----------------------------------------------------------------------- B O S T O N U N I V E R S I T Y Computer Science Department C O L L O Q U I U M Wednesday, March 27, 11:00 AM (Coffee served at 10:45 AM) Seminar Room / MCS 135 Switching Linear Dynamic Models: Learning, Tracking, and Synthesis of Human Motion Vladimir Pavlovic Boston University Abstract The human figure exhibits complex and rich dynamic behavior that is both nonlinear and time-varying. However, most work on tracking and analysis of figure motion has employed either generic or highly specific hand-tailored dynamic models superficially coupled with models of motion regimes. I will present an alternative class of learned dynamic models known as switching linear dynamic systems (SLDSs). Cast in the framework of dynamic Bayesian networks, SLDS can be applied to analysis, tracking, and synthesis of the human figure motion. Since exact inference in SLDS is intractable, I will introduce two novel approximate inference algorithms based on structured variational and winner-takes-all approximations. Experimental results on learning, analysis, and synthesis of figure dynamics from video data indicate significant advantages and potential of the SLDS approach. Bio: Vladimir Pavlovic is a research assistant professor in the Bioinformatics Program at Boston University. Before joining Boston University, Vladimir was a member of research staff at Compaq's Cambridge Research Laboratory (CRL) in Cambridge, MA. His interests include applications of machine learning and probabilistic inference to problems in computer vision, HCI, bioinformatics, and modeling of complex systems. Vladimir received a Ph.D. in electrical and computer engineering from the University of Illinois at Urbana-Champaign in 1999. Host: Stan Sclaroff ------------------------------------------------------------------------- For colloquium info, including directions, see http://cs-www.bu.edu/colloquium -------------------------------------------------------------------------