Human Motion Identification

A Multidisciplinary University Research Initiative

  

Overview | People | Participating Institutions | Slide presentation | Contacts                                                     Site Last Modified: February 26, 2003


© 2003  All Rights Reserved                                                                                                                Human Motion Identification

Overview


In this MURI initiative, we are investigating human motion from a wide range of perspectives. The research can be categorized into five main areas: The tables below summarize sub-categories of research in each of the main areas, and describe where this research is taking place. For additional details on research in a particular area, please see the specific web site associated with the particular institution, which is linked from the list of participating institutions.

Tracking

Tracking research by institution
As pointsCourse scaleKinematic
Berkeley-Using optical flowUsing point features
Using body parts
Caltech--Using point motion features
NYU---
USCUsing range data-Using space carving

Motion representation

Motion representation research by institution
DeformableImage basedShort and medium time scaleLong time scale
Berkeley-Using optical flowClustering marker positionsModified clustering
Caltech-Parsing silhouettesClustering marker positionsHMM's
NYUBy tracking points/regions---
USC----

Motion classification

Motion classification research by institution
From 2DFrom 3DCompositional annotationsBulk motion annotation
BerkeleyUsing optical flowLifted tracker outputSupport vector machinesOnline learning
Motion capture
CaltechUsing point features-- -
Using clustered motions
NYU----
USC-Using clustered motions--

Motion synthesis

Motion synthesis research by institution
Cut and paste 2DCut and paste 3D3D from annotationBy statistical model
BerkeleyMatchingSearch motion graphSearch weighted by annotation-
Caltech----
NYU---Multi-resolution motion texture
USC---From clusters

Data collection

Data collection efforts
IndividualsMotion capture dataLinked motion capture and videoRange dataBulk video
Berkeley-Electronic Arts, SonyFrom Stanford-Hours, Week, Month
CaltechVideoMotion capture setupMotion capture setup--
NYU-Motion capture setup---
USC---Pointlike people-



People


There are eight faculty doing active research on human motion identification as part of this MURI program. Their web pages are linked below. Additional post-doctoral researchers, graduate students, and staff can be found at the university-specific sites.
  • Chris Bregler (NYU)
  • John Canny (Berkeley)
  • David Forsyth (Berkeley)
  • Michael Jordan (Berkeley)
  • Jitendra Malik (PI) (Berkeley)
  • Maja Mataric' (USC)
  • Pietro Perona (Caltech)
  • Stuart Russell (Berkeley)

    Participating Institutions


    Research for the human motion identification MURI is taking place at the following four research universities: Each of the links above takes you to a web page describing the specific research efforts at that site.


    Contacts


    If you have questions about this MURI on human motion identification, please contact Erik Miller by email at egmil@eecs.berkeley.edu