Computational Principles for the Recognition of Biological Movements: Model-Based Versus Feature-Based Approaches

Martin A. Giese, Knoblich, Marc Grosjean, Ian M. Thornton, Maggie ShiffrarMPG.PuRe 2005

Abstract

The recognition of body movements is a complex computational problem. Multiple theories about possible underlying computational mechanisms have been discussed in the psychological literature, and in computer vision a variety of technical algorithms for body tracking and human movement recognition have been proposed. However, models for neural processes in the visual cortex must fulfill additional constraints. We discuss two opposing explanations for the recognition of complex body movements: The fitting of kinematic models, and an analysis in terms of simple dynamic motion and form features. We present a physiologically plausible neural model that accounts for biological motion recognition by the analysis of simple form and motion features that accounts for many known experimental results. We discuss advantages and limitations of the two approaches as models for the cortical processing of biological movement stimuli. Also we provide some ideas how the neural model can be extended to account for an influence of motor planning on visual perception.
BibTeX
@inproceedings{giese2005computational,
  author = {Martin A. Giese and Knoblich and Marc Grosjean and Ian M. Thornton and Maggie Shiffrar},
  title = {{Computational Principles for the Recognition of Biological Movements: Model-Based Versus Feature-Based Approaches}},
  booktitle = {MPG.PuRe},
  year = {2005},
}

Loading related content...

Figure