September 2017
Volume 17, Issue 10
Open Access
Vision Sciences Society Annual Meeting Abstract  |   August 2017
Fat-tailed propagation noise model of visual object tracking
Author Affiliations
  • Byeong-Hee Gwak
    School of Life Sciences, UNIST, Ulsan, South Korea
  • Hanan Mohamed
    School of Life Sciences, UNIST, Ulsan, South Korea
  • Oh-Sang Kwon
    Department of Human Factors Engineering, UNIST, Ulsan, South Korea
Journal of Vision August 2017, Vol.17, 413. doi:https://doi.org/10.1167/17.10.413
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Byeong-Hee Gwak, Hanan Mohamed, Oh-Sang Kwon; Fat-tailed propagation noise model of visual object tracking. Journal of Vision 2017;17(10):413. https://doi.org/10.1167/17.10.413.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Background: According to the optimal tracking model of visual motion perception (Kwon, Tadin, & Knill, 2015), the visual system integrates noisy sensory inputs with the prediction of a forward model that reflects natural dynamics to estimate the states of a moving object continuously. The model provides a unifying account of several visual illusions, such as motion induced position shift, curveball illusion, and peripheral slowing of perceived speed. However, illusory perception of flicker-defined motion (Mulligan & Stevenson, VSS demo, 2014) provides a counterexample against the prediction of the model. The flicker-defined motion appears to jump, even when it moves continuously. The current version of the optimal tracking model, a Kalman filter model, cannot generate jumping percepts of a continuously moving object regardless of the parameter values. Simulation results: The propagation noise of the tracking model represents the visual system's assumption of the random changes of velocity over time. The current model assumes that the propagation noise follows a Gaussian distribution. However, various movements observed in nature follow fat-tailed distribution, rather. (Kleinberg, 2000; Sims et al., 2008). We built a model that assumes fat-tailed distribution as a propagation noise, and numerically simulated the performance of the model using particle filtering. The simulation results are qualitatively consistent with the observed illusory percepts. First, the model can account for the three visual illusions mentioned above (motion induced position shift, curveball illusion, peripheral slowing of perceived speed). Second, the model predicts jumping percepts given the flicker-defined motion as sensory inputs when the target moves smoothly. Furthermore, the model predicts that the jumping frequency increases as the speed of the target motion increases, which is consistent with the qualitative observation. Conclusions: Simulation results show that fat-tailed propagation noise model can account for a wider range of perceptual phenomena than the current optimal tracking model. It suggests that the visual system assumes fat-tailed propagation noise reflecting objects' movements often observed in nature.

Meeting abstract presented at VSS 2017

×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×