hide
Free keywords:
-
Abstract:
Human motion capturing can be regarded as an optimization problem where one
searches for the pose that minimizes a previously defined error function based
on some image features. Most approaches for solving this problem use iterative
methods like gradient descent approaches. They work quite well as long as they
do not get distracted by local optima.
We introduce a novel approach for global optimization that is suitable for the
tasks as they occur during human motion capturing. We call the method
interacting simulated annealing since it is based on an interacting particle
system that converges to the global optimum similar to simulated annealing.
We provide a detailed mathematical discussion that includes convergence results
and annealing properties. Moreover, we give two examples that demonstrate
possible applications of the algorithm, namely a global optimization problem
and a multi-view human motion capturing task including segmentation,
prediction, and prior knowledge. A quantative error analysis also indicates the
performance and the robustness of the interacting simulated annealing
algorithm.
