Capturing Articulated Body Motion from Video

The human body motion is one example of articulated motion. Video-based human motion capturing provides a non-invasive means to analyze human actions, which is desirable in many applications such as gait analysis, gesture recognition, and rehabilitation. The big challenge lies in the fact that the human body is a very articulated kinematical structure with over 30 degrees of freedom that demands tremendous computation for conventional methods.

We advocate the collaborative approach to this problem. Our method is based on a dynamic Markov network model, a generative model which characterizes the dynamics, the image observations of individual subparts, and the motion constraints among them. The motion estimates can be approximated by the mean field fixed-point of this network. Based on this analytical result, we have designed the mean field Monte Carlo (MFMC) algorithm, in which a set of low dimensional particle filters collaborate to solve the high dimensional motion estimation problem with linear complexity. In our later work, we proposed a new method to estimate the human body poses from a single image, by combining the probabilistic variational analysis with Markov chain Monte Carlo strategies for an efficient data-driven sampling. This gives an automatic initialization for MFMC to continuously capture the body articulation in video sequences.


Demo sequences:
 
2-link arm
3-link finger
6-link upperbody
10-link full body


Publication:
  1. Gang Hua and Ying Wu, "Variational Maximum a Posteriori by Annealed Mean Field Analysis", IEEE Trans. on Pattern Analysis and Machine Intelligence, Vol.27, No.11, pp.1747-1781, Nov., 2005.  [PDF]

  2.  
  3. Gang Hua and Ying Wu, "Sequential Mean Field Variational Analysis of Structured Deformable Shapes", Computer Vision and Image Understanding, Vol.101, No.2., pp.87-99, Feb., 2006.  [PDF]

  4.  
  5. Gang Hua and Ying Wu, "Capturing Human Body Motion from Video for Perceptual Interfaces by Sequential Variational MAP", invited, in Proc.11th Int'l Conf. on Human-Computer Interaction (HCII'05), Las Vegas, Nevada, July 2005.   [PDF]

  6.  
  7. Ying Wu, Gang Hua and Ting Yu, "Tracking Articulated Body by Dynamic Markov Network", In Proc. IEEE Int'l Conf. on Computer Vision (ICCV'03), pp.1094-1101, Nice, France, Oct., 2003.  [PDF]


Return to HD-Motion Research


Updated 10/2005. Copyright © 2001-2006 Ying Wu