Computer Assisted Skill Acquisition
Terry Caelli
Department of Computing Science
University of Alberta, Canada
Mittwoch, 09.07.2003, 14 Uhr c.t., H 16
In this presentation we consider how 3D sensors can be used to encode and
predict complex actions using hidden Markov models (HMM). Of particular
interest is the development of a type of stochastic differential geometry of
component action trajectories, their decomposition into screws and the
degrees to which HMMs can encode such actions for both recognition and
prediction. In addition, we consider some tests for evaluating the
usefulness of HMMs in terms of the properties of their parameters and just
how each component of HMMs contributes to predicting optimal state
sequences.
Bio
Dr. Terry Caelli. His interests lie in Computer Vision, Pattern Recognition
and Artificial Intelligence and their applications to intelligent sensing,
image interpretation and computer assisted perception and action systems. He is
Professor of Computing Science at the University of Alberta, Canada. He is a
Fellow of the International Association for Pattern Recognition and a
Fellow of the Institute for Electronic and Electrical Engineers (IEEE).