I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l En
g
ineering
a
nd
Co
m
pu
t
er
Science
Vo
l.
38
,
No
.
1
,
A
p
r
il
20
25
,
p
p
.
1
~
21
I
SS
N:
2
502
-
4
7
52
,
DOI
: 1
0
.
1
1
5
9
1
/ijee
cs
.v
38.
i
1
.
p
p
1
-
21
1
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs
.
ia
esco
r
e.
co
m
A review
of conv
o
lutiona
l neural n
e
tworks
f
o
r
clas
sif
y
ing
power
qua
lity pro
blems
using
K
eras
API
Ada
m
u Sa
’
idu
1,
2
,
Na
s
iru B.
K
a
da
nd
a
ni
1
,
B
a
la
B
o
y
i Bu
k
a
t
a
1
1
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
i
c
a
l
En
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
E
n
g
i
n
e
e
r
i
n
g
,
B
a
y
e
r
o
U
n
i
v
e
r
si
t
y
,
K
a
n
o
,
N
i
g
e
r
i
a
2
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
i
c
a
l
/
El
e
c
t
r
o
n
i
c
s
Te
c
h
n
o
l
o
g
y
,
S
c
h
o
o
l
o
f
S
e
c
o
n
d
a
r
y
E
d
u
c
a
t
i
o
n
(
T
e
c
h
n
i
c
a
l
)
,
F
e
d
e
r
a
l
C
o
l
l
e
g
e
o
f
Ed
u
c
a
t
i
o
n
(
T
e
c
h
n
i
c
a
l
)
B
i
c
h
i
,
K
a
n
o
S
t
a
t
e
,
N
i
g
e
r
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l
4
,
2
0
2
4
R
ev
is
ed
No
v
3
,
2
0
2
4
Acc
ep
ted
No
v
23
,
2
0
2
4
Th
e
m
a
jo
r
c
a
u
se
s
o
f
e
lec
tri
c
p
o
w
e
r
q
u
a
li
t
y
(
P
Q)
p
ro
b
lem
s
a
re
m
a
in
ly
d
u
e
to
th
e
i
n
c
re
a
se
d
u
ti
li
z
a
ti
o
n
o
f
n
o
n
l
in
e
a
r
lo
a
d
s
,
c
a
p
a
c
it
o
r
a
n
d
l
o
a
d
sw
it
c
h
in
g
e
v
e
n
ts,
tran
sfo
rm
e
r
e
n
e
r
g
iza
ti
o
n
,
a
n
d
o
c
c
u
rre
n
c
e
o
f
a
ss
o
rted
fa
u
lt
s
a
t
th
e
d
istri
b
u
ti
o
n
c
o
rrid
o
r.
Th
e
p
ro
b
l
e
m
s
o
ften
in
tr
o
d
u
c
e
h
a
rm
o
n
ics
a
n
d
o
th
e
r
wa
v
e
fo
rm
a
n
o
m
a
li
e
s
li
k
e
v
o
lt
a
g
e
sa
g
s,
v
o
lt
a
g
e
sw
e
ll
s
a
n
d
in
t
e
rru
p
ti
o
n
s
a
lo
n
g
t
h
e
p
o
we
r
s
y
ste
m
s.
A
ti
m
e
ly
c
las
sifica
ti
o
n
o
f
su
c
h
p
ro
b
lem
s
is
im
p
o
rtan
t
i
n
u
n
d
e
rsta
n
d
i
n
g
th
e
ir
imp
a
c
t
o
n
c
o
stly
p
o
we
r
sy
ste
m
e
c
o
n
o
m
y
.
Th
e
p
a
p
e
r
e
x
p
lo
re
s
c
o
m
p
re
h
e
n
siv
e
re
v
iew
o
f
P
Q
issu
e
s,
o
p
e
ra
ti
o
n
a
l
c
o
n
c
e
p
t
o
f
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
two
r
k
(CNN
)
a
n
d
it
s
u
ti
li
z
a
ti
o
n
in
s
o
lv
in
g
P
Q
p
ro
b
lem
s.
No
v
e
l
d
e
e
p
lea
rn
in
g
(DL)
a
p
p
r
o
a
c
h
u
si
n
g
v
a
rian
t
o
f
De
n
se
Ne
t
CNN
tec
h
n
i
q
u
e
in
Ke
ra
s
API
p
lat
fo
rm
is
d
e
p
l
o
y
e
d
t
o
e
x
trac
t
t
h
e
fe
a
tu
re
s
o
f,
a
n
d
c
las
sify
P
Q
p
r
o
b
lem
s.
Th
e
p
ro
p
o
se
d
tec
h
n
i
q
u
e
imp
r
o
v
e
s
c
la
ss
ifi
c
a
ti
o
n
p
e
rfo
rm
a
n
c
e
with
a
n
a
c
c
u
ra
c
y
o
f
9
9
.
9
6
%
.
It
s
h
o
ws
re
m
a
rk
a
b
le
imp
ro
v
e
m
e
n
t
o
v
e
r
th
e
trad
it
i
o
n
a
l
tec
h
n
iq
u
e
s
in
th
e
li
tera
tu
re
w
h
ich
we
re
7
3
.
5
3
%
t
o
9
9
.
9
2
%
a
c
c
u
ra
te
fo
r
a
p
e
rio
d
fro
m
2
0
1
8
to
2
0
2
3
.
Th
e
m
o
st
p
r
o
m
is
in
g
p
a
rt
o
f
th
e
m
e
th
o
d
is
t
h
e
imp
r
o
v
e
m
e
n
t
sh
o
w
n
i
n
th
e
c
las
sifica
ti
o
n
p
e
rfo
rm
a
n
c
e
wh
e
n
c
o
m
p
a
re
d
with
th
a
t
o
b
tain
e
d
in
th
e
li
tera
tu
re
.
Th
e
tec
h
n
i
q
u
e
c
a
n
a
lso
b
e
a
p
p
li
e
d
i
n
re
a
l
ti
m
e
to
c
a
ter fo
r
re
a
l
P
Q p
ro
b
lem
s.
K
ey
w
o
r
d
s
:
C
o
n
v
o
lu
tio
n
al
n
eu
r
al
Dee
p
lear
n
in
g
Dis
tr
ib
u
tio
n
co
r
r
id
o
r
Ker
as API
Netwo
r
k
s
Po
wer
q
u
ality
p
r
o
b
lem
s
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Ad
am
u
Sa’
id
u
Dep
ar
tm
en
t o
f
E
lectr
ical
E
n
g
i
n
ee
r
in
g
,
Facu
lty
o
f
E
n
g
in
ee
r
in
g
,
B
ay
er
o
Un
iv
e
r
s
ity
P.
M.
B
.
3
0
1
1
,
Kan
o
,
Nig
e
r
ia
E
m
ail: a
d
am
u
g
iwa7
7
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
Po
wer
q
u
ality
(
PQ)
o
f
an
elec
tr
ical
p
o
wer
s
y
s
tem
n
etwo
r
k
is
af
f
ec
ted
with
n
u
m
b
er
o
f
is
s
u
es
[
1
]
th
at
m
ay
ca
u
s
e
m
alf
u
n
ctio
n
in
g
,
d
am
ag
e
o
f
a
p
p
lian
ce
s
an
d
e
v
e
n
f
u
n
d
lo
s
s
es.
