I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
.
14
, N
o.
3
,
J
une
2025
, pp.
2302
~
2314
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
14
.i
3
.pp
2302
-
2314
2302
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
H
yb
r
i
d
m
e
t
h
od
f
or
op
t
i
m
i
z
i
n
g e
m
ot
i
on
r
e
c
ogn
i
t
i
on
m
od
e
l
s o
n
e
l
e
c
t
r
oe
n
c
e
p
h
al
ogr
am
si
gn
a
l
s
I
M
ad
e
A
gu
s
Wi
r
aw
an
1
, K
ad
e
k
Y
ot
a E
r
n
an
d
a A
r
yan
t
o
1
,
I
N
.
S
u
k
aj
aya
2
, N
i
N
yom
an
M
e
s
t
r
i
A
gu
s
t
in
i
3
,
D
e
w
i
A
r
u
m
Wi
d
h
iy
an
t
i
M
e
t
r
a
P
u
t
r
i
4
1
D
a
t
a
S
c
i
e
nc
e
L
a
b, D
e
pa
r
t
m
e
nt
of
I
nf
or
m
a
t
i
c
E
ngi
ne
e
r
i
ng,
F
a
c
ul
t
y
o
f
E
ngi
ne
e
r
i
ng a
nd V
oc
a
t
i
ona
l
, U
ni
ve
r
s
i
t
a
s
P
e
ndi
di
ka
n G
a
ne
s
ha
,
S
i
nga
r
a
j
a
, I
ndone
s
i
a
2
D
e
pa
r
t
m
e
nt
of
M
a
t
he
m
a
t
i
c
s
,
F
a
c
ul
t
y
of
M
a
t
he
m
a
t
i
c
s
a
nd N
a
t
ur
a
l
S
c
i
e
nc
e
s
,
U
n
i
ve
r
s
i
t
a
s
P
e
ndi
di
ka
n
G
a
ne
s
ha
, S
i
nga
r
a
j
a
, I
ndone
s
i
a
3
D
e
pa
r
t
m
e
nt
N
e
ur
ol
ogy a
nd M
e
di
c
a
l
R
e
ha
bi
l
i
t
a
t
i
on
,
F
a
c
ul
t
y of
M
e
di
c
i
ne
, U
ni
v
e
r
s
i
t
a
s
P
e
ndi
di
ka
n G
a
ne
s
ha
, S
i
nga
r
a
j
a
, I
ndone
s
i
a
4
E
duc
a
t
i
on of
C
hi
l
dr
e
n, A
dol
e
s
c
e
nt
s
, A
dul
t
s
a
nd T
he
i
r
C
ha
r
a
c
t
e
r
i
s
t
i
c
s
L
a
b
,
F
a
c
ul
t
y of
E
duc
a
t
i
on
, U
ni
ve
r
s
i
t
a
s
P
e
ndi
di
ka
n G
a
ne
s
ha
,
S
i
nga
r
a
j
a
, I
ndone
s
i
a
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
A
ug 30, 2024
R
e
vi
s
e
d
F
e
b 3, 2025
A
c
c
e
pt
e
d
M
a
r
15, 2025
Two
critical
factors
that
need
to
be
studied
in
emotion
recognition
are
the
differences
in
electroenceph
alogram
(EEG)
signal
patterns
caus
ed
by
participant
characteristics
and
EEG
signals
with
spatial
information.
These
factors
significant
ly
affect
the
resulting
accuracy.
The
model
propo
sed
in
this
study
can
consider
these
factors.
This
model
consists
of
the
m
odified
weighted
mean
filter
method
for
the
basic
EEG
signal
smoothing
p
rocess,
the diffe
rential e
ntropy method
for the f
eature
extractio
n process
, th
e relativ
e
difference method for
the baseline reduction, the
3D cube method for
f
eature
representation,
and
the
continuous
capsule
network
method
for
the
classifi
cation
process.
Based
on
testing
on
three
public
datasets,
this
hybrid
method
can
overcome
factors
affecting
emotion
recognition
accurac
y.
This
statement
is
based
on
the
accurac
y
produced
by
this
model,
which
outperformed the acc
uracy validated in previous studies.
K
e
y
w
o
r
d
s
:
3D
c
ube
C
ont
in
uous
c
a
ps
ul
e
ne
twor
k
D
if
f
e
r
e
nt
ia
l
e
nt
r
opy
E
le
c
tr
oe
nc
e
pha
lo
gr
a
m
E
m
ot
io
n r
e
c
ogni
ti
on
M
odi
f
ie
d w
e
ig
ht
e
d m
e
a
n f
il
te
r
R
e
la
ti
ve
di
f
f
e
r
e
nc
e
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
I
M
a
de
A
gus
W
ir
a
w
a
n
D
a
ta
S
c
ie
nc
e
L
a
b, D
e
p
a
r
tm
e
nt
of
I
nf
or
m
a
ti
c
E
ng
in
e
e
r
in
g, F
a
c
ul
ty
of
E
ngi
ne
e
r
in
g a
nd V
oc
a
ti
ona
l
U
ni
ve
r
s
it
a
s
P
e
ndi
di
ka
n G
a
ne
s
ha
U
da
ya
na
R
oa
d, S
in
ga
r
a
ja
,
B
a
li
,
I
ndone
s
ia
E
m
a
il
:
im
a
de
.a
gus
w
ir
a
w
a
n@
undi
ks
ha
.
a
c
.i
d
1.
I
N
T
R
O
D
U
C
T
I
O
N
I
t
is
im
por
ta
nt
to
r
e
c
ogni
z
e
e
m
ot
io
na
l
r
e
a
c
ti
ons
in
hum
a
n
s
to
unde
r
s
ta
nd
hum
a
n
m
e
nt
a
l
s
ta
te
s
a
nd
pe
r
f
or
m
a
nc
e
[
1]
,
[
2]
.
S
om
e
e
m
ot
io
n
r
e
c
ogni
ti
on
a
pp
r
oa
c
he
s
c
a
n
be
pe
r
f
or
m
e
d
e
xt
e
r
na
ll
y.
H
ow
e
ve
r
,
in
te
r
na
l
r
e
c
ogni
ti
on
of
e
m
ot
io
ns
,
p
r
im
a
r
il
y
vi
a
e
le
c
tr
oe
nc
e
pha
lo
gr
a
m
(
E
E
G
)
s
ig
na
ls
,
ha
s
s
e
ve
r
a
l
a
dva
nt
a
ge
s
,
s
uc
h
a
s
:
i)
ha
vi
ng s
pa
ti
a
l
in
f
or
m
a
ti
on t
ha
t
r
e
pr
e
s
e
nt
s
huma
n a
f
f
e
c
ti
ve
e
xpe
r
ie
nc
e
s
[
3]
, a
nd
ii
)
r
e
s
pond to e
a
r
ly
c
ha
nge
s
in
hum
a
n
e
m
ot
io
na
l
r
e
a
c
ti
ons
[
4]
,
[
5]
.
A
lt
hough
va
r
io
us
m
e
th
ods
ha
ve
be
e
n
pr
opos
e
d
to
im
pr
ove
th
e
pe
r
f
or
m
a
nc
e
of
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
(
A
I
)
m
ode
ls
in
r
e
c
ogni
z
in
g
e
m
ot
io
ns
,
in
te
r
f
e
r
e
nc
e
in
th
e
E
E
G
s
ig
na
l
a
nd
lo
s
s
of
s
pa
ti
a
l
in
f
or
m
a
ti
on
f
r
om
th
e
E
E
G
s
ig
na
l
c
a
n
r
e
s
ul
t
in
lo
w
m
ode
l
a
c
c
ur
a
c
y
[
6]
.
T
he
r
e
f
or
e
,
e
xa
m
in
in
g
s
e
ve
r
a
l
a
ppr
opr
ia
te
m
e
th
ods
to
pr
oduc
e
opt
im
a
l
A
I
m
ode
ls
to
s
ol
ve
th
e
s
e
two
pr
obl
e
m
s
is
e
s
s
e
nt
ia
l.
G
e
ne
r
a
l
ly
,
tw
o
e
s
s
e
n
ti
a
l
a
s
pe
c
t
s
a
r
e
c
on
s
id
e
r
e
d
i
n
d
e
v
e
l
opi
ng
a
n
e
m
ot
i
on
r
e
c
o
gni
ti
o
n
m
od
e
l
on
E
E
G
s
i
gn
a
l
s
:
p
a
r
ti
c
i
pa
nt
a
nd
E
E
G
s
i
gn
a
l
c
ha
r
a
c
t
e
r
i
s
t
ic
s
[
6]
.
P
a
r
ti
c
i
pa
n
t
c
ha
r
a
c
te
r
i
s
ti
c
s
in
c
lu
d
e
pe
r
s
on
a
li
ty
tr
a
it
s
,
ge
n
de
r
,
c
u
lt
ur
e
,
a
nd
in
te
ll
e
c
t
ua
l
a
bi
li
t
ie
s
.
T
h
is
a
s
pe
c
t
pr
o
du
c
e
s
di
f
f
e
r
e
nt
E
E
G
s
i
gn
a
l
pa
tt
e
r
n
va
r
ia
n
c
e
s
f
or
e
a
c
h
pa
r
t
ic
ip
a
nt
.
I
f
th
i
s
E
E
G
s
i
gn
a
l
pa
tt
e
r
n
va
r
ia
nc
e
is
no
t
a
dd
r
e
s
s
e
d,
it
w
il
l
im
pa
c
t
lo
w
e
m
o
ti
o
n
r
e
c
og
ni
ti
on
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
H
y
br
id
m
e
th
od f
or
opt
imi
z
in
g
e
m
ot
io
n r
e
c
ogni
ti
on m
ode
l
s
on
…
(
I
M
ade
A
gus
W
ir
aw
an
)
2303
a
c
c
ur
a
c
y
[
7
]
.
T
o
o
ve
r
c
om
e
th
i
s
pr
ob
le
m
,
W
ir
a
w
a
n
e
t
al
.
[
8]
pr
opo
s
e
d
a
ba
s
e
l
in
e
r
e
du
c
ti
on
pr
oc
e
s
s
u
s
in
g
th
e
r
e
l
a
ti
ve
d
if
f
e
r
e
nc
e
m
e
th
od
.
T
he
ba
s
e
li
ne
r
e
d
uc
ti
o
n
pr
o
c
e
s
s
r
e
d
uc
e
s
th
e
e
xp
e
r
im
e
n
ta
l
s
i
gn
a
l
f
e
a
tu
r
e
s
w
it
h
th
e
ba
s
e
li
ne
s
ig
n
a
l
f
e
a
t
ur
e
s
[
7]
.
T
h
e
e
xp
e
r
im
e
n
ta
l
s
ig
n
a
l
i
s
a
s
i
gna
l
th
a
t
r
e
pr
e
s
e
nt
s
a
n
e
m
ot
i
on
a
l
c
on
di
t
io
n
.
M
e
a
nw
h
il
e
,
th
e
b
a
s
e
li
n
e
s
ig
na
l
r
e
p
r
e
s
e
nt
s
a
ne
ut
r
a
l
c
on
di
ti
on
[
9
]
.
