Virtual Character Definition and Animation within the MPEG-4 Standard - 3D Modeling and Animation: Synthesis and Analysis Techniques for the Human Body - page 56

Since the compression schemes developed in the MPEG-4 FBA framework
offer good performances, the two algorithms (predictive and DCT-based) have
been adopted for compressing the SMS animation data.

SMS versus FBA

A comparative analysis of the FBA and SMS frameworks is synthesized in Table 2
below.
While the FBA is founded on the representation of avatars as segmented
characters, that makes it an appropriate framework for cartoon-like applications,
SMS offers a higher degree of realistic representation, dealing with the concept
of skeleton-driven animation.
When dealing with the avatar body, the FBA standardizes a fixed number of
animation parameters (296) by attaching to each anatomical segment up to three
rotation angles. The SMS framework does not limit the number of animation
parameters (bones and muscles). Moreover, the animation parameters refer to
an extended set of geometrical transformations (rotations, translations, scaling
factors).
Shape deformations are present in both frameworks. For example, the FBA
standardizes a number of control points in order to perform facial deformations,
while the SMS allows us to add curve-based deformers at any level of the skin.
In FBA, the deformation tools are cluster-based. In SMS, they are curve-based.
The FBA standardizes the number and location of the control points, while SMS

Table 2. Main FBA and SMS features (Preda, 2002b).

Criteria FBA SMS

Model type Virtual human character Generic virtual character

Geometry definition Segmented character Seamless character

Hierarchy Standardized Hierarchy Hierarchy build on a generic

skeleton

Local deformation Cluster based for face,

deformation tables for body

Curve-based deformation

Scene graph nodes Define a Face and Body Node and

use H-Anim PROTOs for
specifying the model geometry

Define a own set of 6 nodes

Animation
parameters

296 for body, 68 for face Undefined number of parameters,

arbitrary number of bones and
muscles are supported

Animation editing
support

Forward kinematics Forward kinematics, inverse

kinematics, temporal frame
interpolation

Compression Frame predictive-based, DCT

based

Frame predictive-based, DCT based