Acc
o
r
d
in
g
to
t
h
e
I
E
E
E
STD
5
1
9
-
1
9
9
2
d
o
cu
m
e
n
t,
PQ
h
as
b
ee
n
id
en
tifie
d
o
n
th
e
id
ea
o
f
ea
r
th
in
g
elec
tr
o
n
ic
co
m
p
o
n
en
t
s
in
a
way
th
at
i
s
ap
p
r
o
p
r
iate
to
th
e
p
er
f
o
r
m
an
c
e
o
f
th
at
eq
u
ip
m
en
t
an
d
ag
r
e
ea
b
le
with
th
e
p
r
em
is
e
wir
in
g
s
y
s
tem
an
d
o
th
er
co
n
n
ec
ted
d
ev
ices
[
2
]
.
Oth
er
PQ
d
ef
in
itio
n
s
h
av
e
also
b
ee
n
ad
v
an
ce
d
,
s
ee
f
o
r
ex
am
p
le
[3
]
-
[
6
]
.
T
h
e
co
n
ce
p
t
o
f
g
o
o
d
o
r
b
a
d
PQ
d
ep
en
d
s
o
n
th
e
cu
s
to
m
er
’
s
p
er
ce
p
t
io
n
.
C
o
n
v
er
s
ely
,
if
th
e
eq
u
i
p
m
en
t
o
p
er
atio
n
is
s
atis
f
ac
to
r
ily
,
th
e
cu
s
to
m
er
l
ab
els
th
e
p
o
wer
as
th
at
o
f
g
o
o
d
q
u
ality
,
ot
h
er
wis
e
it
is
o
f
b
ad
q
u
ality
[
7
]
.
B
etwe
en
th
e
two
ex
tr
em
es,
s
ev
er
al
g
r
ad
es
o
f
PQ
m
ay
ex
is
t,
ag
ain
d
ep
e
n
d
in
g
o
n
th
e
c
u
s
to
m
er
’
s
p
er
ce
p
tio
n
.
T
h
ese
g
r
ad
es
m
ay
b
e
as
a
r
esu
lt
o
f
th
e
escalatio
n
o
f
n
o
n
lin
ea
r
d
e
v
ices
with
in
th
e
p
o
wer
s
y
s
tem
,
u
n
p
r
ec
e
d
en
ted
f
au
lts
,
as
well
as
p
o
wer
o
u
tag
es.
As
a
r
esu
lt,
m
o
s
t
elec
tr
o
n
ics
eq
u
ip
m
en
t
an
d
o
th
er
s
en
s
itiv
e
d
ev
ices
at
th
e
cu
s
to
m
er
’
s
lo
ad
s
ten
d
to
eith
er
s
h
u
t
d
o
wn
o
r
g
et
co
m
p
letely
d
a
m
ag
ed
,
th
er
eb
y
ca
u
s
in
g
s
er
io
u
s
d
is
s
atis
f
ac
tio
n
to
th
e
cu
s
to
m
er
.
T
h
e
PQ
d
is
tu
r
b
an
ce
s
af
f
ec
t
b
o
th
th
e
cu
s
to
m
e
r
as
well
as
th
e
u
tili
ty
b
y
,
f
o
r
ex
am
p
le,
ca
u
s
in
g
tr
an
s
m
is
s
io
n
lin
e
lo
s
s
es
an
d
s
h
u
ttin
g
d
o
wn
o
f
g
e
n
er
ato
r
s
.
U
n
d
er
s
tan
d
in
g
an
d
tack
lin
g
PQ
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
38
,
No
.
1
,
Ap
r
il
20
25
:
1
-
21
2
ev
en
ts
h
as
h
e
n
ce
b
ec
o
m
e
p
a
r
am
o
u
n
t
in
im
p
r
o
v
in
g
s
er
v
ic
e
q
u
ality
at
cu
s
to
m
e
r
s
an
d
u
tili
ty
’
s
d
is
tr
ib
u
tio
n
c
o
r
r
id
o
r
s
.
I
t
h
as
b
ee
n
r
ep
o
r
te
d
in
[
8
]
th
at
n
o
n
lin
ea
r
r
elate
d
PQ
p
r
o
b
lem
s
d
ir
ec
tly
af
f
ec
t
s
en
s
itiv
e
d
is
tr
ib
u
tio
n
s
y
s
tem
eq
u
ip
m
en
t
lik
e
elec
tr
ic
d
r
iv
es,
p
r
o
g
r
am
m
a
b
le
lo
g
ic
c
o
n
tr
o
ller
s
an
d
s
in
g
le
-
l
o
o
p
c
o
n
tr
o
ller
s
.
T
h
er
ef
o
r
e,
an
en
v
is
ag
ed
tim
ely
id
en
tific
a
tio
n
an
d
class
if
icatio
n
o
f
s
u
ch
p
r
o
b
lem
s
is
r
ec
o
n
s
id
er
ed
in
t
h
is
p
ap
er
.
I
n
v
iew
o
f
th
is
,
co
n
v
en
tio
n
al
way
s
in
d
ea
lin
g
with
PQ
is
s
u
es
in
v
o
lv
e
f
ea
tu
r
e
e
x
tr
ac
tio
n
m
eth
o
d
as
well
as
clas
s
if
icatio
n
p
r
o
ce
s
s
u
s
in
g
s
ig
n
al
s
y
n
th
esis
an
aly
s
is
.
Su
ch
tech
n
iq
u
es
em
p
lo
y
ed
f
o
r
f
ea
tu
r
e
s
y
n
th
esis
in
liter
atu
r
e
in
clu
d
e
f
o
u
r
ier
tr
a
n
s
f
o
r
m
(
FT)
,
s
h
o
r
t
tim
e
f
o
u
r
ie
r
tr
an
s
f
o
r
m
(
STFT
)
,
wav
elet
t
r
an
s
f
o
r
m
(
W
T
)
,
S
-
tr
an
s
f
o
r
m
(
ST)
,
wig
n
er
v
ille
d
is
tr
ib
u
tio
n
(
W
VD)
,
em
p
ir
i
ca
l
m
o
d
e
d
ec
o
m
p
o
s
itio
n
(
E
MD
)
,
in
d
ep
e
n
d
en
t
co
m
p
o
n
en
t
an
aly
s
is
(
I
C
A)
an
d
v
ar
iatio
n
al
m
o
d
e
d
ec
o
m
p
o
s
itio
n
(
VM
D)
s
ee
f
o
r
in
s
tan
t
[
9
]
.
Mo
s
t
o
f
th
ese
m
eth
o
d
s
ar
e
ass
o
ciate
d
with
d
r
awb
ac
k
s
f
o
r
ex
am
p
le;
FT
is
s
im
p
le
b
u
t
u
n
s
u
itab
le
f
o
r
ch
ao
tic
s
y
s
tem
s
wi
th
v
ar
iab
le
d
is
tu
r
b
an
ce
s
lar
g
el
y
b
ec
au
s
e
o
f
its
r
is
k
in
tim
e
-
f
r
e
q
u
en
cy
r
ep
r
esen
tat
io
n
a
b
ilit
y
.
Ho
wev
er
,
STFT
ca
n
u
s
e
a
s
lid
in
g
win
d
o
w
to
s
et
tle
FT’
s
tim
e
-
f
r
eq
u
en
cy
lo
ca
lizatio
n
p
r
o
b
lem
in
ch
a
o
tic
s
y
s
tem
s
.
B
u
t
it
i
s
r
estricte
d
with
th
e
d
im
en
s
io
n
o
f
th
e
s
lid
in
g
win
d
o
w
to
b
e
f
u
r
th
er
d
ep
lo
y
ed
.
