VISION

An athlete performs while being filmed by an ordinary camcorder. From the resulting video, we will be able to model his motion in 3-D, re-synthesize the video as it would have been from a different angle to help him or her visualize what was done right and what was done wrong, and virtually modify the movement to show what should have been done.

SCIENTIFIC TOPICS

We will explore, develop and experiment new methods and techniques for creating parametric models of athletic motions, such as a golf swing or a walking pattern, from video sequences. They will be used both to increase the robustness of computer animation techniques and to allow motion re-synthesis for pedagogic purposes.

These models will eventually form the basis of a teaching tool to be used by professional coaches. A very similar scenario can be envisioned if one replaces the athlete discussed above by an orthopedics patient whose gait must be analyzed and diagnosed by a physician.

Tracking the transition between walking and running from a single video. In the first four frames the subject is running. The transition occurs in the following three frames and the sequence ends with running. Our system tracks the motion and represents it using the 3-D skeleton whose projection appears overlaid in white on the images.

One frame of a synthetic video showing a specific running style transposed to four virtual humans of different sizes .

APPLICATIONS

Athletic training. Diagnostic and outcome evaluation for orthopedic patients. Sophisticated man-machine interfaces.

RESULTS

We have introduced new methods and algorithms to modify an existing motion movement (e.g., golf swing). To visualize and validate the capabilities of the proposed methods a system was constructed. The system works correctly. The interface is easy to interact; the motion adaptations can be easily done by using the mouse pointer. The new methods and algorithms were validated through the editing of a golf swing motion. The proposed approach is able to provide a motion solution that preserves the intrinsic characteristics of the original movement and with a low computation cost.

MAIN PUBLICATIONS

Carvalho,S.R., Boulic,R.,Thalmann,D.: Motion Pattern Preserving IK Operating in the Motion Principal Coefficients Space. 15-th International Conference in Central Europe on Computer Graphics, Visualization and Computer Vision (WSCG2007). January 29 - February 1, 2007.