Purchase this article with an account.
Leonid Fedorov, Martin Giese; An integrated model for the shading and silhouette cues in the perception of biological motion.. Journal of Vision 2016;16(12):273. doi: 10.1167/16.12.273.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
In absence of stereo and texture cues the perception of biological motion can show multistability, where different walking directions can be attributed to the same physical stimulus (Vangeneugden 2012, Vanrie 2004, 2006). This bistability disappears for walkers that consist of shaded volumetric elements, where the disambiguation depends on the light-source direction. No neural models exist that account for this multi-stability of biological motion perception, or its dependence on shading cues. METHODS: We extended an established neural model for biological motion perception (Giese & Poggio, 2013) by two essential mechanisms: (i) A two-dimensional neural field that consists of snapshot neurons that are selective for keyframes of actions with different preferred views. The dynamics of this recurrent neural networks reproduces the multi-stability of biological perception. (ii) A new hierarchical pathway that processes intrinsic shading cues, independent of contours generated by the silhouette of the walker. RESULTS: The model reproduces the observed bistability of biological motion perception, perceptual switching, and its dependence on adaptation. We are able to reproduce the disambiguation by additional shading cues, and demonstrate that the robustness of processing of such internal shading cues is improved by the novel shading pathway. CONCLUSION: Body motion perception integrates multiple cues, including surface shading. Straight-forward extensions of existing neural models can account for these effects.
Meeting abstract presented at VSS 2016
This PDF is available to Subscribers Only