I
de
a
l
ly
,
n
e
ut
r
a
l
c
ond
it
i
on
s
h
a
v
e
lo
w
e
r
s
i
gn
a
l
a
m
p
li
t
ud
e
va
lu
e
s
t
h
a
n
e
m
ot
io
na
l
c
o
ndi
ti
o
n
s
[
10]
, [
11]
.
H
ow
e
v
e
r
,
th
e
b
a
s
e
li
n
e
s
ig
na
l
do
e
s
no
t
r
e
pr
e
s
e
nt
a
ne
ut
r
a
l
c
on
di
t
io
n
b
e
c
a
us
e
th
e
b
a
s
e
li
ne
a
nd
e
x
pe
r
im
e
nt
a
l
s
ig
na
ls
s
ti
l
l
h
a
v
e
t
he
s
a
m
e
a
m
pl
i
tu
d
e
va
lu
e
s
[
1
2]
–
[
14]
.
T
h
e
hi
gh
a
m
p
li
tu
d
e
of
t
h
e
b
a
s
e
l
in
e
s
i
gn
a
l
is
c
a
us
e
d
b
y i
nt
e
r
n
a
l
a
nd
e
x
te
r
n
a
l
a
r
te
f
a
c
t
s
/i
nt
e
r
f
e
r
e
nc
e
[
9]
,
[
1
5]
–
[
17]
.
W
ir
a
w
a
n
e
t
al
.
[
18]
ha
ve
s
uc
c
e
s
s
f
ul
ly
de
ve
lo
pe
d
a
nd
im
pl
e
m
e
nt
e
d
th
e
m
odi
f
ie
d
w
e
ig
ht
e
d
m
e
a
n
f
il
te
r
m
e
th
od
to
opt
im
iz
e
th
e
ba
s
e
li
ne
r
e
duc
ti
on
pr
oc
e
s
s
w
it
h
th
e
r
e
la
ti
ve
di
f
f
e
r
e
nc
e
m
e
th
od.
H
ow
e
ve
r
,
th
is
m
ode
l
ne
e
ds
to
c
on
s
id
e
r
th
e
s
pa
ti
a
l
c
ha
r
a
c
te
r
is
ti
c
s
of
E
E
G
s
ig
na
ls
[
18]
.
S
e
ve
r
a
l
r
e
s
e
a
r
c
he
r
s
h
a
ve
s
tu
di
e
d
th
e
a
s
pe
c
ts
of
E
E
G
s
ig
na
l
c
ha
r
a
c
te
r
is
ti
c
s
f
r
om
th
e
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
pr
oc
e
s
s
,
f
e
a
tu
r
e
r
e
pr
e
s
e
nt
a
ti
on,
a
nd
c
la
s
s
if
ic
a
ti
on.
I
n
th
e
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
pr
oc
e
s
s
,
th
e
di
f
f
e
r
e
nt
ia
l
e
nt
r
opy
m
e
th
od
c
a
n
r
e
pr
e
s
e
nt
s
pa
ti
a
l
in
f
or
m
a
ti
on
a
nd
is
s
ta
bl
e
in
e
m
ot
io
n
r
e
c
ogni
ti
on
c
om
pa
r
e
d
to
ot
he
r
f
e
a
tu
r
e
s
[
7]
,
[
19]
–
[
21]
.
I
n
th
e
f
e
a
tu
r
e
r
e
pr
e
s
e
nt
a
ti
on
pr
oc
e
s
s
,
th
e
3D
c
ub
e
m
e
th
od
c
a
n
m
a
in
ta
in
s
pa
ti
a
l
in
f
or
m
a
ti
on
be
twe
e
n
c
ha
nne
ls
a
nd
be
twe
e
n
th
e
ta
, a
lp
ha
, be
ta
, a
nd ga
m
m
a
f
r
e
que
nc
i
e
s
[
22]
. I
n t
e
r
m
s
of
c
la
s
s
if
ic
a
ti
on, s
e
ve
r
a
l
r
e
s
e
a
r
c
he
r
s
ha
ve
s
tu
di
e
d t
he
m
ul
ti
-
c
la
s
s
s
uppor
t
ve
c
to
r
m
a
c
hi
ne
[
23]
,
gr
a
ph
r
e
gul
a
r
iz
e
d
e
xt
r
e
m
e
le
a
r
ni
ng
m
a
c
hi
ne
[
24]
,
c
onvolut
io
n
ne
ur
a
l
ne
twor
k
[
25]
,
[
26
]
,
a
nd
th
e
c
om
bi
na
ti
on
of
lo
ng
s
hor
t
-
te
r
m
m
e
m
or
y
w
it
h
c
onvolut
io
n
ne
ur
a
l
ne
twor
k
[
27]
.
H
ow
e
ve
r
,
of
th
e
s
e
ve
r
a
l
c
la
s
s
if
ic
a
ti
on
m
e
th
ods
s
tu
di
e
d,
th
e
c
a
p
s
ul
e
n
e
twor
k
m
e
th
od
ha
s
be
e
n
pr
ove
n
to
r
e
pr
e
s
e
nt
s
pa
ti
a
l
in
f
or
m
a
ti
on
in
th
e
c
la
s
s
if
ic
a
ti
on
p
r
oc
e
s
s
[
22]
.
W
ir
a
w
a
n
e
t
al
.
[
22]
pr
opos
e
d
a
c
ont
in
uous
c
a
ps
ul
e
n
e
twor
k
m
e
th
od
to
pr
e
s
e
r
ve
s
pa
ti
a
l
in
f
or
m
a
ti
on
f
r
om
E
E
G
s
ig
na
ls
f
or
c
la
s
s
if
ic
a
ti
on.
H
ow
e
ve
r
, t
he
noi
s
e
i
n E
E
G
s
ig
na
ls
i
n t
hi
s
m
ode
l
ha
s
ye
t
to
be
o
ve
r
c
om
e
.
T
he
pr
opos
e
d
pr
obl
e
m
-
s
ol
vi
ng
a
ppr
oa
c
h
f
or
th
e
f
ir
s
t
pr
obl
e
m
is
to
a
ppl
y
th
e
m
odi
f
ie
d
w
e
ig
ht
e
d
m
e
a
n
f
il
te
r
m
e
th
od
to
r
e
m
ove
in
te
r
f
e
r
e
nc
e
in
th
e
ba
s
e
li
ne
s
ig
na
l
a
nd
a
ppl
y
th
e
r
e
la
ti
ve
di
f
f
e
r
e
nc
e
m
e
th
od
t
o
e
li
m
in
a
te
in
te
r
f
e
r
e
nc
e
in
th
e
e
xpe
r
im
e
nt
a
l
s
ig
na
l.
F
or
th
e
s
e
c
ond
pr
obl
e
m
,
th
is
r
e
s
e
a
r
c
h
pr
opos
e
s
th
e
di
f
f
e
r
e
nt
ia
l
e
nt
r
opy
m
e
th
od
to
e
xt
r
a
c
t
ba
s
e
li
ne
a
nd
e
xpe
r
im
e
nt
a
l
s
ig
na
l
f
e
a
tu
r
e
s
,
th
e
3D
c
ube
m
e
th
od
t
o
r
e
pr
e
s
e
nt
e
xpe
r
im
e
nt
a
l
s
ig
n
a
l
f
e
a
tu
r
e
s
,
a
nd
th
e
c
ont
in
uo
us
c
a
p
s
ul
e
ne
twor
k
m
e
th
od
f
or
e
m
ot
io
n
c
la
s
s
if
ic
a
ti
on.
T
hi
s
hybr
id
m
e
th
od
is
e
xpe
c
te
d
to
im
pr
ove
th
e
p
e
r
f
or
m
a
nc
e
of
e
m
ot
io
n
r
e
c
ogni
ti
on
m
ode
ls
on
E
E
G
s
ig
na
ls
.
T
he
m
ode
l
w
a
s
te
s
te
d
w
it
h
th
r
e
e
e
m
ot
io
n
da
ta
s
e
ts
,
s
uc
h
a
s
D
E
A
P
,
D
R
E
A
M
E
R
,
a
nd
A
M
I
G
O
S
,
to
m
e
a
s
ur
e
th
e
pe
r
f
or
m
a
nc
e
of
th
e
pr
opos
e
d
m
ode
l.
T
he
s
e
th
r
e
e
da
ta
s
e
t
s
ha
ve
di
f
f
e
r
e
nt
c
ha
r
a
c
te
r
is
ti
c
s
r
e
ga
r
di
ng
th
e
a
m
ount
of
da
ta
,
num
be
r
of
c
ha
nn
e
ls
, a
nd
r
e
c
or
di
ng
s
c
e
n
a
r
io
s
.
T
o
pr
oduc
e
a
n
opt
im
a
l
m
ode
l
f
or
r
e
c
ogni
z
in
g e
m
ot
io
ns
vi
a
E
E
G
s
ig
na
ls
.
2.
M
E
T
H
O
D
B
a
s
e
d
on
th
e
s
tu
di
e
s
,
th
e
r
e
s
e
a
r
c
he
r
s
[
18]
,
[
22]
pr
opos
e
a
m
ode
l
c
ons
is
ti
ng
of
s
e
ve
r
a
l
m
e
th
ods
,
s
uc
h
a
s
th
e
m
odi
f
ie
d
w
e
ig
ht
e
d
m
e
a
n
f
il
te
r
m
e
th
od
f
or
r
e
m
ovi
ng
ba
s
e
li
ne
E
E
G
s
ig
na
l
a
r
te
f
a
c
ts
,
th
e
di
f
f
e
r
e
nt
ia
l
e
nt
r
opy
m
e
th
od
f
or
f
e
a
tu
r
e
e
xt
r
a
c
ti
on,
th
e
r
e
la
ti
ve
di
f
f
e
r
e
nc
e
f
or
ba
s
e
li
ne
r
e
duc
ti
on,
th
e
3D
c
ube
m
e
th
od
f
or
f
e
a
tu
r
e
r
e
pr
e
s
e
nt
a
ti
on,
a
nd
th
e
c
ont
in
uous
c
a
ps
ul
e
ne
twor
k
m
e
th
od
f
or
th
e
e
m
ot
io
n
c
la
s
s
if
ic
a
ti
on
pr
oc
e
s
s
.
T
hi
s
e
xp
e
r
im
e
nt
w
il
l
be
c
onduc
te
d
on
th
r
e
e
publ
ic
da
t
a
s
e
t
s
.
T
he
f
lo
w
di
a
gr
a
m
r
e
la
te
d
to
th
e
r
e
s
e
a
r
c
h
pr
oc
e
dur
e
s
th
a
t
w
il
l
be
im
p
le
m
e
nt
e
d
is
in
F
ig
ur
e
1.
B
a
s
e
d
on
F
ig
ur
e
1,
th
e
r
e
s
e
a
r
c
h
pr
oc
e
dur
e
c
a
n
be
de
s
c
r
ib
e
d
in
t
he
s
ub
s
e
c
ti
on a
s
f
ol
lo
w
s
.
F
ig
ur
e
1. R
e
s
e
a
r
c
h f
lo
w
di
a
gr
a
m
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
3
,
J
une
2025
:
2302
-
2314
2304
2.1.
