Game Development Reference
In-Depth Information
We use the motion capture sequence to train the model. Thirty states are used
in the experiment. Each state is observed via MUPs, which are modeled using
a Gaussian model. We assume the covariance matrices of the states to be
diagonal in training. The centers of the trained states are used as key shapes (i.e.,
control points) for the NURBS interpolation scheme. The interpolation-based
temporal model is used in the key-frame-based face animation, such as text-
driven animation in iFACE.
Model-Based Facial Motion Analysis
In this section, we describe model-based facial motion analysis. In the existing
3D non-rigid face tracking algorithm using 3D facial deformation model, the
subspace spanned by the Action Units (AUs) is used as the constraints of low-
level image motion. Similar to MUs, AUs are defined in such a way that arbitrary
facial deformation is approximated by a linear combination of AUs. However,
the AUs are usually manually designed. For these approaches, our automatically
learned MUs can be used in place of the manually designed AUs. In this way,
extensive manual intervention can be avoided and natural facial deformation can
be approximated better.
We choose to use the learned MUs in the 3D non-rigid face tracking system
proposed in Tao (1998) because it has been shown to be robust and real-time.
The facial motion observed in an image plane can be represented by
M(R
(
0
V
+
L
P
)
+
T
)
(3)
where
M
is the projection matrix,
V
is the neutral face,
P
L
defines the non-rigid
deformation,
R
is the 3D rotation decided by three rotation angles
[
, and
T
stands for 3D translation.
L
is an
N
×
M
matrix that
]
t
w
,
w
,
w
=
W
x
y
z
[
]
t
contains
M
AUs, each of which is an
M
dimensional vector.
P
=
p
,...,
p
is
1
M
the coefficients of the AUs. To estimate facial motion parameters
{
T
,
W
,
P
}
d
from 2D inter-frame motion
V
2
, the derivative of equation (3) is taken with
D
d
respect to
{
T
,
W
,
P
}
. Then, a linear equation between
V
2
and
{d
T
,
d
W
,
d
P
}
D
Search WWH ::
Custom Search