The influence of eye model complexity and parameter variations on the predictive power of simulated eye-tracking data
Date
2023-11
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : Stellenbosch University
Abstract
ENGLISH ABSTRACT: Gesimuleerde oogvolgdata vergemaklik die vinnige, herhaalde, en onafhanklike evaluasie van oogvolgmetodes sonder dat daar 'n hulpbron-intensiewe gebruikersstudie vereis word. Gesimuleerde data is veral voordelig gedurende die vroeƫ ontwikkeling van 'n oogvolger waarin verskeie konfigurasies van apparatuur en algoritmiese komponente oorweeg word. Die meeste van die gesimuleerde oogvolgdata wat in die oogvolg-verwante literatuur gebruik word, is gegenereer deur gebruik te maak van lae-kompleksiteit oogmodelle met parameters gebaseer op bevolkingsgemiddeldes wat die kornea simuleer as 'n enkele sferiese oppervlak. Hierdie dissertasie het ondersoek ingestel of die verhoging van die oogmodel se kompleksiteit en parameters wat gebruik word om gesimuleerde data te genereer, kan lei tot verbeterings in werklike oogvolguitkomste. Die bevindinge van die dissertasie toon dat die gebruik van 'n oogmodel wat 'n a-sferiese voorste korneale oppervlak en realistiese variasies in die asferisiteit van die voorste korneale oppervlak insluit in 'n gesimuleerde oogvolgomgewing, die voorspellende vermoƫ van die gesimuleerde data sal verbeter. Dit sal moontlik op sy beurt lei tot die ontwikkeling van oogvolgmetodes met beter resultate.
AFRIKAANSE OPSOMMING: Simulated eye-tracking data facilitate the rapid, repeated, and independent evaluation of eye-tracking methods without requiring resource-intensive user studies. Simulated data are particularly advantageous during the early development of an eye-tracker in which various configurations of hardware and algorithmic components are considered. Most of the simulated eye-tracking data used in the eye-tracking-related literature were generated using low-complexity eye models with parameters based on population averages that simulate the cornea as a single spheric surface. This dissertation investigated whether increasing the eye model complexity and parameter variations used to generate simulated data can lead to improvements in real-world eye-tracking outcomes. The findings of the dissertation demonstrate that using an eye model that includes an aspheric anterior corneal surface and realistic variations in the asphericity of the anterior corneal surface in a simulated eye-tracking environment will improve the predictive power of the simulated data, which in turn will likely lead to the development of eye-tracking methods with better outcomes.
AFRIKAANSE OPSOMMING: Simulated eye-tracking data facilitate the rapid, repeated, and independent evaluation of eye-tracking methods without requiring resource-intensive user studies. Simulated data are particularly advantageous during the early development of an eye-tracker in which various configurations of hardware and algorithmic components are considered. Most of the simulated eye-tracking data used in the eye-tracking-related literature were generated using low-complexity eye models with parameters based on population averages that simulate the cornea as a single spheric surface. This dissertation investigated whether increasing the eye model complexity and parameter variations used to generate simulated data can lead to improvements in real-world eye-tracking outcomes. The findings of the dissertation demonstrate that using an eye model that includes an aspheric anterior corneal surface and realistic variations in the asphericity of the anterior corneal surface in a simulated eye-tracking environment will improve the predictive power of the simulated data, which in turn will likely lead to the development of eye-tracking methods with better outcomes.
Description
Thesis (PhD)--Stellenbosch University, 2023.