Motion Synthesis with Decoupled Parameterization

1
Computer Graphics International 2008
Motion Synthesis with Decoupled Parameterization
Dongwook Ha and JungHyun Han Korea University
1
2
Motion blending
2
3
Parameterization

• Parameterization provides natural controls.

• To approximate the inverse function, scattered
  data interpolation techniques are used.
• Rose et al. 98 Rose et al. 01 Kovar and
  Gleicher 04

3
4
Parameterization
Parameter space
4
5
Basic Idea

• In the previous approaches,
• the number of required example motions is
  exponential in the dimensionality of the
  parameter space.
• Our Solution - Decoupled parameterization

exponential growth
5
6
Motion control with decoupled parameters
6
7
Time-warping

• Temporal alignment within the example motions

• Scheme for time-warping Shin and Oh 06
• computes the reference time for the current frame.

7
8
Double-layered time-warping

• 1st layer Time-warping for motion blending
• 2nd layer Time-warping for splicing

• 1st layer

8
9
Double-layered time-warping

• 2nd Layer
• Temporal alignment between two reference motions
• Scheme for time-warping

9
10
Double-layered time-warping

• without double-layered time-warping
• with double-layered time-warping

10
11
Re-composition of parameter space

• The samples for the upper body are dynamic.
• Moreover, our attaching operation changes the
  orientation of the pelvis adequately. Heck and
  Gleicher 06
• But, the relative positions among samples are
  preserved.

11
12
Re-composition of parameter space

• The parameters of the samples are approximated.
• The error range of the method is less than 1
  centimeter.

12
13

Video
13
14
Summary

• Motion synthesis with decoupled parameterization
• hybrid approach
• blending-based parametric motion synthesis
• motion splicing
• reduction of the number of required motion data
• allows high-dimensional parameterization.
• can reuse the parametric motion spaces.

14