C
ol
le
c
t
io
n
of
d
at
as
e
t
s
T
he
da
ta
s
e
ts
u
s
e
d
in
th
i
s
r
e
s
e
a
r
c
h
a
r
e
D
E
A
P
,
D
R
E
A
M
E
R
,
a
nd
A
M
I
G
O
S
[
12]
–
[
14
]
.
T
he
s
e
th
r
e
e
da
ta
s
e
ts
us
e
di
f
f
e
r
e
nt
E
E
G
r
e
c
or
di
ng
t
ool
s
, r
e
c
or
di
ng du
r
a
ti
ons
, a
nd r
e
c
or
di
ng s
c
e
na
r
io
s
. T
he
s
e
di
f
f
e
r
e
nc
e
s
i
n
c
ha
r
a
c
te
r
is
ti
c
s
w
il
l
be
i
n
s
tr
um
e
nt
a
l
in
t
e
s
ti
ng t
he
r
e
li
a
bi
li
ty
of
t
he
m
ode
l
be
in
g de
ve
lo
pe
d.
2.2.
S
e
gm
e
n
t
at
io
n
T
he
s
e
gm
e
nt
a
ti
on
pr
oc
e
s
s
s
e
pa
r
a
te
s
th
e
ba
s
e
li
ne
s
ig
na
l
a
nd
th
e
tr
ia
l
s
ig
na
l.
T
he
ba
s
e
li
ne
s
ig
na
l
is
a
n
E
E
G
s
ig
na
l
th
a
t
r
e
pr
e
s
e
nt
s
t
he
pa
r
ti
c
ip
a
nt
'
s
ne
ut
r
a
l
c
ondi
ti
on. T
he
t
r
ia
l
s
ig
na
l
is
a
n E
E
G
s
ig
na
l
th
a
t
r
e
pr
e
s
e
nt
s
th
e
pa
r
ti
c
ip
a
nt
'
s
e
m
ot
io
na
l
r
e
a
c
ti
on.
I
n
th
e
D
E
A
P
da
ta
s
e
t,
th
e
f
ir
s
t
th
r
e
e
s
e
c
onds
a
r
e
th
e
ba
s
e
li
ne
s
ig
na
l
s
,
a
nd
th
e
f
our
th
to
s
ix
ty
-
th
ir
d
s
e
c
onds
a
r
e
th
e
e
xpe
r
im
e
nt
a
l
s
ig
na
ls
.
I
n
D
R
E
A
M
E
R
a
nd
A
M
I
G
O
S
,
th
e
f
i
r
s
t
f
iv
e
s
e
c
onds
a
r
e
th
e
ba
s
e
li
ne
s
ig
na
l,
a
nd
th
e
f
ol
lo
w
in
g
s
e
c
onds
a
r
e
th
e
tr
ia
l
s
ig
na
l
[
12]
–
[
14
]
.
E
ve
r
y
s
e
c
ond
of
E
E
G
s
ig
na
l
da
ta
on
D
E
A
P
,
D
R
E
A
M
E
R
,
a
nd
A
M
I
G
O
S
c
ons
is
ts
of
a
128
H
z
s
a
m
pl
in
g
r
a
te
.
F
ig
ur
e
2
is
a
n
il
lu
s
tr
a
ti
on
of
s
e
gm
e
nt
a
ti
on
in
th
e
A
M
I
G
O
S
da
ta
s
e
t.
S
e
gm
e
nt
a
ti
on
is
pe
r
f
or
m
e
d
on
a
ll
c
ha
nne
ls
.
T
he
D
E
A
P
da
ta
s
e
t
c
ont
a
in
s
32 c
ha
nne
l
s
, w
hi
le
t
he
D
R
E
A
M
E
R
a
nd A
M
I
G
O
S
da
ta
s
e
ts
c
ont
a
in
14
[
12]
–
[
14]
.
F
ig
ur
e
2. T
he
E
E
G
s
ig
na
l
s
e
gm
e
nt
a
ti
on pr
oc
e
s
s
f
or
A
F
3 a
nd A
F
4 c
ha
nne
ls
on t
he
A
M
I
G
O
S
da
ta
s
e
t
2.3. Re
m
ovi
n
g
ar
t
if
ac
t
s
T
he
m
odi
f
ie
d
w
e
ig
ht
e
d
m
e
a
n
f
il
te
r
m
e
th
od
r
e
m
ove
s
a
r
te
f
a
c
t
s
by
s
m
oot
hi
ng
th
e
b
a
s
e
li
ne
s
ig
na
l.
H
ow
e
ve
r
,
be
f
or
e
th
a
t,
th
e
ba
s
e
li
ne
s
ig
na
l
da
ta
w
a
s
nor
m
a
li
z
e
d
us
in
g
th
e
Z
-
s
c
or
e
nor
m
a
li
z
a
ti
on
m
e
th
od
[
5]
,
[
28]
.
T
hi
s
pr
oc
e
s
s
a
im
s
to
pr
oduc
e
w
e
ig
ht
va
lu
e
s
f
r
om
a
ba
s
e
li
n
e
s
ig
na
l
th
a
t
is
ty
pi
c
a
ll
y
di
s
tr
ib
ut
e
d
(
0<
x<
1)
[
25]
. T
hi
s
w
e
ig
ht
va
lu
e
w
il
l
be
us
e
d i
n t
he
m
odi
f
ie
d w
e
ig
ht
e
d m
e
a
n f
il
te
r
m
e
th
od (
1)
.
=
(
∑
+
+
=
−
(
2
+
1
)
∑
+
=
−
)
(
1)
T
he
va
lu
e
j
=
n,
n+1,
n+2…
.,
m
+2n,
n
i
s
th
e
w
in
dow
le
ngt
h,
w
hi
le
th
e
va
lu
e
m
is
th
e
a
m
ount
of
da
ta
.
∑
+
=
−
is
a
pr
oc
e
s
s
t
o nor
m
a
li
z
e
t
he
w
e
ig
ht
va
lu
e
s
t
o m
e
e
t
th
e
r
e
qui
r
e
m
e
nt
s
∑
=
1
[
29]
.
2.4. De
c
om
p
os
it
io
n
D
e
c
om
pos
it
io
n
a
im
s
to
s
e
pa
r
a
te
f
our
f
r
e
que
nc
y
ba
nd
s
in
th
e
b
a
s
e
li
ne
a
nd
tr
ia
l
s
ig
na
ls
.
T
hi
s
pr
oc
e
s
s
us
e
s
th
e
B
ut
te
r
w
or
th
f
il
te
r
m
e
th
od
(3
rd
or
de
r
)
,
w
he
r
e
th
e
E
E
G
s
ig
na
l
f
r
e
que
nc
y
f
il
te
r
in
g
pr
oc
e
s
s
is
ba
s
e
d
on
th
e
lo
w
-
pa
s
s
a
nd
hi
gh
-
pa
s
s
r
a
nge
(
ba
ndpa
s
s
f
il
te
r
)
.
T
he
de
c
om
pos
it
io
n
pr
oc
e
s
s
w
a
s
c
a
r
r
ie
d
out
f
or
a
ll
c
ha
nne
ls
f
or
ba
s
e
li
ne
a
nd
tr
ia
l
s
ig
na
ls
[
7]
.
2.5.
F
e
at
u
r
e
e
xt
r
ac
t
io
n
T
he
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
pr
oc
e
s
s
is
c
a
r
r
ie
d
out
to
obt
a
in
r
e
le
va
nt
E
E
G
s
ig
na
l
f
e
a
tu
r
e
s
.
E
a
c
h
f
r
e
que
nc
y
ba
nd
pe
r
f
or
m
s
it
e
ve
r
y
s
e
c
ond
(
128
H
z
s
a
m
pl
in
g
r
a
te
)
f
or
ba
s
e
li
ne
a
nd
tr
ia
l
s
ig
na
ls
.
T
hi
s
r
e
s
e
a
r
c
h
u
s
e
s
th
e
di
f
f
e
r
e
nt
ia
l
e
nt
r
opy me
th
od (
2)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
H
y
br
id
m
e
th
od f
or
opt
imi
z
in
g
e
m
ot
io
n r
e
c
ogni
ti
on m
ode
l
s
on
…
(
I
M
ade
A
gus
W
ir
aw
an
)
2305
ℎ
(
)
=
1
2
l
o
g
(
2
2
(
)
)
(
2)
W
he
r
e
e
is
E
ul
e
r
'
s
c
ons
ta
nt
,
δ
2
i
is
th
e
i
th
va
r
ia
nc
e
in
de
x,
X
is
th
e
th
e
ta
,
a
lp
ha
,
be
ta
,
or
ga
m
m
a
f
r
e
que
nc
y
ba
nd, a
nd
h
i
is
t
he
di
f
f
e
r
e
nt
ia
l
e
nt
r
opy va
lu
e
f
or
t
he
E
E
G
s
ig
na
l
a
t
th
e
i
th
in
de
x.
2.6.
B
as
e
li
n
e
r
e
d
u
c
t
io
n
T
hi
s
pr
oc
e
s
s
a
im
s
to
pr
oduc
e
f
e
a
tu
r
e
v
a
lu
e
s
on
tr
ia
l
s
ig
na
l
s
th
a
t
c
a
n
c
ha
r
a
c
te
r
iz
e
pa
r
ti
c
ip
a
nt
s
'
e
m
ot
io
na
l
r
e
a
c
ti
ons
to
c
la
s
s
if
y
[
7]
,
[
30]
.
T
hi
s
s
tu
dy
u
s
e
s
th
e
r
e
la
ti
ve
di
f
f
e
r
e
nc
e
m
e
th
od
to
r
e
duc
e
th
e
tr
ia
l
s
ig
na
l
f
e
a
tu
r
e
s
w
it
h
th
e
ba
s
e
li
ne
s
ig
na
l
f
e
a
tu
r
e
s
[
8]
,
[
31]
.
A
s
a
n
in
it
ia
l
s
ta
ge
,
th
e
ba
s
e
li
ne
s
ig
na
l'
s
a
ve
r
a
g
e
f
e
a
tu
r
e
va
lu
e
f
or
e
a
c
h f
r
e
que
nc
y ba
nd
w
a
s
c
a
lc
ul
a
te
d (
3
)
.
(
)
=
∑
(
)
=
1
(
3)
(
)
is
th
e
di
f
f
e
r
e
nt
ia
l
e
nt
r
opy
f
e
a
tu
r
e
va
lu
e
of
f
r
e
que
nc
y
ba
nd
X
a
t
th
e
k
th
in
de
x
f
or
th
e
ba
s
e
li
ne
s
ig
na
l.
(
)
is
th
e
a
ve
r
a
ge
di
f
f
e
r
e
nt
ia
l
e
nt
r
opy
f
e
a
tu
r
e
va
lu
e
f
or
f
r
e
que
nc
y
ba
nd
X
in
th
e
ba
s
e
li
ne
s
ig
na
l.
T
he
va
lu
e
k
=
1,
2,..,
N
,
w
he
r
e
N
de
f
in
e
s
th
e
num
be
r
o
f
di
f
f
e
r
e
nt
ia
l
e
nt
r
opy
f
e
a
tu
r
e
va
lu
e
s
f
or
e
a
c
h
e
xpe
r
im
e
nt
,
c
ha
nne
l,
a
nd
p
a
r
ti
c
ip
a
nt
f
r
om
th
e
ba
s
e
li
ne
s
ig
na
l
[
8]
.
