Please use this identifier to cite or link to this item: http://repository.kln.ac.lk/handle/123456789/15706
Title: Statistical Analysis of Eye Movement Trajectories
Authors: Dissanayaka, P.
Su, J.
Ghosh, B.K.
Keywords: Eye movements
Binocular eye movement
Trajectories
Smoothing splines
Segmentation
Issue Date: 2016
Publisher: Faculty of Science, University of Kelaniya, Sri Lanka
Citation: Dissanayaka, P., Su, J. and Ghosh, B.K. 2016. Statistical Analysis of Eye Movement Trajectories. In Proceedings of the International Research Symposium on Pure and Applied Sciences (IRSPAS 2016), Faculty of Science, University of Kelaniya, Sri Lanka. p 53.
Abstract: Eye movement refers to the voluntary or involuntary movement of the eyes. Eye trajectories explain human search behavior. It has been revealed that variations of these trajectories reflect the dynamics of the oculomotor system. Eye movement trajectories under free exploration contain a lot of noise with saccades and fixations. As a result eye movement trajectories cannot be treated as any other trajectory. It has been a challenge to analyze gaze during free exploration, preserving spatial and temporal characteristics of eye movements. Therefore, most of the experiments are carried in laboratories under restrictions. Here we address the problem of analyzing the binocular eye movement trajectory under free exploration, to understand the underlying patterns in the movement. First eye movement trajectory is mapped onto the unit sphere as a set of time indexed points. The obtained eye movement trajectories on the sphere are not simple curves and with repeated movements. Therefore, the whole trajectory was segmented using the speed at each time instance in order to obtain simple trajectories. For each segment, we have estimated a smoothed curve given by a set of time-indexed points on the sphere so that the estimated curve would approximate the data points best at given time instants while being regular as possible. These smoothed curves as cubic splines can be used to analyze patterns in the whole trajectory. Furthermore, the estimated curves as a set of time indexed points were used in interpolation and clustering. Distances between different curves are calculated using the geodesic distance on the unit sphere. Using the distance matrix of the segmented smoothed curves and the software called cystoscope, curves are grouped to obtain four different clusters. Each cluster from the binocular eye movement was analyzed for both left and right eye movement to obtain curves overlapped where both eyes move together in the same direction. These overlapped curves can be analyzed further to compare eye movement patterns in different individuals.
URI: http://repository.kln.ac.lk/handle/123456789/15706
ISBN: 978-955-704-008-0
Appears in Collections:IRSPAS 2016

Files in This Item:
File Description SizeFormat 
53.pdf217.72 kBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.