GraFIX: a semiautomatic approach for parsing low- and high-quality eye-tracking data

Area of Science:

• Cognitive Science
• Developmental Psychology
• Human-Computer Interaction

Background:

• Fixation durations (FD) are key metrics for attention and information processing.
• Existing fixation detection methods struggle with low-quality eye-tracking data, particularly in infants, leading to inaccuracies.
• Manual coding is time-consuming, while automatic methods lack precision with variable infant data.

Purpose of the Study:

• To introduce GraFIX, a novel semiautomatic method for accurate and efficient fixation detection in eye-tracking data.
• To address the challenges of variable data quality in special populations like infants.
• To improve the reliability and stability of fixation duration measures.

Main Methods:

• GraFIX employs a two-step process: initial parsing with user-adapted velocity-based algorithms, followed by graphical interface manipulation for fine-tuning.
• Algorithms include data smoothing, interpolation of missing points, and artifactual fixation removal based on user-defined criteria.
• The method incorporates visualization tools to aid manual coding and allows adaptation of parameters like velocity threshold and interpolation latency.

Main Results:

• Intercoder reliability analysis demonstrated that GraFIX yields more reliable and stable fixation duration measures compared to previous methods.
• The method proved effective across both low- and high-quality infant eye-tracking data.
• Adaptable detection criteria enhance the robustness of fixation duration analysis.

Conclusions:

• GraFIX provides a significant advancement in analyzing eye-tracking data, particularly for challenging datasets from infant populations.
• The semiautomatic approach balances accuracy and efficiency, making fixation duration analysis more accessible.
• This method enhances the validity of research on attention and information processing in infants and other special populations.