T
he
m
e
a
n
va
lu
e
obt
a
in
e
d
is
th
e
n
us
e
d
f
or
th
e
ba
s
e
li
ne
r
e
duc
ti
on pr
oc
e
s
s
f
or
t
he
t
r
ia
l
s
ig
na
l
(
4)
.
(
)
=
(
)
(
)
(
4)
W
he
r
e
(
)
is
th
e
f
e
a
tu
r
e
va
lu
e
of
f
r
e
que
nc
y
ba
nd
j
(
X
)
r
e
pr
e
s
e
nt
s
th
e
ba
s
e
li
ne
r
e
du
c
ti
on
va
lu
e
of
f
r
e
que
nc
y
ba
nd
X
a
t
th
e
j
th
in
de
x
f
or
th
e
E
E
G
e
xpe
r
im
e
nt
a
l
s
ig
na
l.
V
a
lu
e
j=
1,
2,
..,
M
,
w
he
r
e
M
is
th
e
num
be
r
of
f
e
a
tu
r
e
va
lu
e
s
f
or
e
a
c
h e
xpe
r
im
e
nt
, c
ha
nne
l,
a
nd p
a
r
ti
c
ip
a
nt
i
n t
he
t
r
ia
l
s
ig
na
l
.
2.7.
F
e
at
u
r
e
r
e
p
r
e
s
e
n
t
at
io
n
A
f
te
r
t
he
ba
s
e
li
ne
r
e
duc
ti
on pr
oc
e
s
s
, t
he
t
r
ia
l
s
ig
na
l
f
e
a
tu
r
e
va
lu
e
s
a
r
e
r
e
pr
e
s
e
nt
e
d us
in
g t
he
3D
c
ube
m
e
th
od. T
he
3D
c
ube
i
s
ba
s
e
d on the
i
nt
e
r
na
ti
ona
l
s
ys
te
m
10
-
20 s
ta
nda
r
d f
or
c
ha
nne
l
pl
a
c
e
m
e
nt
on t
he
s
c
a
lp
.
T
he
in
te
r
na
ti
ona
l
s
ys
te
m
s
ta
nda
r
d
10
-
20
m
a
p
s
s
pa
ti
a
l
in
f
or
m
a
ti
on
be
twe
e
n
c
ha
nne
ls
by
r
e
pr
e
s
e
nt
in
g
tr
ia
l
s
ig
na
l
f
e
a
tu
r
e
s
in
to
a
9×
9
m
a
tr
ix
.
T
h
e
s
iz
e
of
th
is
r
e
pr
e
s
e
nt
a
ti
on
is
ba
s
e
d
on
th
e
m
a
xi
m
um
num
be
r
of
E
E
G
c
ha
nne
ls
pl
a
c
e
d
on
th
e
he
a
d.
T
hi
s
r
e
pr
e
s
e
nt
a
ti
on
a
im
s
to
r
e
pr
e
s
e
nt
s
p
a
ti
a
l
in
f
or
m
a
ti
on
be
twe
e
n
a
dj
a
c
e
nt
c
ha
nne
ls
.
F
ig
ur
e
3
r
e
pr
e
s
e
nt
s
di
f
f
e
r
e
nt
ia
l
e
nt
r
opy
f
e
a
tu
r
e
s
m
a
p
pe
d
in
to
a
9×
9
m
a
tr
ix
[
7]
.
T
hi
s
r
e
pr
e
s
e
nt
a
ti
on
is
ba
s
e
d
on
e
a
c
h
f
r
e
que
nc
y
ba
nd,
w
he
r
e
r
e
d,
ye
ll
ow
,
gr
e
e
n,
a
nd
bl
ue
r
e
pr
e
s
e
nt
ga
m
m
a
,
be
ta
,
a
lp
ha
,
a
nd
th
e
ta
f
r
e
que
nc
ie
s
[
7]
.
T
he
D
E
A
P
a
nd
D
R
E
A
M
E
R
da
ta
s
e
ts
pr
oduc
e
2400
a
nd
3728
3D
c
ube
s
f
or
e
a
c
h
pa
r
ti
c
ip
a
nt
.
F
ur
th
e
r
m
or
e
,
e
a
c
h
pa
r
ti
c
ip
a
nt
ge
ne
r
a
te
d
6192
3D
c
ube
s
f
or
lo
ng
e
xpe
r
im
e
nt
s
on
th
e
A
M
I
G
O
S
da
ta
s
e
t
a
nd
1394 3D c
ube
s
f
or
s
hor
t
e
xpe
r
im
e
nt
s
.
F
ig
ur
e
3. F
e
a
tu
r
e
r
e
pr
e
s
e
nt
a
ti
on f
or
e
a
c
h f
r
e
que
nc
y ba
nd us
in
g
a
3D
c
ube
2.7. Cla
s
s
if
i
c
at
io
n
T
he
c
ont
in
uou
s
c
a
p
s
ul
e
n
e
twor
k
m
e
th
od
r
e
pr
e
s
e
nt
s
s
p
a
ti
a
l
in
f
o
r
m
a
ti
on
in
th
e
c
la
s
s
if
ic
a
ti
on
pr
oc
e
s
s
.
T
he
a
r
c
hi
te
c
tu
r
e
of
th
e
c
ont
in
uous
c
a
ps
ul
e
ne
twor
k
m
e
th
od
is
s
how
n
in
F
ig
u
r
e
4
[
22]
.
T
he
c
ont
in
uous
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
3
,
J
une
2025
:
2302
-
2314
2306
c
a
ps
ul
e
ne
twor
k a
r
c
hi
te
c
tu
r
e
in
F
ig
ur
e
4
is
di
vi
de
d
in
to
th
r
e
e
p
a
r
ts
:
c
ont
in
uous
c
onvolut
io
n,
pr
im
a
r
y
c
a
p
s
ul
e
,
a
nd e
m
ot
io
n c
a
ps
ul
e
.
F
ig
ur
e
4. C
ont
in
uous
c
a
ps
ul
e
ne
twor
k a
r
c
hi
te
c
tu
r
e
[
22]
2.7.1. Cont
in
u
ou
s
c
on
vol
u
t
io
n
A
t
th
i
s
s
t
a
g
e
,
th
i
s
pr
o
c
e
s
s
i
s
c
a
r
r
i
e
d
o
ut
f
our
t
im
e
s
t
o pr
o
du
c
e
f
o
ur
f
e
a
tu
r
e
m
a
p
s
.
T
h
e
f
e
a
tu
r
e
m
a
p
s
w
i
ll
be
c
r
e
a
t
e
d
i
n
th
e
1
st
c
o
nvo
lu
t
io
n,
2
nd
c
on
vol
ut
i
on,
3
rd
c
onv
ol
u
ti
on,
a
nd
4
th
c
on
vo
lu
ti
on
w
it
h
s
i
z
e
s
of
9
×
9
×
64,
9
×
9
×
12
8,
9
×
9
×
2
56
,
a
n
d
9
×
9
×
64
,
r
e
s
p
e
c
ti
v
e
l
y.
T
h
e
f
i
lt
e
r
s
u
s
e
d
a
r
e
2
×
2
×
4
w
it
h
64
f
or
1
st
c
on
vol
ut
i
on,
2
×
2
×
4
w
it
h
1
28
f
or
2
nd
c
onv
ol
u
ti
o
n,
2
×
2
×
4
w
it
h
256
f
o
r
3
rd
c
o
nvo
lu
ti
on
,
a
n
d
1
×
1
×
4
w
i
th
6
4
f
or
4
th
c
onv
ol
ut
i
on.
F
ur
th
e
r
m
or
e
,
th
e
s
tr
id
e
v
a
lu
e
i
n
e
a
c
h
c
on
vol
ut
i
on
i
s
1.
I
n
a
d
di
ti
on,
th
e
a
c
t
iv
a
ti
on
f
or
e
a
c
h
c
on
vol
ut
i
on
i
s
r
e
c
ti
f
i
e
d
l
in
e
a
r
u
ni
t
(
R
e
L
U
)
.
I
n
t
he
c
ont
in
uou
s
c
on
vol
ut
i
on
a
ppr
oa
c
h,
t
he
p
a
d
di
ng
u
s
e
d
is
t
he
S
A
M
E
a
nd
d
oe
s
not
us
e
p
ool
in
g,
s
o
th
e
di
m
e
ns
io
ns
of
t
he
s
i
z
e
of
t
he
f
e
a
t
ur
e
m
a
p
s
pr
od
uc
e
d
in
a
ll
c
o
nvo
lu
t
io
n
s
r
e
m
a
i
n t
h
e
s
a
m
e
;
onl
y
th
e
d
e
pt
h
of
th
e
f
e
a
t
ur
e
m
a
p
c
ha
ng
e
s
.
T
h
is
pr
o
c
e
s
s
a
im
s
t
o
m
a
in
t
a
i
n
th
e
s
p
a
ti
a
l
in
f
or
m
a
ti
o
n
of
a
ll
E
E
G
c
h
a
nn
e
l
s
a
c
c
or
di
ng
t
o
th
e
p
os
it
i
on
i
n
th
e
f
ir
s
t
da
ta
i
np
ut
[
7]
.
T
h
e
n
um
b
e
r
o
f
pa
r
a
m
e
t
e
r
s
g
e
n
e
r
a
t
e
d
i
n
th
e
1
st
c
on
vo
lu
t
io
n
,
t
h
e
2
nd
c
onv
ol
ut
i
on,
t
he
3
rd
c
onv
ol
ut
i
on,
a
nd
th
e
4
th
c
on
vo
lu
t
io
n
a
r
e
1,
08
8
(
64
×
(2
×
2
×
4)
+
1)
,
3
2.
896
(
128
×
(2
×
2
×
64
)
+
1)
,
13
1.
328
(
2
56
×
(2
×
2
×
12
8)
+
1)
,
a
nd
1
6.4
48
(
6
4
×
(1
×
1
×
2
56)
+
1)
.
2.7.2. P
r
im
ar
y c
ap
s
u
le
A
t
th
is
s
ta
ge
,
th
e
f
e
a
tu
r
e
m
a
p
ge
n
e
r
a
te
d
f
r
om
th
e
4
th
c
onvolut
io
n
is
s
e
p
a
r
a
te
d
in
to
e
ig
ht
bl
oc
ks
, e
a
c
h
bl
oc
k
w
it
h
a
s
iz
e
9
×
9
×
8.
N
e
xt
,
e
a
c
h
bl
oc
k
is
r
e
s
h
a
pe
d
in
to
a
la
ye
r
of
8
×
1
c
a
ps
ul
e
s
(
u
i
)
s
o
th
a
t
one
bl
oc
k
w
il
l
pr
oduc
e
81
c
a
ps
ul
e
s
[
22]
. T
he
pr
im
a
r
y c
a
ps
ul
e
r
e
pr
e
s
e
nt
s
t
he
n
um
be
r
of
c
a
ps
ul
e
s
a
t
th
e
l
ow
e
r
l
e
ve
l
i
(
u
i
)
w
it
h
a
s
iz
e
of
1
×
8,
s
o t
he
r
e
a
r
e
648 c
a
p
s
ul
e
l
a
ye
r
s
g
e
ne
r
a
te
d f
r
om
t
he
e
ig
ht
bl
oc
ks
.
