OSCILLATING MODELS FOR PERCEPTION OF HUMAN MOTION
Date
2003-02-13
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
There are two fundamentally different approaches to machine
analysis of human motion: model-based, that use a high-level kinematic
model, and model-free, that use low-level representations of the motion.
Although each approach has its advantages, there is currently a perceptual
gap between the two. This paper describes a new type of kinematic model
that enables us to bridge the perceptual gap. The model is the perceptual
equivalent of passive mechanical models that walk without any control
mechanism. Thus the model not only describes the linkages in a kinematic
chain, it also has an innate resonance that is a gait. If we introduce
control to force the model to synchronize with low-level oscillations
perceived in a video sequence, the result is a gait model that resonates
with an observed gait. We describe a system that demonstrates this new
model and the connection between the model-free and model-based
representations.
Description
Keywords
Computer Science