W
T
ca
n
im
p
r
o
v
e
t
h
e
tim
e
-
f
r
eq
u
e
n
cy
r
eso
lu
tio
n
i
n
PQ
d
is
tu
r
b
an
ce
an
aly
s
is
h
o
wev
er
it
is
r
ea
ctiv
e
to
n
o
is
e.
ST
co
m
b
in
es
STFT
with
W
T
an
d
s
u
b
d
u
es
s
o
m
e
o
f
t
h
e
W
T
d
r
awb
ac
k
s
.
Hen
ce
,
th
e
ST
is
p
r
ed
o
m
in
an
t
to
FT,
STFT
a
n
d
W
T
ap
p
r
o
ac
h
es
s
p
ec
if
ically
in
th
e
n
o
is
e
-
r
ich
s
y
s
tem
s
.
E
v
en
th
o
u
g
h
,
th
e
g
e
n
er
ic
ad
ap
tatio
n
o
f
ST
r
estricte
d
its
ap
p
licatio
n
b
ec
au
s
e
o
f
its
co
m
p
lex
ities
.
R
ec
en
tly
,
W
VD
m
eth
o
d
h
as
g
ain
ed
m
o
r
e
atten
tio
n
as
o
n
e
o
f
th
e
an
aly
tical
d
ev
ices
f
o
r
n
o
n
s
tatio
n
ar
y
s
ig
n
als
d
u
e
to
its
h
ig
h
tim
e
-
f
r
e
q
u
en
cy
r
eso
lu
tio
n
an
d
h
ig
h
p
er
f
o
r
m
an
ce
in
th
e
ex
is
ten
ce
o
f
n
o
is
e
ag
ain
[
9
]
.
Ho
wev
er
,
th
is
m
eth
o
d
r
eq
u
ir
e
s
m
o
r
e
tim
e
to
tr
an
s
f
er
1
D
to
2
D
im
ag
e
f
ile
wh
ich
ca
u
s
es
d
elay
s
in
th
e
tr
ain
i
n
g
p
r
o
ce
s
s
.
T
h
e
E
MD
b
ein
g
a
d
a
p
tiv
e
in
n
at
u
r
e
is
a
tim
e
-
f
r
eq
u
en
cy
a
p
p
r
o
ac
h
f
o
r
s
y
n
th
esizin
g
n
o
n
s
tatio
n
ar
y
s
ig
n
als
wh
ich
d
is
in
teg
r
ate
s
ig
n
al
in
to
a
f
in
ite
n
u
m
b
er
o
f
i
n
tr
in
s
ic
m
o
d
e
f
u
n
ctio
n
s
(
I
MFs)
[
1
0
]
.
Als
o
,
it
h
as
s
o
m
e
in
h
er
en
t p
r
o
b
lem
s
lik
e
m
o
d
e
m
ix
in
g
an
d
b
o
u
n
d
ar
y
a
f
f
ec
ts
th
at
led
to
im
p
r
o
p
e
r
I
MF
d
ec
o
m
p
o
s
itio
n
a
b
ilit
y
.
E
MD
-
I
C
A
tech
n
iq
u
e
was
in
tr
o
d
u
ce
d
to
a
d
d
r
ess
th
ese
is
s
u
es
.
T
h
e
tech
n
i
q
u
e
is
ef
f
icien
t
in
r
em
o
v
in
g
th
e
m
o
d
e
m
ix
i
n
g
e
f
f
ec
t
b
u
t
f
ac
es
th
e
lack
o
f
th
e
am
p
litu
d
e
in
f
o
r
m
atio
n
d
u
e
t
o
th
e
in
clu
d
ed
I
C
A.
On
th
e
o
th
er
h
an
d
,
t
h
e
VM
D
d
is
m
an
tles
a
m
u
ltimo
d
al
s
ig
n
al
to
f
in
ite
n
u
m
b
er
o
f
b
an
d
-
lim
ited
I
MFs.
W
h
en
co
m
p
ar
in
g
E
MD
b
ased
ap
p
r
o
ac
h
es,
th
e
VM
D
is
a
m
o
r
e
s
tr
o
n
g
an
d
r
eliab
le
tech
n
iq
u
e
to
n
o
is
e
as
well
as
s
am
p
lin
g
er
r
o
r
s
wh
ic
h
g
en
e
r
alize
a
class
ical
W
ien
er
f
ilter
in
to
m
u
ltip
les
an
d
a
d
ap
tiv
e
b
a
n
d
s
.
T
h
e
s
ig
n
al
p
r
o
ce
s
s
in
g
ap
p
r
o
ac
h
es
d
is
cu
s
s
ed
s
o
f
ar
ar
e
alwa
y
s
e
m
p
lo
y
ed
to
e
x
tr
ac
t
f
ea
tu
r
es
f
r
o
m
m
an
y
ty
p
es
o
f
PQ p
r
o
b
lem
s
.
Fu
r
th
er
m
o
r
e
,
as
s
o
o
n
as
f
ea
t
u
r
e
is
ex
tr
ac
ted
,
PQ
p
r
o
b
lem
s
cl
ass
if
icatio
n
p
r
o
ce
s
s
s
tar
ts
u
s
in
g
s
p
ec
if
ic
tech
n
iq
u
e.
Fo
r
in
s
tan
ce
,
in
C
ai
et
a
l.
[
9
]
,
PQ
d
is
tu
r
b
an
ce
s
class
if
icatio
n
tech
n
iq
u
es
in
cl
u
d
e
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
es
(
SVM)
,
ar
tific
ial
n
eu
r
al
n
etwo
r
k
(
ANN)
,
p
r
o
b
a
b
ilis
tic
n
eu
r
al
n
etwo
r
k
s
(
PN
N)
an
d
d
ec
is
io
n
t
r
ee
(
DT
)
.
Am
o
n
g
s
t
th
e
tech
n
i
q
u
es
m
en
tio
n
ab
o
v
e,
SVM
is
th
e
m
o
s
t
wid
ely
u
tili
ze
m
eth
o
d
wi
th
litt
le
s
am
p
les
an
d
s
tr
u
ctu
r
al
r
is
k
s
in
v
o
lv
ed
ar
e
r
e
d
u
ce
d
.
So
also
,
DT
is
d
ec
is
io
n
m
ak
in
g
tech
n
iq
u
e
i
n
tr
ee
-
lik
e
p
atter
n
g
r
a
p
h
u
s
ed
to
h
ig
h
lig
h
t
th
e
r
elatio
n
s
h
ip
o
f
v
ar
io
u
s
f
ea
tu
r
es
th
at
m
ak
es
ca
teg
o
r
izatio
n
o
f
PQ
p
r
o
b
lem
s
ea
s
y
[
1
1
]
.
Ho
wev
er
,
SVM
an
d
DT
tech
n
iq
u
es
g
av
e
ac
cu
m
u
lativ
e
er
r
o
r
s
in
th
e
class
if
icatio
n
p
r
o
ce
s
s
o
f
PQ
p
r
o
b
lem
s
.
T
ac
k
lin
g
th
is
is
s
u
e,
ANN
-
b
ased
class
if
ier
s
ar
e
g
en
er
ally
ap
p
lied
v
ia
ef
f
icien
t
lear
n
in
g
p
r
o
ce
s
s
.
T
h
e
ap
p
r
o
ac
h
elim
in
ates
th
e
p
r
esen
ce
o
f
iter
atio
n
s
o
r
ac
cu
m
u
lativ
e
er
r
o
r
s
.