2.7.3. E
m
ot
io
n
c
ap
s
u
le
T
he
a
f
f
in
e
tr
a
ns
f
or
m
a
ti
on
a
nd
dyna
m
ic
r
out
in
g
pr
oc
e
s
s
e
s
a
r
e
c
a
r
r
ie
d
out
a
t
th
is
s
ta
g
e
.
F
ir
s
t,
th
e
e
m
ot
io
n
c
a
ps
ul
e
pr
oc
e
s
s
be
gi
ns
by
in
it
ia
li
z
in
g
j
=
1
a
nd
i
=
1.
T
h
e
va
r
ia
bl
e
j
r
e
pr
e
s
e
nt
s
th
e
num
be
r
of
e
m
ot
io
n
c
la
s
s
e
s
to
be
r
e
c
ogni
z
e
d
(
v
j
)
,
w
hi
le
th
e
va
r
ia
bl
e
i
r
e
pr
e
s
e
nt
s
th
e
num
be
r
of
c
a
p
s
ul
e
da
t
a
(
)
.
W
he
n
th
e
va
r
ia
bl
e
s
i
a
nd
j
a
r
e
1,
th
e
c
a
p
s
ul
e
da
ta
1
w
il
l
be
th
e
f
ir
s
t
i
nput
da
ta
in
th
e
e
m
ot
io
n
c
a
p
s
ul
e
pr
oc
e
s
s
a
nd
r
e
la
te
d
to
th
e
1
st
out
put
c
la
s
s
(
v
1
)
.
T
he
f
ol
lo
w
in
g
p
r
oc
e
s
s
of
th
e
e
m
ot
io
n
c
a
ps
ul
e
is
a
n
a
f
f
in
e
tr
a
ns
f
or
m
a
ti
on.
H
ow
e
ve
r
,
be
f
or
e
th
is
pr
oc
e
s
s
is
c
a
r
r
ie
d
out
,
th
e
in
it
ia
l
in
it
ia
li
z
a
ti
on
pr
oc
e
s
s
of
th
e
w
e
ig
ht
m
a
tr
ix
,
is
c
a
r
r
ie
d
out
f
ir
s
t.
T
he
a
f
f
in
e
tr
a
ns
f
or
m
a
ti
on
p
r
oc
e
s
s
w
il
l
pr
oduc
e
u
̂
,
(
ve
c
to
r
)
.
A
f
te
r
̂
,
is
obt
a
in
e
d,
th
e
r
out
in
g
in
it
ia
li
z
a
ti
on
p
r
oc
e
s
s
(
r
=0
)
is
c
a
r
r
ie
d
out
a
nd
c
ont
in
ue
d
w
it
h
th
e
in
it
ia
li
z
a
ti
on
of
th
e
va
lu
e
b
i
,
j
=
0.
T
he
r
va
lu
e
de
te
r
m
in
e
s
how
m
a
ny
dyna
m
ic
r
out
in
g
pr
oc
e
s
s
e
s
a
r
e
c
a
r
r
ie
d
out
.
T
he
dyna
m
ic
r
out
in
g
pr
oc
e
s
s
is
c
a
r
r
ie
d
out
by
de
f
a
ul
t
th
r
ic
e
[
22]
.
T
he
dyn
a
m
ic
r
out
in
g
p
r
oc
e
s
s
c
ons
is
t
s
of
c
a
lc
ul
a
ti
ng
th
e
c
oupl
in
g
c
oe
f
f
ic
ie
nt
c
i
,
j
,
w
e
ig
ht
e
d
s
um
(
s
j
)
,
a
c
ti
va
ti
ng
s
qua
s
hi
ng
(
v
j
)
,
a
nd
th
e
upda
te
d
va
lu
e
of
b
i
,
j
.
T
he
b
i
,
j
va
lu
e
de
te
r
m
in
e
s
th
e
c
oupl
in
g
c
oe
f
f
ic
ie
nt
c
i
,
j
va
lu
e
.
A
f
te
r
th
e
va
lu
e
o
f
c
i
,
j
is
obt
a
in
e
d,
th
e
w
e
ig
ht
e
d
s
um
pr
oc
e
s
s
is
c
a
r
r
ie
d
out
to
pr
oduc
e
th
e
ve
c
to
r
va
lu
e
s
j
.
T
he
va
lu
e
of
s
j
obt
a
in
e
d
is
f
ur
th
e
r
a
c
ti
va
te
d
w
it
h
th
e
s
qua
s
hi
ng
f
unc
ti
on
to
m
a
ke
th
e
ve
c
to
r
va
lu
e
v
j
.
A
f
te
r
th
e
va
lu
e
of
th
e
ve
c
to
r
v
j
is
r
e
c
e
iv
e
d,
th
e
r
out
in
g
va
lu
e
(
r
)
is
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
H
y
br
id
m
e
th
od f
or
opt
imi
z
in
g
e
m
ot
io
n r
e
c
ogni
ti
on m
ode
l
s
on
…
(
I
M
ade
A
gus
W
ir
aw
an
)
2307
c
he
c
ke
d.
T
h
e
dyna
m
ic
r
out
in
g
pr
oc
e
s
s
is
r
e
pe
a
te
d
if
th
e
va
lu
e
of
r
<
3.
T
hi
s
pr
oc
e
s
s
be
gi
ns
w
it
h
upda
ti
ng
th
e
va
lu
e
of
b
i
,
j
,
th
e
n
c
a
lc
ul
a
ti
ng
th
e
w
e
ig
ht
e
d
s
um
(
s
j
)
a
nd
th
e
s
qu
a
s
hi
ng
a
c
ti
va
ti
on
pr
oc
e
s
s
(
v
j
)
.
T
he
f
in
a
l
va
lu
e
of
v
j
in
th
e
3
rd
r
out
in
g
is
th
e
n
pr
oc
e
s
s
e
d
to
de
t
e
r
m
in
e
th
e
va
lu
e
of
N
or
m
||
v
j
||
a
nd
L
2
R
e
gul
a
r
iz
a
ti
on.
D
e
te
r
m
in
in
g
th
e
va
lu
e
of
||v
j
||
(
N
o
r
m
)
is
th
e
e
nd
of
th
e
e
m
ot
io
n
c
a
ps
ul
e
pr
oc
e
s
s
f
or
th
e
f
ir
s
t
c
a
ps
ul
e
da
ta
(
1
)
in
t
he
1
st
e
poc
h.
F
ur
th
e
r
m
or
e
,
f
or
th
e
e
m
ot
io
n
c
a
ps
ul
e
pr
oc
e
s
s
on
th
e
f
ol
lo
w
in
g
c
a
ps
ul
e
da
ta
(
i
)
,
th
e
ba
t
c
h
va
lu
e
c
he
c
ki
ng pr
oc
e
s
s
i
s
c
a
r
r
ie
d out f
ir
s
t;
i
f
t
he
ba
tc
h va
lu
e
i
s
a
m
ul
ti
pl
e
of
2, t
he
n t
he
w
e
ig
ht
va
lu
e
upda
te
pr
oc
e
s
s
,
is
c
a
r
r
ie
d
out
.
T
he
w
e
ig
ht
upda
te
pr
oc
e
s
s
us
e
s
th
e
R
M
S
P
r
op
m
e
th
od
w
it
h
m
om
e
nt
um
=
0.09.
A
f
te
r
c
he
c
ki
ng t
he
ba
tc
h va
lu
e
, t
he
ne
xt
s
te
p i
s
t
o c
he
c
k t
he
va
lu
e
of
va
r
ia
bl
e
i
. I
f
t
he
va
lu
e
of
va
r
ia
bl
e
i
is
l
e
s
s
t
ha
n
648
(
i
<
648)
,
th
e
n
th
e
in
c
r
e
m
e
nt
pr
oc
e
s
s
(
i
+
+
)
of
va
r
ia
bl
e
i
is
c
a
r
r
ie
d
out
.
T
hi
s
in
c
r
e
m
e
nt
pr
oc
e
s
s
d
e
te
r
m
in
e
s
th
e
f
ol
lo
w
in
g
c
a
ps
ul
e
da
ta
(
i
)
to
be
pr
oc
e
s
s
e
d
in
th
e
e
m
ot
io
n
c
a
ps
ul
e
.
A
f
te
r
th
a
t,
th
e
a
f
f
in
e
tr
a
ns
f
or
m
a
ti
on
a
nd
dyna
m
ic
r
out
in
g
pr
oc
e
s
s
e
s
a
r
e
e
xe
c
ut
e
d
a
ga
in
.
T
hi
s
pr
oc
e
s
s
i
s
r
e
pe
a
te
d
unt
il
th
e
648
th
c
a
ps
ul
e
da
ta
.
W
he
n
th
e
648
th
c
a
p
s
ul
e
da
ta
ha
s
f
in
is
h
e
d
e
xe
c
ut
in
g,
th
e
ne
x
t
s
te
p
is
to
c
he
c
k
th
e
v
a
lu
e
of
va
r
ia
bl
e
j
.
I
f
va
r
ia
bl
e
j
is
le
s
s
th
a
n
4
(
j
<
4)
,
th
e
va
lu
e
in
c
r
e
m
e
nt
pr
oc
e
s
s
of
va
r
ia
bl
e
j
is
c
a
r
r
ie
d
out
(
j
+
+
)
.
T
hi
s
in
c
r
e
m
e
nt
pr
oc
e
s
s
de
te
r
m
in
e
s
th
e
c
la
s
s
v
j
th
a
t
w
il
l
be
pr
oc
e
s
s
e
d
ne
xt
in
th
e
e
m
ot
io
n
c
a
ps
ul
e
.
T
hi
s
pr
oc
e
s
s
w
il
l
s
to
p
w
he
n
648
da
ta
c
a
ps
ul
e
s
(
)
ha
ve
be
e
n
tr
a
in
e
d
f
or
a
ll
e
m
ot
io
n
c
la
s
s
e
s
(
j
=
4)
.
F
ig
u
r
e
5
onl
y
de
s
c
r
ib
e
s
th
e
e
m
ot
io
n
c
a
ps
ul
e
f
lo
w
f
or
648
da
ta
c
a
p
s
ul
e
s
(
)
in
4
c
la
s
s
e
s
(
v
j
)
in
1
st
e
poc
h
[
30]
,
[
32]
,
[
33]
.
T
he
f
our
c
la
s
s
e
s
out
put
,
na
m
e
ly
:
hi
gh
a
r
ous
a
l
pos
it
iv
e
va
le
nc
e
(
H
A
P
V
)
,
hi
gh
a
r
ous
a
l
ne
ga
ti
ve
va
le
nc
e
(
H
A
N
V
)
,
lo
w
a
r
ous
a
l
ne
ga
ti
ve
va
le
nc
e
(
L
A
N
V
)
, a
nd l
ow
a
r
ous
a
l
po
s
it
iv
e
va
le
nc
e
(
L
A
P
V
)
.
F
ig
ur
e
5. F
lo
w
c
ha
r
t
e
m
ot
io
n c
a
ps
ul
e
2.9. E
val
u
at
io
n
m
e
t
h
od
T
hi
s
s
tu
dy
us
e
s
th
e
k
-
f
ol
d
c
r
os
s
-
va
li
da
ti
on
e
va
lu
a
ti
on
m
e
th
od
(
k=
10)
to
e
va
lu
a
te
th
e
pr
opos
e
d
m
ode
l.
