Database generation
In recent years, much work has been devoted to increasing the robustness of people tracking algorithms by introducing motion models[1]. The models usually are composed of a single activity and are linear, such as the well known Principal Component Analysis (PCA). In this projet the student will do the following:
 |
1. Build a multiactivity database: for example: playing basketball, dancing, walking in different types of terrain including turning, etc. 2. Apply different learning techniques to "learn" such database in a probabilistic way: mixture of gaussians, kernel density estimation, gaussian processes[2], etc. |
Figure 1: Basket motion from CMU database.
Reading:
[1] R. Urtasun and P. Fua. 3D Human Body Tracking using Deterministic Motion Models. In European Conference in Computer Vision (ECCV), Prague, Czech Republic, May 2004.
[2] K. Grochow, S. L. Martin, A. Hertzmann and Z. Popovic. Style-based Inverse Kinematics. Computer Graphics, SIGGRAPH proceedings (2004).