|
|
TWK: 8th Tübingen Perception Conference
25th - 27th Feb 2005
|
|
|
|
Mid- and High-Level Motion Analysis - From Image Features to Behavior
Martin A. Giese (University Clinic Tübingen) & Heiko Neumann (University of Ulm)
Much of the previous research on motion perception has focused on simple mechanisms for motion detection and integration, such as explaining motion coherence or the solution of the aperture problem. Recently, a vivid interest in the study of higher functions in motion percep-tion has emerged and research has started to address questions like: How do attention and top-down effects influence the processing of motion information? How does, potentially learned, information about movements and shapes interact with the perception of motion? What are the processes that underlie the perception of motion of complex articulated objects and biological organisms? Another interesting question, which has attracted much attention, is whether the recognition of complex body movements is linked to the motor planning of movements, poten-tially exploiting a common representation for perception and action. The symposium tries to give an overview of several recent lines of research in mid- and high-level motion vision, pre-senting results from psychophysics, electrophysiology, imaging, and computational modelling.
The talk by Heiko Neumann will illustrate how a local integration of motion and form informa-tion can be accomplished by exploiting relatively simple physiologically plausible dynamic neural networks. By the appropriate combination of feed-forward and feedback connections, and appropriate connection between different cortical areas an accurate estimation of object motion in presence of occlusions and discontinuities of the optic flow field can be accom-plished. Stefan Treue will present physiological data from monkey cortex that sheds some light in the neural basis of the modulation of motion processing by attention. Neural correlates of spatial and feature-based attentions are discussed. The contribution of Marc Grosjean will dis-cuss psychophysical evidence demonstrating interference between execution and perception of different motor actions. Such evidence supports the idea of a common substrate for the encod-ing of events in perception and action. Ian Thornton will discuss how internal representations of static and dynamic stimuli interact in high-level perception, addressing problems like the perception of body and facial movements. Particularly, he will present psychophysical evidence for an important role of top-down influences in the processing of complex movement and form stimuli. The contribution of Martin Giese discusses experimental evidence from psychophysics and functional imaging that supports the hypothesis that the recognition of biological move-ments and actions is based on the learning of prototypical patterns. These results support a computational theory that assumes that action patterns are neurally encoded in terms of spatio-temporal templates for sequences form and optic flow patterns. |
|
