3D Motion Editing Through B-Spline Fitting with Constraints

This paper proposes a novel motion editing algorithm that uses human motion capture data for animating 3D characters. First, the algorithm fits the 2D root joint trajectory with the cubic B-Spline through least-square minimization. In this process, it finds the optimal number of control points based on an error threshold. Once it gets n control points, users are allowed to change the positions of control points, which is able to create a new trajectory. The new trajectory is then fed into the original motion so that it is modified to reflect the new trajectory. To find exact parameter t of the spline curve representing root joint position, the algorithm performed the arc-length parameterization on the curve. Since the motions are forced to change the root joint positions, the result may violate the fidelity of the original motions, which may cause some artifacts such as foot skating. To fix them, the IK (Inverse Kinematics) solver is applied to motions to change the limb orientation. Although the IK solver can change the orientation of original motions, if the differences between the modified trajectory and original trajectory are too big, then the result motion produces awkward poses over times. In order to prevent them, our algorithm puts constraints on the control points of curve automatically so that users are able to edit the trajectory freely without considering whether it produces natural motions or not.