要旨
Body motion perception from impoverished stimuli, such as point-light motion, shows interesting multi-stability, which disappears in presence of shading cues that suggest depth. Existing models for body motion perception account neither for multi-stability nor for the specific influence of such shading cues.We propose a new model that captures these phenomena. METHOD: We extended a classical hierarchical recognition model for body motion by: (i) a two-dimensional dynamic neural field of snapshot, which reproduces decisions about walking direction and their multi-stability; (ii) a novel hierarchical pathway that processes specifically intrinsic luminance gradients, being invariant against the strong contrast edges on the boundary of the body. RESULTS:We show that the model reproduces the observed multi-stability of perception for unshaded walker silhouettes. The addition of intrinsic shading cues results in monostable unambiguous perception in the model,
consistent with psychophysical results. In addition, we show that the novel shading pathway is necessary to accomplish a robust recognition of the relevant intrinsic luminance gradients.
CONCLUSIONS: By straight-forward extensions of a classical physiologically-inspired model for body motion recognition we were able to account for the perceptual multistability and the dependence on shading cues of body motion perception.