T
h
e
PNN
was
o
b
tain
ed
f
r
o
m
B
ay
esian
n
etwo
r
k
an
d
k
er
n
el
f
is
h
er
d
is
cr
im
in
an
t
alg
o
r
ith
m
.
T
h
e
PNN
was
ac
c
ep
ted
to
b
e
q
u
ick
er
an
d
m
o
r
e
r
eliab
le
th
an
ANN
an
d
its
s
tay
aliv
e
m
eth
o
d
s
co
m
p
r
is
in
g
o
f
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
class
if
icatio
n
tec
h
n
iq
u
es
h
ad
b
ee
n
d
em
o
n
s
tr
ated
to
b
e
f
r
u
itf
u
l
[
1
2
]
.
Yet,
it
h
as
th
r
ee
d
r
awb
ac
k
s
f
ir
s
tly
;
f
ea
tu
r
e
ex
tr
ac
tio
n
m
eth
o
d
s
ar
e
n
o
t
au
to
m
atic,
b
ec
au
s
e
v
ar
io
u
s
k
i
n
d
s
as
well
as
q
u
an
titi
es
o
f
f
ea
tu
r
es
h
av
e
v
ar
y
in
g
ef
f
ec
ts
o
n
th
e
class
if
icatio
n
p
er
f
o
r
m
an
ce
s
.
Hen
ce
,
ac
cu
r
a
cy
o
f
th
e
class
if
ier
i
s
o
f
ten
u
n
d
er
m
in
e
d
as
k
ey
f
ea
tu
r
es
co
u
ld
b
e
m
is
s
ed
o
u
t.
Seco
n
d
ly
,
f
ea
tu
r
e
ex
tr
ac
tio
n
t
ec
h
n
iq
u
e
an
d
class
if
icatio
n
s
t
ag
e
ar
e
two
in
d
iv
id
u
alis
tic
ac
tiv
ities
y
et
v
ar
ia
b
les
o
f
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
s
co
u
ld
b
e
en
h
a
n
ce
d
as
p
er
P
Q
s
i
g
n
als
an
aly
s
is
.
No
twith
s
tan
d
in
g
,
v
ar
ia
b
les
o
f
f
ea
tu
r
e
ex
tr
ac
tio
n
ar
e
s
ec
u
r
e
d
im
m
ed
iately
wh
en
th
e
o
p
er
a
tio
n
is
ac
co
m
p
lis
h
ed
an
d
lim
its
ac
cu
r
ac
y
o
f
th
e
class
if
icatio
n
r
esu
lts
.
Hen
ce
,
th
e
attr
ib
u
tes
o
f
PQ
p
r
o
b
lem
s
ca
n
n
o
t
b
e
r
ec
o
n
d
itio
n
ed
in
th
e
d
u
e
c
o
u
r
s
e.
T
h
ir
d
ly
,
co
n
v
en
tio
n
al
m
eth
o
d
s
ar
e
s
h
allo
w
(
litt
le
o
r
s
lig
h
t
d
e
p
th
)
lear
n
in
g
m
ec
h
a
n
is
m
in
n
atu
r
e
.
T
h
e
class
if
icatio
n
p
er
f
o
r
m
an
ce
s
ar
e
lo
wer
th
an
d
ee
p
lear
n
in
g
(
D
L
)
tech
n
iq
u
es
b
ec
au
s
e
th
e
latter
h
av
e
d
e
ep
lay
er
n
etwo
r
k
an
d
b
ig
d
ata
s
u
p
p
o
r
t [
1
3
]
.
L
ater
o
n
,
v
a
r
io
u
s
tech
n
iq
u
es
wer
e
d
ev
elo
p
ed
to
au
to
m
atica
lly
d
etec
t
an
d
class
if
ie
s
P
Q
p
r
o
b
lem
s
,
p
ar
ticu
lar
ly
b
ased
o
n
s
ig
n
al
p
r
o
ce
s
s
in
g
tech
n
iq
u
es
[
1
4
]
.
No
wad
ay
s
,
f
ew
ap
p
r
o
ac
h
es
h
av
e
b
ee
n
p
u
t
in
p
lace
to
co
m
e
u
p
with
a
u
to
m
atic
class
if
icatio
n
o
f
PQ
p
r
o
b
lem
s
u
tili
zin
g
v
o
lu
m
in
o
u
s
d
ata
i
n
v
o
lv
i
n
g
m
ac
h
in
e
lear
n
in
g
(
ML
)
tech
n
iq
u
es
d
ir
ec
tly
.
ML
is
a
g
en
er
al
ter
m
th
at
r
ef
er
s
to
alg
o
r
ith
m
s
wh
ich
lear
n
s
f
r
o
m
v
ast
am
o
u
n
t
o
f
d
ata.
R
ec
en
tly
,
ML
h
as
r
ec
ei
v
ed
m
u
ch
awa
r
en
ess
d
u
e
to
t
h
e
ev
o
lu
tio
n
o
f
m
o
r
e
p
r
o
m
is
in
g
alg
o
r
ith
m
s
,
m
o
r
e
tr
ain
in
g
d
ata
av
ailab
ilit
y
as
w
ell
as
m
o
r
e
c
o
m
p
u
tatio
n
al
r
es
o
u
r
ce
s
g
lo
b
ally
[
1
5
]
.
I
t
ca
n
al
s
o
b
e
ap
p
lied
f
o
r
a
wid
e
ar
ea
o
f
ap
p
licatio
n
s
s
u
c
h
as
cr
ed
it
-
ca
r
d
f
r
a
u
d
id
en
tifi
ca
tio
n
,
s
p
ee
ch
r
ec
o
g
n
itio
n
an
d
m
ed
ical
d
iag
n
o
s
is
.
DL
p
o
in
t
o
u
t
th
e
co
m
m
o
n
ty
p
e
o
f
ML
tech
n
iq
u
es
em
p
lo
y
ed
f
o
r
lear
n
in
g
d
is
cr
im
in
ativ
e
ch
ar
ac
ter
is
tics
f
r
o
m
a
g
iv
en
d
ata
in
c
h
r
o
n
o
lo
g
ical
o
r
d
er
u
s
in
g
ass
em
b
led
,
la
y
er
wis
e
s
tr
u
ctu
r
es
.
Am
o
n
g
s
t
wh
ich
ar
e
C
NN,
lo
n
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
A
r
ev
iew
o
f c
o
n
vo
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
ks fo
r
cla
s
s
i
fyin
g
p
o
w
er q
u
a
lity p
r
o
b
lems u
s
in
g
…
(
A
d
a
mu
S
a
’
id
u
)
3
s
h
o
r
t
-
ter
m
m
e
m
o
r
y
n
etwo
r
k
s
(
L
STM
s
)
,
co
n
v
o
lu
tio
n
al
a
u
to
en
co
d
er
s
(
C
AE
s
)
an
d
L
STM
au
to
en
co
d
er
s
[
1
6
]
.
DL
m
o
d
els
d
em
o
n
s
tr
ated
h
ig
h
an
ticip
ated
ca
p
ab
ilit
ies
in
im
ag
e
an
d
s
p
ee
ch
r
ec
o
g
n
itio
n
,
n
atu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
(
NL
P),
an
d
in
tellig
en
t
g
am
if
icatio
n
[
1
7
]
.