T
hi
s
m
ode
l
w
il
l
m
e
a
s
ur
e
th
e
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F
1
s
c
or
e
va
lu
e
s
f
or
th
e
th
r
e
e
publ
ic
da
ta
s
e
ts
s
tu
di
e
d:
D
E
A
P
,
D
R
E
A
M
E
R
,
a
nd
A
M
I
G
O
S
.
T
hi
s
va
lu
e
is
us
e
d
a
s
th
e
ba
s
is
f
or
m
e
a
s
ur
in
g
th
e
pe
r
f
or
m
a
nc
e
of
t
he
pr
opos
e
d m
ode
l.
3.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
T
he
hybr
id
m
ode
l
c
a
n
pr
oduc
e
hi
gh
a
c
c
ur
a
c
y
on
th
r
e
e
publ
ic
da
ta
s
e
ts
ba
s
e
d
on
th
e
e
va
lu
a
ti
on
m
e
th
od.
T
hi
s
m
ode
l
is
a
ls
o
s
upe
r
io
r
to
s
e
ve
r
a
l
e
m
ot
io
n
r
e
c
o
gni
ti
on
m
ode
ls
in
p
r
e
vi
ous
s
tu
di
e
s
.
F
ig
ur
e
6
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
3
,
J
une
2025
:
2302
-
2314
2308
pr
e
s
e
nt
s
th
e
pr
opos
e
d
m
ode
l'
s
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F
1
s
c
or
e
va
lu
e
s
f
or
r
e
c
ogni
z
in
g
f
our
e
m
ot
io
n
c
la
s
s
e
s
f
r
om
e
a
c
h pa
r
ti
c
ip
a
nt
on t
he
D
E
A
P
da
ta
s
e
t.
F
ig
ur
e
6. A
c
c
ur
a
c
y, pr
e
c
is
io
n, r
e
c
a
ll
, a
nd F
1 s
c
or
e
gr
a
phs
f
or
r
e
c
ogni
z
in
g f
our
e
m
ot
io
n c
la
s
s
e
s
on t
h
e
D
E
A
P
da
ta
s
e
t
B
a
s
e
d
on
F
ig
ur
e
6,
s
e
v
e
r
a
l
pa
r
ti
c
ip
a
nt
s
w
it
h
I
D
1,
2,
3,
4,
5,
6, 7,
8,
9,
10,
13,
14,
15,
16,
17,
18,
20,
21,
23,
26,
27,
28,
30,
31,
a
nd
32
m
a
na
ge
d
to
a
c
hi
e
ve
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F
1
s
c
or
e
of
100%
.
I
n
c
ont
r
a
s
t,
th
e
lo
w
e
s
t
a
c
c
ur
a
c
y,
pr
e
c
i
s
io
n,
r
e
c
a
ll
,
a
nd
F
1
w
e
r
e
a
c
hi
e
ve
d
by
pa
r
ti
c
ip
a
nt
I
D
22
a
t
98.29%
,
98.20%
,
98.31%
,
a
nd
98.24%
.
O
ve
r
a
ll
,
th
e
a
ve
r
a
ge
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F
1
s
c
or
e
s
a
c
hi
e
ve
d
w
e
r
e
99.91%
,
99.92%
,
99.92%
,
a
nd
99.91%
,
r
e
s
pe
c
ti
ve
ly
.
I
n
a
ddi
ti
on,
s
ta
nda
r
d
de
vi
a
ti
on
va
lu
e
s
f
or
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F
1
a
r
e
0.32,
0.33,
0.31,
a
nd
0.32,
r
e
s
pe
c
t
iv
e
ly
.
S
o,
it
c
a
n
be
s
ta
te
d
th
a
t
th
e
di
s
tr
ib
ut
io
n
of
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F
1
s
c
or
e
of
e
a
c
h
pa
r
ti
c
ip
a
nt
is
c
lo
s
e
r
to
th
e
a
ve
r
a
ge
va
lu
e
c
om
pa
r
e
d
to
th
e
m
ode
l
pr
opos
e
d
in
[
18]
,
[
22]
.
F
u
r
th
e
r
m
or
e
,
F
ig
u
r
e
7
pr
e
s
e
nt
s
th
e
pr
opos
e
d
m
ode
l'
s
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
, a
nd F
1 s
c
or
e
f
or
e
a
c
h pa
r
ti
c
ip
a
nt
on t
he
D
R
E
A
M
E
R
da
t
a
s
e
t.
F
ig
ur
e
7. A
c
c
ur
a
c
y, pr
e
c
is
io
n, r
e
c
a
ll
, a
nd F
1 s
c
or
e
gr
a
phs
f
or
r
e
c
ogni
z
in
g f
our
e
m
ot
io
n c
la
s
s
e
s
on t
h
e
D
R
E
A
M
E
R
da
ta
s
e
t
B
a
s
e
d
on
F
ig
ur
e
7,
pa
r
ti
c
ip
a
nt
s
w
it
h
I
D
12
a
c
hi
e
ve
d
a
c
c
ur
a
c
y,
r
e
c
a
ll
,
a
nd
F
1
s
c
or
e
s
of
99.03%
,
98.93%
,
a
nd
98.97%
,
r
e
s
pe
c
ti
ve
ly
,
w
hi
le
pa
r
ti
c
ip
a
nt
s
w
it
h
I
D
01
a
c
hi
e
ve
d
th
e
hi
ghe
s
t
pr
e
c
is
io
n
a
t
99.07%
.
O
n
th
e
ot
he
r
ha
nd,
th
e
lo
w
e
s
t
a
c
c
ur
a
c
y
a
nd
pr
e
c
is
io
n
w
e
r
e
a
c
hi
e
ve
d
by
p
a
r
ti
c
ip
a
nt
I
D
20
a
t
92.55%
a
nd
92.99%
,
r
e
s
pe
c
ti
ve
ly
.
I
n
c
om
pa
r
is
on,
th
e
lo
w
e
s
t
r
e
c
a
ll
a
nd
F
1
s
c
or
e
s
w
e
r
e
a
c
hi
e
ve
d
by
p
a
r
ti
c
ip
a
nt
I
D
13
a
t
89.43%
a
nd
91.55%
,
r
e
s
pe
c
ti
ve
ly
.
A
lt
hough
non
e
a
c
hi
e
ve
d
1
00%
a
c
c
ur
a
c
y,
ove
r
a
ll
,
th
e
a
ve
r
a
ge
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F
1
w
e
r
e
a
c
hi
e
ve
d
a
t
96.98%
,
96.82%
,
95
.89%
,
a
nd
96.28%
,
r
e
s
p
e
c
ti
ve
ly
.
B
a
s
e
d
on
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
H
y
br
id
m
e
th
od f
or
opt
imi
z
in
g
e
m
ot
io
n r
e
c
ogni
ti
on m
ode
l
s
on
…
(
I
M
ade
A
gus
W
ir
aw
an
)
2309
s
ta
nda
r
d
de
vi
a
ti
on
v
a
lu
e
s
f
or
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F
1
s
c
or
e
w
e
r
e
a
c
hi
e
ve
d
a
t
1.53,
1.62,
2.32,
a
nd
1.97, r
e
s
pe
c
ti
ve
ly
. S
o, i
t
c
a
n be
s
ta
te
d t
ha
t
th
e
di
s
tr
ib
u
ti
on of
a
c
c
ur
a
c
y, pr
e
c
is
io
n, r
e
c
a
ll
, a
nd F
1 s
c
or
e
of
e
a
c
h
pa
r
ti
c
ip
a
nt
i
s
qui
te
c
lo
s
e
t
o t
he
a
v
e
r
a
ge
va
lu
e
c
om
p
a
r
e
d t
o t
he
m
ode
l
pr
opos
e
d
in
[
18]
, [
22]
.
F
in
a
ll
y,
in
F
ig
ur
e
8,
13
pa
r
ti
c
ip
a
nt
s
w
it
h
I
D
1,
2,
3,
6,
7,
9,
19
,
20,
25,
26,
28,
31,
a
nd
34
a
c
hi
e
ve
d
a
c
c
ur
a
c
ie
s
,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F
1
s
c
or
e
of
100%
e
a
c
h.
F
ur
th
e
r
m
or
e
,
19
ot
he
r
pa
r
ti
c
ip
a
nt
s
pr
oduc
e
d
a
n
a
ve
r
a
ge
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F
1
s
c
or
e
a
bove
99
%
.
O
ve
r
a
ll
,
th
e
a
ve
r
a
ge
da
ta
f
or
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F
1
s
c
or
e
w
e
r
e
99.94%
,
99.93%
,
99.9
3%
,
a
nd
99.93%
.
I
n
a
ddi
ti
on,
th
e
s
t
a
nda
r
d
de
vi
a
ti
on
va
lu
e
s
f
or
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
a
nd
F
1
s
c
or
e
a
r
e
0.10,
0.11,
0.13,
a
nd
0.12,
r
e
s
pe
c
ti
ve
ly
.
S
o,
it
c
a
n be
s
ta
te
d t
ha
t
th
e
di
s
tr
ib
ut
io
n o
f
a
c
c
ur
a
c
y, pr
e
c
is
io
n, r
e
c
a
ll
, a
nd F
1 s
c
or
e
f
r
om
e
a
c
h pa
r
ti
c
ip
a
nt
i
s
c
lo
s
e
r
to
th
e
a
ve
r
a
ge
va
lu
e
c
om
pa
r
e
d
to
th
e
m
ode
l
pr
opos
e
d
in
[
18]
,
[
22]
.
F
ur
th
e
r
m
or
e
,
to
m
e
a
s
ur
e
th
e
s
uc
c
e
s
s
of
th
e
i
m
pl
e
m
e
nt
a
ti
on of
t
he
p
r
opos
e
d m
ode
l,
t
he
a
ve
r
a
ge
a
c
c
ur
a
c
y va
lu
e
s
i
n t
hi
s
e
xpe
r
im
e
nt
a
r
e
c
om
pa
r
e
d w
it
h
th
e
a
ve
r
a
ge
a
c
c
ur
a
c
y
va
lu
e
s
f
r
om
s
e
v
e
r
a
l
pr
e
vi
ous
s
tu
di
e
s
,
e
s
pe
c
ia
ll
y
f
or
th
e
r
e
c
ogni
ti
on
of
f
our
e
m
ot
io
n
c
la
s
s
e
s
, a
s
i
n T
a
bl
e
1.
F
ig
ur
e
8. A
c
c
ur
a
c
y, pr
e
c
is
io
n, r
e
c
a
ll
, a
nd F
1 s
c
or
e
gr
a
phs
f
or
r
e
c
ogni
z
in
g f
our
e
m
ot
io
n c
la
s
s
e
s
on t
h
e
A
M
I
G
O
S
da
ta
s
e
t
T
a
bl
e
1. C
om
pa
r
is
on of
t
he
a
v
e
r
a
ge
a
c
c
ur
a
c
y of
t
he
f
our
e
m
ot
io
n c
la
s
s
e
s
i
n t
hi
s
s
tu
dy w
it
h
s
e
ve
r
a
l
pr
e
vi
ous
s
tu
di
e
s
M
e
t
hod
D
E
A
P
(%)
D
RE
A
M
E
R
(%)
A
M
IG
O
S
(%)
F
e
a
t
ure
e
xt
ra
c
t
i
on
:
a
ngl
e
pl
ot
,
c
l
a
s
s
i
fi
c
a
t
i
on:
m
ul
t
i
-
c
l
a
s
s
s
upport
ve
c
t
or
m
a
c
hi
ne
[23]
81.
