Adapting and Reconfiguring Human Figure Motion Capture Data through the Application of Inverse Kinematics and Biomechanics-Based Optimisation

M. Meredith, "Adapting and Reconfiguring Human Figure Motion Capture Data through the Application of Inverse Kinematics and Biomechanics-Based Optimisation", PhD, 2005. (Supervisor: Dr Steve Maddock). [pdf]

PhD Abstract: This thesis investigates the issue of modifying motion capture data, specifically the reconfiguration process which includes retargeting and individualisation. To perform modifications, a series of novel algorithms are introduced, where the first is grounded in the domain of inverse kinematics and the second is in dynamics. By applying the algorithms to existing motions, it is shown how the tasks of simple retargetting problem, individualisation and injury simulation can be achieved. These are the limit of the inverse kinematics technique. In contrast, the dynamics-based algorithm also provides the ability to add in plausible environmental or force-based changes.

Aside from the algorithms themselves, the reconfiguration of motions demonstrates the most significant portion of this work in that it is possible to take a single piece of motion data from a source actor and spawn many different versions of it in order to produce motions that better portray the build and biomechanical structure of a target character. This addresses the issue of using the same motion for each and every character regardless of its shape and size, which looks unrealistic. The reconfigured motions are produced using an example motion of a source actor and the biomechanical information of the target actor. Comparing the reconfigured motions to the real motions of target actors provides a validation for these techniques.

In addition to the two main threads of work that come from the inverse kinematics and dynamics-based modification algorithms, a new method of processing positional motion capture marker data to result in an animated hierarchical data structure is presented.

Publications