Alth
o
u
g
h
,
th
e
r
e
is
ex
is
ten
ce
o
f
s
ev
er
al
ex
ce
llen
t
r
e
v
iew
p
ap
er
s
in
th
e
f
ield
,
th
e
f
o
cu
s
is
n
o
t
o
n
tim
ely
f
ea
tu
r
e
ex
tr
a
ctio
n
p
r
ec
is
io
n
an
d
ac
cu
r
ate
c
lass
if
icatio
n
o
f
P
Q
p
r
o
b
lem
s
.
Acc
o
r
d
in
g
to
th
e
a
f
o
r
em
e
n
tio
n
ed
is
s
u
es
r
elatin
g
to
DL
ca
p
ab
ilit
y
,
th
is
p
ap
er
is
aim
ed
at
r
ev
iewin
g
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
(
C
NN)
b
ased
Den
s
eNe
t
ar
ch
itectu
r
e
f
o
r
ti
m
ely
f
ea
tu
r
e
ex
tr
ac
tin
g
an
d
class
if
y
in
g
PQ
p
r
o
b
lem
s
as
o
n
e
o
f
th
e
co
n
tr
ib
u
tio
n
s
in
attem
p
ts
to
s
o
lv
e
PQ
p
r
o
b
lem
s
at
th
e
d
is
tr
ib
u
tio
n
co
r
n
er
.
I
n
th
is
p
ap
e
r
,
th
e
ap
p
licatio
n
o
f
DL
tech
n
iq
u
e
p
a
r
t
icu
lar
ly
C
NN
b
ased
K
er
as
A
PI
to
au
to
m
atica
lly
class
if
y
PQ
p
r
o
b
le
m
s
will
b
e
co
m
p
r
eh
e
n
s
iv
ely
d
em
o
n
s
tr
ated
.
E
v
en
th
o
u
g
h
,
th
e
p
r
o
p
o
s
ed
DL
m
o
d
el
h
as
th
e
ca
p
ab
ilit
y
o
f
ac
c
u
r
ately
class
if
y
in
g
f
i
v
e
d
if
f
er
en
t
PQ
p
r
o
b
le
m
s
is
also
ex
p
ec
ted
to
o
u
t
p
er
f
o
r
m
o
t
h
er
m
o
d
els
b
r
o
u
g
h
t
f
o
r
war
d
in
th
e
liter
atu
r
e.
Mo
d
el
v
alid
atio
n
is
also
p
e
r
f
o
r
m
e
d
s
o
as
to
a
u
th
en
ticate
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
d
ev
elo
p
e
d
ap
p
r
o
ac
h
.
T
h
e
s
o
u
n
d
co
n
tr
ib
u
tio
n
s
o
f
th
is
p
ap
er
ar
e:
(
1
)
T
h
e
co
m
p
r
e
h
en
s
iv
e
r
ev
iew
o
f
C
NN
f
o
r
PQ
p
r
o
b
lem
s
class
if
icatio
n
p
r
o
ce
s
s
es
(
2
)
I
n
d
is
p
ar
ity
to
th
e
cu
r
r
en
t
PQ
s
ig
n
als
an
aly
s
is
,
C
NN
-
Ker
as
m
o
d
el
is
em
p
lo
y
ed
f
o
r
t
h
e
PQ
p
r
o
b
le
m
s
class
if
icat
io
n
(
3
)
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
o
f
C
NN
b
ased
K
er
as
h
as
s
u
p
er
io
r
p
er
f
o
r
m
an
ce
s
in
im
a
g
e
ca
teg
o
r
izatio
n
(
4
)
T
h
e
tim
e
co
n
s
u
m
i
n
g
s
tag
es
in
f
ea
tu
r
e
e
x
tr
ac
tio
n
,
f
ea
tu
r
e
s
elec
tio
n
an
d
d
ata
s
ize
m
in
im
izatio
n
in
tr
ad
itio
n
al
ML
b
ased
alg
o
r
ith
m
s
ar
e
h
u
g
ely
r
e
d
u
ce
d
an
d
(
5
)
An
ad
a
p
tiv
e
m
o
m
en
t
esti
m
atio
n
(
Ad
am
)
o
p
tim
izatio
n
alg
o
r
ith
m
was
em
p
lo
y
ed
to
g
et
th
e
b
est
h
y
p
er
p
a
r
am
eter
s
f
o
r
tu
n
in
g
C
NN
-
Ker
as
m
o
d
el.
R
em
ain
in
g
p
ar
ts
o
f
th
e
p
ap
e
r
wer
e
o
r
g
an
ize
d
as
f
o
llo
ws.
Sectio
n
2
r
e
v
iews
PQ
p
r
o
b
lem
s
an
d
s
ec
tio
n
3
d
escr
ib
es
a
co
m
p
r
e
h
en
s
iv
e
o
v
e
r
v
iew
o
f
t
h
e
C
NN.
I
n
s
ec
tio
n
4
ap
p
licatio
n
s
o
f
C
NN
in
PQ
p
r
o
b
lem
s
class
if
icatio
n
wer
e
r
ev
iewe
d
.
Sectio
n
5
p
r
esen
ts
m
eth
o
d
o
lo
g
y
o
f
th
e
p
r
o
p
o
s
ed
tec
h
n
iq
u
e.
R
esu
lt
an
d
d
is
cu
s
s
io
n
ar
e
g
iv
en
in
s
ec
tio
n
6
,
an
d
s
ec
tio
n
7
h
ig
h
lig
h
ts
th
e
cu
r
r
en
t
c
h
allen
g
es
o
f
th
e
C
NN.
Fin
ally
,
s
o
m
e
co
n
clu
s
io
n
s
wer
e
d
r
aw
n
in
s
ec
tio
n
8
.
2.
P
O
WE
R
Q
UAL
I
T
Y
DE
F
I
N
E
D
T
h
eo
r
etica
lly
,
PQ
is
co
n
s
id
er
ed
to
b
e
a
m
u
ltifa
ce
ted
ele
ctr
o
m
ag
n
etic
p
h
en
o
m
e
n
o
n
t
h
at
d
is
tu
r
b
s
v
o
ltag
e
a
n
d
c
u
r
r
e
n
t
s
ig
n
al
f
r
o
m
id
ea
l
wav
ef
o
r
m
wh
ich
is
r
ef
er
r
ed
as
th
e
PQ
p
r
o
b
l
em
.
T
h
e
ter
m
PQ
en
co
m
p
ass
es
an
y
f
ac
et
r
elate
d
to
p
ea
k
,
an
g
le
an
d
f
r
e
q
u
en
cy
o
f
v
o
ltag
e
a
n
d
c
u
r
r
e
n
t
wav
esh
ap
es
liv
in
g
in
a
p
o
wer
n
etwo
r
k
[
1
8
]
.
Hen
ce
,
b
ad
o
r
p
o
o
r
PQ
m
ay
e
x
is
t
d
u
e
to
tr
an
s
ien
t
co
n
d
itio
n
s
in
th
e
p
o
wer
n
etwo
r
k
o
r
with
in
th
e
c
o
n
n
e
ctio
n
o
f
n
o
n
lin
ea
r
lo
a
d
s
[
1
9
]
.
Utilizatio
n
o
f
m
o
r
e
s
en
s
itiv
e
lo
ad
s
s
u
ch
as
co
m
p
u
ter
s
,
in
d
u
s
tr
ial
d
r
iv
es,
telec
o
m
m
u
n
i
ca
tio
n
s
an
d
m
e
d
ical
eq
u
i
p
m
en
t
in
p
o
wer
s
y
s
tem
n
etwo
r
k
m
a
y
also
lead
s
to
PQ
p
r
o
b
lem
s
[
2
0
]
.
