An Analysis of Entropy-Based Eye Movement Events Detection
Katarzyna Harezlak1, Dariusz R Augustyn1, Pawel Kasprowski1
1Institute of Informatics, Silesian University of Technology, Akademicka 16, 44-100 Gliwice, Poland.
Entropy (Basel, Switzerland)
|December 3, 2020
Summary
This study introduces a new method using approximate entropy to detect eye movement events like saccades. The Multilevel Entropy Map achieved 83-94% accuracy, offering a promising way to analyze eye movement dynamics.
Area of Science:
- Neuroscience
- Computer Science
- Cognitive Science
Background:
- Eye movement analysis is crucial for understanding human interest and cognition.
- Detecting fixations and saccades is key to interpreting spatiotemporal visual processing.
Purpose of the Study:
- To propose a novel approach for detecting eye movement events using approximate entropy.
- To develop and evaluate a Multilevel Entropy Map for eye movement analysis.
Main Methods:
- Utilized a multiresolution time-domain scheme to create the Multilevel Entropy Map.
- Collected eye position data at a 1000 Hz sampling rate during a 'jumping point' experiment.
- Applied the k-nearest neighbors (knn) classifier for event detection.
Main Results:
- The Multilevel Entropy Map demonstrated effective detection of eye movement events.
- Classification efficiency for identifying the saccadic period ranged from 83% to 94%.
- Accuracy varied based on the sample size utilized in the analysis.
Conclusions:
- The proposed approximate entropy-based method shows potential for describing eye movement dynamics.
- The Multilevel Entropy Map offers a viable tool for eye tracking research.
- This approach may enhance the understanding of cognitive processes through eye movement analysis.
