Quantifying pursuit movements through measures of spatio-temporal similarity

Kevin Parisot 1, 2, 3 Alan Chauvin 3 Anne Guérin-Dugué 1 Ronald Phlypo 1 Steeve Zozor 2
1 GIPSA-VIBS - VIBS
GIPSA-DIS - Département Images et Signal
2 GIPSA-CICS - CICS
GIPSA-DIS - Département Images et Signal
Abstract : In eye movement research, smooth pursuits represent oculomotor events in which the eyes are locked on a perceptual target and track it along the visual field. Such a phenomenon shows high similarity between the target’s displacement in the visual field and the seemingly induced eye movements. In previous work [Parisot et al., 2017, Parisot et al., 2016], we investigated different measures to quantify similarity between two spatio-temporal signals; stimulus motion and gaze. Though qualitatively seemingly simple, similarity between multi-dimensional physiological signals is non-trivial to quantify. We proposed mea- sures based on canonical correlation analysis (CCA) and mutual information (MI) to detect the presence of pursuits within fixational raw gaze signals. We show that this approach gives robust estimations of statistical similarities in variable fixation samples and can provide a measure of the delay or deforma- tion that separate the two signals and can be easily interpreted in gaze signals analysis. We compare this similarity measure to other, existing methods on both simulated and real data sets and propose a quantification for pursuit signals in eye movement research.
Document type :
Poster communications
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-01986047
Contributor : Kevin Parisot <>
Submitted on : Friday, January 18, 2019 - 2:39:35 PM
Last modification on : Saturday, January 19, 2019 - 1:17:22 AM

Identifiers

  • HAL Id : hal-01986047, version 1

Citation

Kevin Parisot, Alan Chauvin, Anne Guérin-Dugué, Ronald Phlypo, Steeve Zozor. Quantifying pursuit movements through measures of spatio-temporal similarity. GDR Vision 2018, Oct 2018, Paris, France. ⟨hal-01986047⟩

Share

Metrics

Record views

38