Eye movement analysis with hidden Markov models (EMHMM) with co-clustering
Janet H Hsiao1,2, Hui Lan3, Yueyuan Zheng4
1Department of Psychology, University of Hong Kong, Pok Fu Lam, Hong Kong. jhsiao@hku.hk.
Behavior Research Methods
|April 30, 2021
Summary
New eye-tracking analysis combines hidden Markov models and co-clustering to reveal distinct visual exploration patterns. These patterns correlate with specific cognitive abilities in scene perception and task performance.
Area of Science:
- Cognitive Psychology
- Computational Neuroscience
- Data Science
Background:
- Eye movement analysis using hidden Markov models (EMHMM) quantifies individual differences in visual patterns.
- Existing EMHMM methods are limited to stimuli with uniform feature layouts, restricting their application.
- A novel approach is needed to analyze eye movements across diverse visual stimuli and tasks.
Purpose of the Study:
- To develop a computational method combining EMHMM with co-clustering for analyzing eye movements across varied stimuli.
- To identify distinct eye-movement patterns in scene perception.
- To investigate the relationship between these patterns and cognitive performance.
Main Methods:
- Combined hidden Markov models (EMHMM) with co-clustering data mining technique.
- Applied the integrated method to analyze eye movements during scene perception tasks.
- Correlated identified eye-movement patterns with performance on object recognition and flanker tasks.
Main Results:
- Discovered two distinct eye-movement patterns: 'explorative' and 'focused'.
- Higher similarity to the explorative pattern correlated with better foreground object recognition.
- Greater similarity to the focused pattern was associated with enhanced feature integration in the flanker task.
Conclusions:
- The EMHMM with co-clustering method effectively quantifies eye-movement patterns across diverse visual stimuli and tasks.
- This approach offers insights into individual differences in cognitive abilities and styles.
- The method has broad applicability for studying cognitive behavior using eye tracking across various disciplines.
