A | B | C | D | E | F 
 G | H | I | J | K | L | M 
 N | O | P | Q | R | S | T 
 U | V | W | X | Y | Z 
max planck institut informatik
mpii logoMinerva of the Max Planck Society

Learning Kinematic Structures for Motion Analysis and Animation Processing



Primary Investigators: E. de Aguiar and C. Theobalt

In order to biomechanically analyze the motion of a person or in order to map real world performances onto virtual characters, captured marker-trajectories, e.g. 3D trajectories of optical beacons attached to the body, have to be transformed into the motion parameters of a kinematic skeleton model. Although commercial tools exist that assist professionals in performing this transformation, the estimation of kinematic skeletons and their motion parameters is still a labor-intensive, error prone and often inflexible process. In order to automate this difficult task, we have developed a novel fully-automatic algorithm to estimate an articulated skeleton model of a moving subject and its motion parameters from body marker trajectories that were measured with an optical motion capture system. Our method does not require a priori information about the shape and proportions of the tracked subject, can be applied to arbitrary motion sequences, and renders dedicated initialization poses unnecessary. Our approach first identifies individual rigid bodies by means of a variant of spectral clustering. Thereafter, it determines joint positions at each time step of motion through numerical optimization, reconstructs the skeleton topology, and finally enforces fixed bone length constraints. Through experiments, we validated the robustness and efficiency of our algorithm and showed that it outperforms related methods from the literature in terms of accuracy and speed. Our algorithm can also be applied to learn kinematic properties from mesh animations. A recent extension also enables kinematically plausible segmentation of moving meshes which renders useful during post-processing and compression of our captured deforming body models.

Videos:

[Video showing the performance of our approach]

References:

[1] E. de Aguiar, C. Theobalt, H.-P. Seidel, Automatic Learning of Articulated Skeletons from 3D Marker Trajectories. In Proc. of Second International Symposium, ISVC 2006, pp. 485-494, Lake Tahoe, USA. [pdf]

@INPROCEEDINGS{deAguiar_isvc2006,
AUTHOR = {de Aguiar, Edilson and Theobalt, Christian and Seidel, Hans-Peter},
EDITOR = {Bebis, George and Boyle, Richard and Parvin, Bahram and Koracin, Darko and Remagnino, Paolo and Nefian, Ara and Meenakshisundaram, Gopi and Pascucci, Valerio and Zara, Jiri and Molineros, Jose and Theisel, Holger and Malzbender, Tom},
TITLE = {Automatic Learning of Articulated Skeletons from {3D} Marker Trajectories},
BOOKTITLE = {Advances in Visual Computing : Second International Symposium, ISVC 2006, Part I},
PUBLISHER = {Springer},
YEAR = {2006},
VOLUME = {4291},
PAGES = {485--494},
SERIES = {Lecture Notes in Computer Science},
ADDRESS = {Lake Tahoe, NV, USA},
MONTH = {November},
}

Search MPII (type ? for help)