eye-tracking data

Area of Science:

• Cognitive Science
• Neuroscience
• Human-Computer Interaction

Background:

• Vertical drift, a common issue in eye-tracking, causes fixation misplacements due to calibration loss.
• This drift is especially problematic for multiline reading experiments, potentially misallocating fixations between lines.
• Manual correction is labor-intensive, error-prone, and inconsistent.

Purpose of the Study:

• To systematically evaluate existing and novel algorithms for automated post hoc correction of vertical drift in eye-tracking data.
• To provide evidence-based recommendations for selecting the most appropriate drift correction method.

Main Methods:

• Documented ten major vertical drift correction algorithms, including two novel approaches.
• Evaluated algorithms using both simulated and naturalistic eye-tracking reading data.
• Compared algorithm performance based on effectiveness in correcting different drift phenomena and suitability for various reading behaviors.

Main Results:

• Dynamic time warping emerged as a highly promising method for vertical drift correction.
• Algorithm performance varied, with some methods proving more effective for specific drift types and reading patterns.
• Systematic evaluation revealed differential suitability of algorithms for distinct experimental conditions.

Conclusions:

• Automated correction of vertical drift is feasible and offers advantages over manual methods.
• Algorithm selection for vertical drift correction should be guided by the specific characteristics of the drift and the reading behavior observed.
• Further research and standardized evaluation are needed to advance drift correction techniques in eye-tracking.