Analyzing Human Motion
Key Difficulty: Complex Time Trajectories
Our Approach: Multi-scale statistical models
Not only do we want to capture the movement of people,we would also like to be able to model that movement in formal ways.
This is a hard problem, because of the complexity of the motions of people and the interplay between those motion components.
We have a new view of this problem, that builds on prior work in statistical image modeling, and projective geometry, that we think may radically improve the ability to model human motion. This is shown in the bottom image, where a set of views of a running person have been combined to create a model of that motion, shown from a view entirely different from any of the acquisition views.