PQ
p
r
o
b
lem
s
in
clu
d
e
v
o
ltag
e
s
ag
o
r
d
ip
,
v
o
l
tag
e
s
well,
p
o
wer
in
ter
r
u
p
tio
n
s
,
v
o
ltag
e
f
lick
er
,
v
o
ltag
e
s
u
r
g
es,
v
o
ltag
e
s
p
ik
e
s
,
s
witch
in
g
tr
an
s
ien
ts
,
f
r
eq
u
en
cy
v
ar
iatio
n
s
,
elec
tr
ical
lin
e
n
o
is
e,
b
r
o
w
n
o
u
ts
,
b
lack
o
u
ts
,
n
o
tch
as c
o
n
tain
e
d
in
T
ab
le
1
.
Fo
r
th
at
r
ea
s
o
n
,
An
an
d
an
d
Sriv
astav
a
[
2
1
]
d
ef
in
e
d
PQ
p
r
o
b
lem
s
as
an
y
d
if
f
icu
lty
d
i
s
p
l
ay
ed
in
r
elatio
n
to
v
o
ltag
e,
cu
r
r
en
t
o
r
lead
in
g
to
f
r
eq
u
en
cy
d
e
v
ia
tio
n
s
wh
ich
y
ield
s
to
f
ailu
r
e
o
r
m
alf
u
n
ctio
n
o
f
cu
s
to
m
er
ap
p
lian
ce
s
.
C
o
n
s
eq
u
en
tly
,
PQ
p
r
o
b
lem
s
h
av
e
r
es
u
lted
in
lo
s
t
tim
e,
lo
s
t
p
r
o
d
u
ctio
n
,
p
r
o
d
u
ctio
n
o
f
s
cr
ap
s
,
lo
s
t
s
ales
an
d
co
n
v
ey
an
ce
d
e
lay
s
as
well
as
d
am
a
g
ed
p
r
o
d
u
ctio
n
eq
u
ip
m
en
t.
T
h
e
s
o
u
r
ce
s
o
f
PQ
p
r
o
b
lem
s
ca
n
v
ar
y
,
r
an
g
in
g
f
r
o
m
n
atu
r
al
p
h
en
o
m
en
a
s
u
c
h
a
s
lig
h
tn
in
g
,
f
l
o
o
d
s
an
d
ea
r
th
q
u
ak
es
to
m
an
m
a
d
e
in
d
u
ce
d
li
k
e
en
er
g
izatio
n
o
f
c
ap
ac
ito
r
b
a
n
k
s
an
d
tr
an
s
f
o
r
m
e
r
s
,
s
witch
in
g
o
r
s
tar
t
-
u
p
o
f
l
ar
g
e
in
d
u
cti
o
n
m
o
to
r
lo
ad
s
,
o
p
er
atio
n
o
f
u
n
s
y
m
m
e
tr
ical
n
o
n
-
lin
ea
r
lo
ad
s
,
f
ailu
r
e
o
f
d
is
tr
ib
u
tio
n
s
y
s
tem
eq
u
ip
m
en
t
an
d
wr
o
n
g
co
n
n
ec
tio
n
s
in
d
is
tr
ib
u
tio
n
s
u
b
s
tatio
n
s
an
d
co
n
s
u
m
e
r
’
s
p
r
e
m
is
es.
T
ab
le
1
.
Descr
ip
tio
n
o
f
s
o
m
e
PQ e
v
en
ts
class
if
icatio
n
,
d
u
r
atio
n
an
d
v
o
ltag
e
m
ag
n
itu
d
e
[
2
2
]
S
/
N
C
a
t
e
g
o
r
y
D
u
r
a
t
i
o
n
V
o
l
t
a
g
e
m
a
g
n
i
t
u
d
e
1
V
o
l
t
a
g
e
s
a
g
0
.
5
c
y
c
l
e
–
1
m
i
n
s
0
.
1
–
0
.
9
p
u
2
V
o
l
t
a
g
e
sw
e
l
l
0
.
5
c
y
c
l
e
–
1
m
i
n
s
0
.
1
–
1
.
8
p
u
3
I
n
t
e
r
r
u
p
t
i
o
n
0
.
5
c
y
c
l
e
–
1
m
i
n
s
<
0
.
1
p
u
4
Tr
a
n
s
i
e
n
t
s
a.
I
mp
u
l
si
v
e
b.
O
sci
l
l
a
t
o
r
y
5
0
n
se
c
–
1
mse
c
5
µ
s
e
c
–
5
0
mse
c
<
0
.
8
p
u
5
O
v
e
r
v
o
l
t
a
g
e
>
1
m
i
n
1
.
1
–
1
.
2
p
u
6
U
n
d
e
r
v
o
l
t
a
g
e
>
1
m
i
n
0
.
8
–
0
.
9
p
u
7
V
o
l
t
a
g
e
i
m
b
a
l
a
n
c
e
S
t
e
a
d
y
st
a
t
e
0
.
5
–
2%
8
H
a
r
mo
n
i
c
S
t
e
a
d
y
st
a
t
e
9
N
o
t
c
h
S
t
e
a
d
y
st
a
t
e
10
N
o
i
se
S
t
e
a
d
y
st
a
t
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
38
,
No
.
1
,
Ap
r
il
20
25
:
1
-
21
4
Fro
m
th
e
f
o
r
eg
o
in
g
d
is
cu
s
s
i
o
n
,
t
h
e
PQ
p
r
o
b
lem
s
h
av
e
b
ec
o
m
e
m
u
ch
m
o
r
e
c
o
m
p
lic
ated
with
p
r
o
life
r
atio
n
o
f
s
o
lid
-
s
tate
co
n
tr
o
ller
s
[
2
3
]
,
u
s
ag
e
o
f
wh
ic
h
co
u
l
d
n
o
t
b
e
o
v
er
l
o
o
k
e
d
b
ec
au
s
e
o
f
th
ei
r
co
s
t
ad
v
an
tag
es,
r
e
d
u
ctio
n
in
s
ize,
en
er
g
y
p
r
eser
v
atio
n
,
ea
s
y
co
n
tr
o
l,
lo
w
wea
r
a
n
d
tear
a
n
d
o
th
er
m
ain
ten
a
n
c
e
ad
v
an
tag
es
th
ey
o
f
f
er
to
t
h
e
m
o
d
er
n
elec
tr
ic
s
y
s
tem
[
2
4
]
.
T
a
b
le
2
p
r
esen
ts
s
o
m
e
co
m
m
o
n
ly
u
s
ed
m
ath
em
atica
l
ex
p
r
ess
io
n
s
f
o
r
p
ar
am
etr
ic
v
ar
iatio
n
s
d
escr
ib
i
n
g
s
o
m
e
PQ
p
r
o
b
lem
s
.
Ho
wev
er
,
th
e
s
o
lid
-
s
tate
co
n
tr
o
l
d
e
v
ic
es,
th
e
cu
s
to
m
er
s
lo
ad
,
as
well
as
th
e
g
en
er
atio
n
s
y
s
tem
h
av
e
all
b
ee
n
id
en
t
if
ied
as
s
o
u
r
ce
s
o
f
PQ
p
r
o
b
lem
s
[
2
5
]
.
Fo
r
th
is
r
e
aso
n
,
PQ
h
a
d
b
ec
o
m
e
an
im
p
o
r
tan
t
f
ield
o
f
r
esear
ch
i
n
ele
ctr
ical
en
g
in
ee
r
in
g
.