67
-
-
F
e
a
t
ure
e
xt
r
a
c
t
i
on:
di
ffe
r
e
nt
i
a
l
e
nt
ropy,
c
l
a
s
s
i
fi
c
a
t
i
on:
gra
ph
re
gu
l
a
ri
z
e
d
e
xt
r
e
m
e
l
e
a
rni
ng
m
a
c
hi
ne
[24]
69.
67
-
-
F
e
a
t
ure
e
xt
ra
c
t
i
on
:
pe
a
rs
on
c
orre
l
a
t
i
on
c
oe
ff
i
c
i
e
nt
,
c
l
a
s
s
i
fi
c
a
t
i
on:
c
onvol
u
t
i
on
ne
ur
a
l
ne
t
w
ork
[26]
73.
1
F
e
a
t
ure
e
xt
r
a
c
t
i
on:
pow
e
r
s
pe
c
t
ra
l
de
ns
i
t
y,
f
e
a
t
ur
e
re
pre
s
e
nt
a
t
i
on:
m
ul
t
i
di
m
e
ns
i
on
a
l
fe
a
t
ure
i
m
a
ge
,
c
l
a
s
s
i
fi
c
a
t
i
on
:
L
S
T
M
+
CN
N
[27]
75.
21
F
e
a
t
ure
e
xt
ra
c
t
i
on
a
nd
c
l
a
s
s
i
fi
c
a
t
i
on:
CN
N
,
b
a
s
e
l
i
ne
re
duc
t
i
on:
di
ffe
re
nc
e
.
F
e
a
t
ure
re
pre
s
e
nt
a
t
i
on:
3D
c
ube
[25]
93.
53
-
95.
95
S
m
oot
hi
ng:
M
W
M
F
,
e
xt
ra
c
t
i
on
fe
a
t
ur
e
:
di
ffe
re
nt
i
a
l
e
nt
ropy,
r
e
pre
s
e
nt
a
t
i
on
fe
a
t
ure
s
:
3D
c
ube
,
b
a
s
e
l
i
ne
re
duc
t
i
on:
re
l
a
t
i
v
e
di
ffe
re
nc
e
,
c
l
a
s
s
i
fi
c
a
t
i
on:
CN
N
[18]
97.
14
89.
71
99.
59
E
xt
ra
c
t
i
on
fe
a
t
ur
e
:
di
ffe
re
n
t
i
a
l
e
nt
ropy,
b
a
s
e
l
i
ne
re
duc
t
i
on:
di
ffe
r
e
nc
e
,
r
e
pre
s
e
nt
a
t
i
o
n
fe
a
t
ure
s
:
3D
c
ube
,
c
l
a
s
s
i
fi
c
a
t
i
on:
c
ont
i
nuous
c
a
ps
ul
e
ne
t
w
ork
[22]
91.
35
94.
23
96.
20
S
m
oot
hi
ng:
m
odi
fi
e
d
w
e
i
ght
e
d
m
e
a
n
fi
l
t
e
r
.
E
xt
ra
c
t
i
on
fe
a
t
ure
:
di
ffe
re
nt
i
a
l
e
nt
rop
y,
Ba
s
e
l
i
ne
re
du
c
t
i
on
:
re
l
a
t
i
ve
di
ff
e
re
nc
e
,
f
e
a
t
ure
re
pre
s
e
nt
a
t
i
on
:
3D
c
ube
,
c
l
a
s
s
i
fi
c
a
t
i
o
n:
c
ont
i
nuous
c
a
ps
ul
e
ne
t
w
ork
(
p
urpos
e
d
m
ode
l
)
99.
91
±
0
.
32
96.
98
±
1
.
53
99.
94
±
1
.
10
A
lt
hough
s
e
ve
r
a
l
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
m
e
th
ods
a
nd
c
la
s
s
if
ic
a
ti
on
m
e
th
ods
ba
s
e
d
on
m
a
c
hi
ne
le
a
r
ni
ng
a
nd
de
e
p
le
a
r
ni
ng
ha
v
e
be
e
n
a
ppl
ie
d
in
pr
e
vi
ous
s
tu
di
e
s
,
th
e
a
c
c
ur
a
c
y
of
e
m
ot
io
n
r
e
c
ogni
ti
on
ha
s
not
r
e
a
c
he
d
opt
im
a
l
(
be
lo
w
85%
)
[
23]
,
[
24]
,
[
26]
,
[
27]
.
T
he
va
r
ia
ti
on
of
th
e
tr
ia
l
s
ig
na
l
pa
tt
e
r
ns
p
r
oduc
e
d
c
a
us
e
s
th
is
pr
obl
e
m
.
S
o,
th
e
c
la
s
s
if
ic
a
ti
on
m
e
th
od
c
a
nnot
r
e
c
ogni
z
e
th
e
pa
tt
e
r
n
f
or
m
e
d.
T
o
ove
r
c
om
e
th
is
pr
obl
e
m
,
Z
ha
o
e
t
al
.
[
2
5]
us
e
d
th
e
di
f
f
e
r
e
nc
e
m
e
th
o
d
t
o
s
tu
dy
th
e
b
a
s
e
li
n
e
r
e
du
c
t
io
n
a
ppr
oa
c
h
.
T
h
e
r
e
d
uc
e
d
tr
i
a
l
s
i
gn
a
l
f
e
a
tu
r
e
v
a
l
ue
s
c
a
n pr
o
du
c
e
m
or
e
r
e
p
r
e
s
e
nt
a
t
iv
e
s
ig
na
l
p
a
t
te
r
n
s
s
o
th
a
t
th
e
E
E
G
s
ig
na
l
p
a
t
te
r
n c
a
n
b
e
r
e
c
og
ni
z
e
d
be
t
te
r
.
H
o
w
e
ve
r
,
of
s
e
v
e
r
a
l
ba
s
e
li
ne
r
e
du
c
t
io
n
m
e
th
od
s
,
t
he
r
e
l
a
ti
ve
di
f
f
e
r
e
n
c
e
m
e
t
ho
d
i
s
s
up
e
r
i
or
t
o
th
e
di
f
f
e
r
e
nc
e
m
e
t
hod
[
8]
.
W
ir
a
w
a
n
e
t
al
.
[
18
]
h
a
s
s
uc
c
e
s
s
f
ul
l
y
a
p
pl
i
e
d
t
he
r
e
l
a
t
iv
e
di
f
f
e
r
e
n
c
e
m
e
th
od
o
pt
i
m
i
z
e
d
w
it
h
th
e
m
o
di
f
i
e
d
w
e
ig
ht
e
d
m
e
a
n
f
il
te
r
m
e
th
od
.
H
ow
e
v
e
r
,
th
e
C
N
N
m
e
th
o
d
u
s
e
d
in
th
is
s
tu
dy
c
a
n
not
c
on
s
i
de
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
3
,
J
une
2025
:
2302
-
2314
2310
th
e
c
ha
r
a
c
t
e
r
i
s
ti
c
s
of
t
he
E
E
G
s
ig
na
l
.
W
ir
a
w
a
n
e
t
al
.
[
22
]
u
s
e
of
t
he
c
ont
in
u
ou
s
c
a
p
s
u
le
n
e
tw
or
k
m
e
th
od
ha
s
be
e
n
pr
o
ve
n
t
o
c
on
s
id
e
r
t
he
c
ha
r
a
c
t
e
r
i
s
ti
c
s
of
t
he
E
E
G
s
i
gn
a
l
.
H
ow
e
v
e
r
,
W
ir
a
w
a
n
e
t
al
.
[
22]
di
d
not
a
ppl
y
th
e
r
e
la
ti
ve
di
f
f
e
r
e
nc
e
a
nd
m
odi
f
ie
d
w
e
ig
ht
e
d
m
e
a
n
f
il
te
r
m
e
th
o
ds
.
B
a
s
e
d
on
th
is
s
tu
dy,
th
e
pr
opos
e
d
m
ode
l
c
om
bi
ne
s
th
e
m
od
e
ls
f
r
om
[
18]
,
[
22]
to
im
pr
ove
e
m
ot
io
n
r
e
c
ogni
ti
on
a
c
c
ur
a
c
y.
B
a
s
e
d
on
e
xpe
r
im
e
nt
s
c
onduc
te
d
on
th
r
e
e
publ
ic
da
ta
s
e
ts
,
th
e
pr
opos
e
d
m
ode
l
in
th
is
s
tu
dy
yi
e
ld
s
hi
ghe
r
a
c
c
ur
a
c
y
th
a
n
th
e
m
ode
ls
in
pr
e
vi
ous
s
tu
di
e
s
in
r
e
c
ogni
z
in
g
f
our
c
la
s
s
e
s
of
e
m
ot
io
ns
.
T
he
r
e
f
or
e
,
f
ur
th
e
r
a
ppl
ic
a
ti
on
r
e
s
e
a
r
c
h
c
a
n
u
s
e
th
is
m
ode
l
to
e
va
lu
a
te
s
tu
de
nt
s
'
unde
r
s
t
a
ndi
ng
in
vol
vi
ng
e
m
ot
io
na
l
r
e
a
c
ti
ons
f
r
om
s
tu
de
nt
s
dur
in
g
th
e
le
a
r
ni
ng
pr
oc
e
s
s
[
34]
,
[
35]
.
I
n
a
ddi
ti
on
to
e
m
ot
io
n
r
e
c
ogni
ti
on,
th
is
m
ode
l
c
a
n
a
ls
o
be
us
e
d
in
di
f
f
e
r
e
nt
dom
a
in
s
, s
uc
h a
s
de
te
c
ti
ng e
pi
le
p
s
y
[
36]
, [
37]
, a
nd mot
or
i
m
a
g
e
r
y
[
38]
.
H
ow
e
ve
r
,
ba
s
e
d
on
th
e
t
e
s
ts
c
onduc
te
d
on
th
e
D
R
E
A
M
E
R
da
ta
s
e
t,
th
e
m
ode
l
pr
opos
e
d
in
th
is
s
tu
dy
pr
oduc
e
s
s
li
ght
ly
lo
w
e
r
a
c
c
ur
a
c
y
va
lu
e
s
th
a
n
th
e
ot
he
r
two
s
e
c
onda
r
y
da
ta
s
e
ts
,
s
uc
h a
s
D
E
A
P
a
nd A
M
I
G
O
S
.
T
hi
s
pr
obl
e
m
is
c
a
u
s
e
d
by
th
e
a
m
pl
it
ude
va
lu
e
of
th
e
E
E
G
s
ig
na
l
in
th
e
D
R
E
A
M
E
R
da
ta
s
e
t
be
in
g
a
bove
a
ve
r
a
ge
.
