December 2022
Volume 22, Issue 14
Open Access
Vision Sciences Society Annual Meeting Abstract  |   December 2022
A neural network model for vector decomposition, reference-frame selection, and relative-motion perception
Author Affiliations
  • Dongcheng He
    University of Denver
  • Haluk Ogmen
    University of Denver
Journal of Vision December 2022, Vol.22, 3040. doi:https://doi.org/10.1167/jov.22.14.3040
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Dongcheng He, Haluk Ogmen; A neural network model for vector decomposition, reference-frame selection, and relative-motion perception. Journal of Vision 2022;22(14):3040. https://doi.org/10.1167/jov.22.14.3040.

      Download citation file:


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

      ×
  • Supplements
Abstract

The perception of motion not only depends on the detection of motion signals but also on choosing and applying reference-frames according to which motion is interpreted. Here we propose a neural-network model that implements the common-fate principle for reference-frame selection. The first layer of the model extracts local retinotopic motion signals. Synaptic projections from this layer to the next decompose retinotopic motion-vectors according to the vector-decomposition principle. A parallel layer of neurons computes common motion across the retinotopic space of the first layer and implements the Gestalt common-fate grouping principle. The common-motion vector extracted by this layer is used as a reference-frame to select a subset of decomposed motion-vectors. We simulated this model for classical relative motion stimuli, viz., the three-dot, rotating-wheel, and point- walker (biological motion) paradigms and found the model performance, in terms of position, direction, and speed errors, to be close to theoretical vector-decomposition values. In the three-dot paradigm, the model made the prediction of perceived curved-trajectories for the central dot when its horizontal speed was slower or faster than the horizontal speed of the flanking dots. We tested this “reference-frame induced curvature” prediction in a psychophysical experiment, where observers were shown a variant of the three-dot stimuli and asked to sketch the perceived trajectories of the central dot. Subjects reported linear vertical trajectories when all dots had the same horizontal velocity. However, perceived trajectories were curved when the horizontal velocities of the central and flanking dots were different. In comparing the predicted and empirical curved trajectories, a good qualitative and quantitative agreement was found between the model and the data. Whereas some models use form information for relative-motion perception, our results show that a simple neural network using solely motion information can account for the perception of relative motion in the classical stimuli.

×
×

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.

×