E
x
ce
p
tio
n
ally
,
in
r
ad
ial
d
is
tr
ib
u
tio
n
n
etwo
r
k
(
R
DN)
c
h
ar
ac
ter
ized
b
y
elev
ated
p
o
wer
lo
s
s
es
lead
in
g
to
h
ig
h
R
/X
r
atio
r
esu
ltin
g
i
n
ap
p
r
o
x
im
ately
1
0
to
1
3
%
l
o
s
s
es
o
f
th
e
p
r
o
d
u
ce
d
p
o
wer
s
er
io
u
s
ly
af
f
ec
t
th
e
s
y
s
tem
n
etwo
r
k
[
2
6
]
.
T
h
is
m
e
n
ac
e
h
as
p
o
s
ed
s
er
io
u
s
c
h
allen
g
es
t
o
b
o
t
h
u
tili
ties
an
d
e
q
u
ip
m
e
n
t
m
an
u
f
ac
t
u
r
e
r
s
in
m
ee
tin
g
th
e
cu
s
to
m
er
’
s
eq
u
ip
m
en
t
PQ
r
eq
u
ir
em
en
ts
as
s
tip
u
lated
b
y
th
e
I
E
E
E
STD
5
1
9
o
f
1
9
9
2
.
C
o
n
v
er
s
ely
,
s
ev
er
al
ap
p
r
o
ac
h
es h
av
e
e
v
o
lv
ed
f
o
r
t
h
e
r
ed
u
ctio
n
o
f
PQ p
r
o
b
lem
s
s
ee
[
2
7
]
as a
n
e
x
am
p
le.
T
ab
le
2
.
E
x
p
r
ess
io
n
s
an
d
p
ar
a
m
eter
v
ar
iatio
n
s
o
f
v
ar
io
u
s
PQ p
r
o
b
lem
s
[
2
8
]
P
Q
p
r
o
b
l
e
m
M
a
t
h
e
ma
t
i
c
a
l
e
x
p
r
e
ssi
o
n
P
a
r
a
me
t
e
r
v
a
r
i
a
t
i
o
n
P
u
r
e
s
i
n
e
w
a
v
e
(
)
=
(
)
=
1
;
=
2
V
o
l
t
a
g
e
s
a
g
(
)
=
(
1
−
(
(
−
1
)
−
(
−
2
)
)
)
si
n
(
)
0
.
1
≤
≤
0
.
9
;
≤
2
−
1
≤
9
V
o
l
t
a
g
e
sw
e
l
l
(
)
=
(
1
+
(
(
−
1
)
−
(
−
2
)
)
)
si
n
(
)
0
.
1
≤
≤
0
.
9
;
≤
2
−
1
≤
9
I
n
t
e
r
r
u
p
t
i
o
n
(
)
=
(
1
−
(
(
−
1
)
−
(
−
2
)
)
)
si
n
(
)
0
.
9
≤
≤
1
;
≤
2
−
1
≤
9
Tr
a
n
s
i
e
n
t
(
)
=
(
si
n
(
)
+
−
(
−
1
)
si
n
(
(
−
1
)
)
)
−
2
≤
≤
2
50
≤
≤
100
500
≤
≤
15
H
a
r
mo
n
i
c
s
(
)
=
(
1
si
n
(
)
+
2
si
n
(
2
)
+
3
si
n
(
3
)
+
5
si
n
(
5
)
+
7
si
n
(
7
)
)
1
=
1
.
0
2
−
7
=
0
.
0
−
0
.
3
F
l
i
c
k
e
r
(
)
=
(
1
+
si
n
(
)
)
si
n
(
)
0
.
01
≤
≤
0
.
25
2
≤
≤
8
N
o
i
se
(
)
=
si
n
(
)
+
(
(
si
n
(
)
)
=
0
.
1
R
ec
en
tly
,
m
an
y
g
r
i
d
s
’
s
tak
eh
o
ld
er
s
p
er
f
o
r
m
s
p
o
n
ta
n
eo
u
s
P
Q
m
o
n
ito
r
in
g
t
o
g
et
r
elev
a
n
t
d
ata
o
f
th
e
p
o
wer
b
ein
g
s
u
p
p
lied
an
d
e
q
u
ip
m
en
t
p
er
f
o
r
m
an
ce
.
T
h
e
tim
e
tak
en
f
o
r
l
o
n
g
ter
m
PQ
m
ea
s
u
r
em
en
ts
r
esu
lted
in
h
u
g
e
d
ata
to
b
e
h
an
d
lin
g
.
Actu
al
p
er
f
o
r
m
a
n
ce
o
f
PQ
m
o
n
ito
r
in
g
d
ev
ice
r
elies
o
n
its
ca
p
ab
ilit
y
to
an
aly
ze
an
d
p
r
esen
ts
v
o
lu
m
in
o
u
s
r
aw
d
ata
o
b
tain
ed
f
r
o
m
m
o
n
ito
r
in
g
ex
er
cise.
B
u
t n
o
twith
s
tan
d
in
g
,
it
co
n
s
u
m
es
m
o
r
e
tim
e
an
d
m
ay
n
o
t
y
ield
s
p
er
f
ec
t
r
esu
lt.
Or
th
o
d
o
x
s
cien
tif
ic
to
o
ls
ar
e
s
til
l
r
eq
u
ir
ed
to
s
p
ee
d
u
p
b
ig
d
ata
in
ter
p
r
etatio
n
a
n
aly
s
is
with
ex
ce
llen
t
r
esu
lt
[
2
9
]
.
T
h
e
PQ
b
i
g
d
ata
is
n
o
th
in
g
b
u
t
h
u
g
e
am
o
u
n
t
o
f
d
ata
s
eq
u
el
to
co
n
tin
u
o
u
s
PQ
m
o
n
ito
r
in
g
[
3
0
]
.
Hen
ce
,
b
ig
d
ata
is
v
o
lu
m
in
o
u
s
o
r
lar
g
e
am
o
u
n
t
o
f
d
ata
with
s
p
ec
if
ic
co
m
p
lex
ities
d
escr
ib
ed
b
y
4
V’
s
(
i.e
.
Vo
lu
m
e,
Velo
city
,
Var
iety
an
d
Ver
ac
ity
)
d
ep
icted
in
T
ab
le
3
.
T
ab
le
3
.
Descr
ip
tio
n
o
f
4
Vs f
o
r
PQ b
ig
d
ata
[
2
9
]
S
/
N
P
a
r
a
me
t
e
r
D
e
scri
p
t
i
o
n
1
V
o
l
u
me
Th
i
s
i
s
a
m
o
u
n
t
,
s
i
z
e
a
n
d
s
c
a
l
e
o
f
d
a
t
a
t
h
a
t
c
o
u
l
d
n
o
t
b
e
m
a
n
a
g
e
d
w
i
t
h
o
u
t
d
e
d
i
c
a
t
e
d
a
n
a
l
y
t
i
c
t
o
o
l
s
.
2
V
e
l
o
c
i
t
y
Th
e
sp
e
e
d
a
t
w
h
i
c
h
d
a
t
a
i
s
g
e
n
e
r
a
t
e
d
a
n
d
h
o
w
f
a
st
t
h
e
d
a
t
a
s
h
o
u
l
d
b
e
p
r
o
c
e
ssed
i
s
t
e
r
m
e
d
v
e
l
o
c
i
t
y
.
3
V
a
r
i
e
t
y
Th
i
s
i
s
h
e
t
e
r
o
g
e
n
e
i
t
y
o
f
d
a
t
a
b
e
i
n
g
u
t
i
l
i
z
e
d
.