F
ig
ur
e
s
9
a
nd
10
s
how
vi
s
u
a
li
z
a
ti
ons
of
th
e
ba
s
e
li
ne
s
ig
na
ls
f
r
om
s
e
ve
r
a
l
pa
r
ti
c
ip
a
nt
s
on
D
E
A
P
a
nd
A
M
I
G
O
S
da
ta
s
e
ts
.
F
ig
ur
e
9. T
he
s
m
oot
hi
ng pr
oc
e
s
s
us
in
g t
he
M
W
M
F
m
e
th
od on t
he
ba
s
e
li
ne
s
ig
na
l
f
r
om
t
he
s
e
c
ond pa
r
ti
c
ip
a
nt
in
t
he
A
M
I
G
O
S
da
ta
s
e
t
F
ig
ur
e
10. T
he
s
m
oot
hi
ng pr
oc
e
s
s
us
in
g t
he
M
W
M
F
m
e
th
od on
t
he
ba
s
e
li
ne
s
ig
na
l
f
r
om
t
he
e
ig
ht
h
pa
r
ti
c
ip
a
nt
i
n t
he
D
E
A
P
da
ta
s
e
t
A
f
te
r
r
e
m
ovi
ng
it
s
a
r
te
f
a
c
ts
us
in
g
th
e
m
odi
f
ie
d
w
e
ig
ht
e
d
m
e
a
n
f
il
te
r
m
e
th
od,
th
e
a
m
pl
it
ude
of
th
e
ba
s
e
li
ne
s
ig
na
l
in
th
e
D
E
A
P
a
nd
A
M
I
G
O
S
da
ta
s
e
ts
r
a
nge
s
±150
µV.
I
n
c
ont
r
a
s
t,
th
e
a
m
pl
i
tu
de
of
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
H
y
br
id
m
e
th
od f
or
opt
imi
z
in
g
e
m
ot
io
n r
e
c
ogni
ti
on m
ode
l
s
on
…
(
I
M
ade
A
gus
W
ir
aw
an
)
2311
ba
s
e
li
ne
s
ig
n
a
l
in
th
e
D
R
E
A
M
E
R
da
ta
s
e
t
ha
s
a
r
a
nge
of
±1
,
60
0
µV.
F
ig
ur
e
11
pr
e
s
e
nt
s
a
vi
s
u
a
li
z
a
ti
on
of
th
e
ba
s
e
li
ne
s
ig
n
a
l
in
t
he
t
e
nt
h pa
r
ti
c
ip
a
nt
be
f
or
e
a
nd a
f
te
r
s
m
oot
hi
ng on the
D
R
E
A
M
E
R
da
ta
s
e
t.
H
ow
e
ve
r
, i
de
a
ll
y, t
he
E
E
G
w
a
ve
i
n pa
r
ti
c
ip
a
nt
s
i
n nor
m
a
l
c
ond
it
io
ns
i
s
50
-
100 µV
[
39]
,
[
40]
. B
a
s
e
d
on
th
e
e
xpe
r
im
e
nt
s
,
it
c
a
n
be
s
ta
te
d
th
a
t
th
e
a
m
pl
it
ude
of
th
e
ba
s
e
li
ne
s
ig
na
l
in
th
e
D
R
E
A
M
E
R
da
ta
s
e
t
s
ti
ll
c
ont
a
in
s
a
r
te
f
a
c
ts
,
e
ve
n
th
ough
th
e
m
odi
f
ie
d
w
e
ig
ht
e
d
m
e
a
n
f
i
lt
e
r
m
e
th
od
ha
s
be
e
n
a
ppl
ie
d.
T
h
e
s
e
a
r
te
f
a
c
ts
c
a
us
e
th
e
a
c
c
ur
a
c
y
of
e
m
ot
io
n
r
e
c
ogni
ti
on
in
th
e
D
R
E
A
M
E
R
da
ta
s
e
t
to
be
s
li
ght
ly
lo
w
e
r
th
a
n
in
th
e
D
E
A
P
a
nd A
M
I
G
O
S
da
ta
s
e
ts
.
F
ig
ur
e
11. T
he
s
m
oot
hi
ng pr
oc
e
s
s
us
in
g t
he
M
W
M
F
m
e
th
od on
t
he
ba
s
e
li
ne
s
ig
na
l
f
r
om
t
he
t
e
nt
h pa
r
ti
c
ip
a
nt
in
t
he
D
R
E
A
M
E
R
da
ta
s
e
t
4.
C
O
N
C
L
U
S
I
O
N
T
w
o c
r
it
ic
a
l
f
a
c
to
r
s
t
o c
ons
id
e
r
i
n e
m
ot
io
n
r
e
c
ogni
ti
on a
r
e
d
if
f
e
r
e
nc
e
s
i
n E
E
G
s
ig
na
l
pa
tt
e
r
ns
c
a
us
e
d
by
pa
r
ti
c
ip
a
nt
c
ha
r
a
c
te
r
is
ti
c
s
a
nd
E
E
G
s
ig
na
ls
ha
vi
ng
s
pa
ti
a
l
in
f
or
m
a
ti
on.
T
he
s
e
two
f
a
c
to
r
s
s
ig
ni
f
ic
a
nt
ly
a
f
f
e
c
t
th
e
r
e
s
ul
ti
ng
a
c
c
ur
a
c
y.
T
h
e
pr
opos
e
d
m
ode
l
in
th
is
s
t
udy
c
a
n
c
ons
id
e
r
th
e
s
e
f
a
c
to
r
s
.
T
hi
s
m
ode
l
c
ons
is
ts
of
th
e
m
odi
f
ie
d
w
e
ig
ht
e
d
m
e
a
n
f
il
te
r
m
e
th
od
f
or
th
e
E
E
G
ba
s
e
li
ne
s
ig
na
l
s
m
oot
hi
ng
pr
oc
e
s
s
,
th
e
di
f
f
e
r
e
nt
ia
l
e
nt
r
opy
m
e
th
od
f
o
r
th
e
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
pr
oc
e
s
s
,
th
e
r
e
la
ti
ve
di
f
f
e
r
e
nc
e
m
e
th
od
f
or
ba
s
e
li
ne
r
e
duc
ti
on,
th
e
3D
c
ub
e
m
e
th
od
f
or
f
e
a
tu
r
e
r
e
pr
e
s
e
nt
a
ti
on,
a
nd
t
he
c
ont
in
uous
c
a
ps
ul
e
ne
twor
k
m
e
th
od
f
or
th
e
c
la
s
s
if
ic
a
ti
on
pr
oc
e
s
s
.
A
ppl
yi
ng
th
e
m
odi
f
ie
d
w
e
ig
ht
e
d
m
e
a
n
f
il
te
r
a
nd
r
e
la
ti
ve
di
f
f
e
r
e
nc
e
m
e
th
ods
c
a
n
ove
r
c
om
e
di
f
f
e
r
e
nc
e
s
in
tr
ia
l
s
ig
na
l
pa
tt
e
r
ns
.
M
e
a
nw
hi
le
,
th
e
d
if
f
e
r
e
nt
ia
l
e
nt
r
opy,
3D
c
ube
s
,
a
nd
c
ont
in
uous
c
a
ps
ul
e
ne
twor
k
m
e
th
ods
c
a
n
c
ons
id
e
r
s
pa
ti
a
l
in
f
or
m
a
ti
on
f
r
o
m
th
e
tr
ia
l
s
ig
na
l.
B
a
s
e
d
on
te
s
ti
ng
on
th
r
e
e
publ
ic
da
ta
s
e
ts
,
th
e
s
e
m
e
th
ods
c
a
n
im
pr
ove
th
e
e
m
ot
io
n
r
e
c
ogni
ti
on
m
ode
l
th
a
n
th
e
m
ode
l
pr
opos
e
d
in
pr
e
vi
ous
s
tu
di
e
s
.
H
ow
e
v
e
r
,
ba
s
e
d
on
e
xp
e
r
im
e
nt
s
c
onduc
te
d
o
n
th
r
e
e
publ
ic
da
ta
s
e
ts
,
it
c
a
n
b
e
s
ta
te
d
th
a
t
th
e
a
m
pl
it
ude
of
th
e
ba
s
e
li
ne
s
ig
na
l
in
th
e
D
R
E
A
M
E
R
da
ta
s
e
t
s
ti
l
l
c
ont
a
in
s
a
r
te
f
a
c
ts
,
e
ve
n
th
ough
th
e
m
odi
f
ie
d
w
e
ig
ht
e
d
m
e
a
n
f
il
te
r
m
e
th
od
ha
s
be
e
n
a
ppl
ie
d.
T
hi
s
a
r
te
f
a
c
t
c
a
us
e
s
th
e
a
c
c
ur
a
c
y
of
e
m
ot
io
n
r
e
c
ogni
ti
on
in
th
e
D
R
E
A
M
E
R
da
ta
s
e
t
to
be
s
li
ght
ly
lo
w
e
r
th
a
n
in
th
e
D
E
A
P
a
nd
A
M
I
G
O
S
da
ta
s
e
ts
.
T
he
r
e
f
or
e
,
th
e
f
ut
ur
e
c
ha
ll
e
nge
is
to
s
tu
dy
th
e
a
ppr
opr
ia
te
w
e
ig
ht
va
lu
e
in
th
e
m
odi
f
ie
d
w
e
ig
ht
e
d
m
e
a
n
f
il
te
r
m
e
th
od
f
or
th
e
a
m
pl
it
ude
of
th
e
ba
s
e
li
ne
E
E
G
s
ig
na
l
a
bove
s
ta
nda
r
d
a
nd
to
s
tu
dy
th
e
te
c
hni
que
of
r
e
m
ovi
ng
a
r
te
f
a
c
ts
f
o
r
e
xpe
r
im
e
nt
a
l
E
E
G
s
ig
na
ls
.
F
U
N
D
I
N
G
I
N
F
O
R
M
A
T
I
O
N
T
hi
s
w
or
k
w
a
s
s
uppor
te
d
e
nt
ir
e
ly
by
F
undi
ng
f
or
F
is
c
a
l
Y
e
a
r
2024,
N
o:
374/
U
N
48.16/L
T
/2
024,
f
r
om
th
e
D
i
r
e
c
to
r
a
te
of
R
e
s
e
a
r
c
h,
T
e
c
hnol
ogy,
a
nd
C
om
m
u
ni
ty
S
e
r
vi
c
e
,
M
in
is
tr
y
of
E
duc
a
ti
on,
C
ul
tu
r
e
,
R
e
s
e
a
r
c
h,
a
nd T
e
c
hnol
ogy, I
ndone
s
ia
.
A
U
T
H
O
R
C
O
N
T
R
I
B
U
T
I
O
N
S
S
T
A
T
E
M
E
N
T
T
hi
s
jo
ur
na
l
us
e
s
th
e
C
ont
r
ib
ut
or
R
ol
e
s
T
a
xonomy
(
C
R
e
di
T
)
to
r
e
c
ogni
z
e
in
di
vi
dua
l
a
ut
hor
c
ont
r
ib
ut
io
ns
, r
e
duc
e
a
ut
hor
s
hi
p di
s
put
e
s
,
a
nd f
a
c
il
it
a
te
c
ol
la
bo
r
a
ti
on.
Evaluation Warning : The document was created with Spire.PDF for Python.