Biomedical Engineering Reference
In-Depth Information
3.8.2 Weight-Normalized Minimum-Norm Filter
The weight-normalized minimum-norm filter has been proposed by Dale et al. and
the method is often called as dynamic statistical parametric mapping (dSPM) [
15
].
The idea is to normalize the minimum-norm weight with its weight's norm to ensure
that the spatial distribution of the noise is uniform. The scalar-type weight is thus
given by
G
−
1
l
G
−
1
l
(
r
)
(
r
)
w
(
r
)
=
)
=
l
T
.
(3.79)
G
−
1
l
(
r
G
−
2
l
(
r
)
(
r
)
The idea of weight normalization can be extended to derive the weight matrix for
the vector-type spatial filter, such that
G
−
1
L
(
r
)
W
(
r
)
=
tr
L
T
)
.
(3.80)
G
−
2
L
(
r
)
(
r
Using this weight matrix, the source vector is estimated as
G
−
1
y
L
T
(
r
)
(
t
)
W
T
s
(
r
,
t
)
=
(
r
)
y
(
t
)
=
tr
L
T
)
.
(3.81)
G
−
2
L
(
r
)
(
r
The weight matrix for the
L
2
-regularized version is given by
)
−
1
L
(
G
+
ʾ
I
(
r
)
W
(
r
)
=
tr
L
T
)
.
(3.82)
(
)(
+
ʾ
)
−
2
L
(
r
G
I
r
3.8.3 sLORETA Filter
Standardized low resolution electromagnetic tomography (sLORETA) was originally
proposed by Pascual-Marqui [
16
]. The method can be reformulated as a nonadaptive
spatial filte
r. In this metho
d, the minimum-norm filter outputs are normalized by the
quantity
l
T
G
−
1
l
. This normalization is called as the standardization. The
scalar-type weight is given by
(
r
)
(
r
)
lG
−
1
w
(
r
)
=
l
T
.
(3.83)
G
−
1
l
(
r
)
(
r
)
The extension to the vector-type sLORETA filter results in the weight matrix
expressed as
G
−
1
L
L
T
G
−
1
L
)
]
−
1
/
2
W
(
r
)
=
(
r
)
[
(
r
)
(
r
,
(3.84)