B
i
g
d
a
t
a
a
l
w
a
y
s
c
o
mes
f
r
o
m
v
a
r
i
o
u
s s
o
u
r
c
e
s,
w
h
i
c
h
c
a
n
b
e
d
i
f
f
e
r
e
n
t
i
n
t
y
p
e
s,
f
o
r
ma
t
,
s
e
m
a
n
t
i
c
a
n
d
v
o
l
u
m
e
.
4
V
e
r
a
c
i
t
y
Q
u
a
l
i
t
y
o
f
c
o
l
l
e
c
t
e
d
d
a
t
a
i
s refer
s
t
o
a
s v
e
r
a
c
i
t
y
.
I
t
i
s
c
o
n
c
e
r
n
e
d
w
i
t
h
b
i
a
ses,
n
o
i
s
e
a
n
d
a
b
n
o
r
m
a
l
i
t
y
i
n
t
h
e
d
a
t
a
.
A
c
c
u
r
a
c
y
o
f
a
n
y
a
n
a
l
y
t
i
c
p
r
o
c
e
ss
a
p
p
l
i
e
d
t
o
t
h
e
d
a
t
a
d
e
p
e
n
d
s
g
r
e
a
t
l
y
o
n
t
h
e
v
e
r
a
c
i
t
y
o
f
t
h
e
s
o
u
r
c
e
d
a
t
a
.
3.
CO
NVO
L
U
T
I
O
NA
L
NE
UR
AL
NE
T
WO
RK
W
ith
in
th
e
s
p
h
er
e
o
f
DL
r
esear
ch
,
C
NN
tech
n
iq
u
e
is
th
e
m
o
s
t
u
tili
ze
d
s
tr
aig
h
tf
o
r
war
d
alg
o
r
ith
m
[
3
1
]
.
C
h
ief
s
ig
n
if
ican
t
ch
ar
ac
ter
is
tic
o
f
C
NN
co
m
p
ar
ed
to
its
an
tece
d
en
t
is
th
at
it
au
to
m
atica
lly
r
ec
o
g
n
izes
p
er
tin
en
t
attr
i
b
u
tes
in
th
e
ab
s
e
n
ce
o
f
h
u
m
an
s
u
p
e
r
v
is
io
n
[
3
2
]
.
T
h
e
C
NNs
h
ad
b
ee
n
am
p
ly
em
p
lo
y
ed
in
wid
e
ar
ea
o
f
d
if
f
er
en
t
f
ield
s
,
in
c
lu
d
in
g
co
m
p
u
ter
v
is
io
n
,
s
p
e
ec
h
p
r
o
ce
s
s
in
g
,
f
ac
e
r
ec
o
g
n
itio
n
,
an
d
im
ag
e
class
if
icatio
n
.
T
ab
le
4
d
escr
ib
ed
ev
o
lu
tio
n
o
f
C
NN
ar
ch
itectu
r
es a
n
d
th
eir
ch
ar
ac
ter
is
tics
o
v
er
tim
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
A
r
ev
iew
o
f c
o
n
vo
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
ks fo
r
cla
s
s
i
fyin
g
p
o
w
er q
u
a
lity p
r
o
b
lems u
s
in
g
…
(
A
d
a
mu
S
a
’
id
u
)
5
T
ab
le
4
.
E
v
o
lu
tio
n
o
f
C
NN
ar
ch
itectu
r
es o
v
er
tim
e
[
3
3
]
S
/
N
C
N
N
T
y
p
e
Y
e
a
r
C
h
a
r
a
c
t
e
r
i
s
t
i
c
s
D
a
t
a
s
e
t
1
Le
N
e
t
1
9
9
8
I
t
w
a
s t
h
e
f
i
r
s
t
p
o
p
u
l
a
r
C
N
N
a
r
c
h
i
t
e
c
t
u
r
e
,
i
t
h
a
s s
e
v
e
n
l
a
y
e
r
s,
i
t
u
s
e
s
R
e
LU
a
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
M
N
I
S
T
2
A
l
e
x
N
e
t
2
0
1
2
I
t
i
s d
e
e
p
e
r
a
n
d
w
i
d
e
r
c
o
m
p
a
r
e
d
t
o
Le
N
e
t
,
I
t
u
ses
R
ELU
,
d
r
o
p
o
u
t
a
n
d
o
v
e
r
l
a
p
p
o
o
l
i
n
g
G
P
U
s NV
I
D
I
A
G
TX
5
8
0
I
mag
e
N
e
t
3
V
G
G
(
V
i
su
a
l
g
e
o
me
t
r
y
g
r
o
u
p
)
2
0
1
4
I
t
u
ses
sm
a
l
l
-
si
z
e
d
f
i
l
t
e
r
s
a
n
d
h
a
s
h
o
m
o
g
e
n
e
o
u
s
t
o
p
o
l
o
g
y
I
mag
e
N
e
t
4
G
o
o
g
L
e
N
e
t
2
0
1
5
I
t
i
s
t
h
e
f
i
r
st
C
N
N
a
r
c
h
i
t
e
c
t
u
r
e
t
o
i
n
t
r
o
d
u
c
e
b
l
o
c
k
c
o
n
c
e
p
t
a
n
d
d
i
f
f
e
r
e
n
t
f
i
l
t
e
r
si
z
e
.
I
t
u
s
e
s s
p
l
i
t
t
r
a
n
sf
o
r
m
t
o
m
e
r
g
e
i
d
e
a
I
mag
e
N
e
t
5
I
n
c
e
p
t
i
o
n
V
-
3
2
0
1
5
I
t
i
s a
b
l
e
t
o
t
a
k
e
c
a
r
e
o
f
b
o
t
t
l
e
n
e
c
k
i
ss
u
e
a
n
d
u
s
e
d
sma
l
l
f
i
l
t
e
r
s
i
n
s
t
e
a
d
o
f
l
a
r
g
e
f
i
l
t
e
r
s.
I
t
h
a
s
b
e
t
t
e
r
f
e
a
t
u
r
e
r
e
p
r
e
se
n
t
a
t
i
o
n
.
I
mag
e
N
e
t
6
I
n
c
e
p
t
i
o
n
V
-
4
2
0
1
6
I
t
u
ses
a
sy
m
met
r
i
c
f
i
l
t
e
r
s w
i
t
h
s
p
l
i
t
t
r
a
n
sf
o
r
m a
n
d
i
n
t
e
g
r
a
t
i
o
n
c
o
n
c
e
p
t
I
mag
e
N
e
t
7
R
e
sN
e
t
2
0
1
6
I
t
i
s ro
b
u
s
t
a
g
a
i
n
st
o
v
e
r
f
i
t
t
i
n
g
d
u
e
t
o
s
y
mm
e
t
r
y
ma
p
p
i
n
g
-
b
a
s
e
d
sk
i
p
c
o
n
n
e
c
t
i
o
n
s wi
t
h
r
e
si
d
u
a
l
l
e
a
r
n
i
n
g
I
mag
e
N
e
t
,
C
I
F
A
R
-
10
8
X
c
e
p
t
i
o
n
2
0
1
7
I
t
h
a
s
d
e
p
t
h
-
b
a
se
d
c
o
n
v
o
l
u
t
i
o
n
f
o
l
l
o
w
e
d
b
y
p
o
i
n
t
-
b
a
se
d
c
o
n
v
o
l
u
t
i
o
n
I
mag
e
N
e
t
9
R
e
si
d
u
a
l
A
t
t
e
n
t
i
o
n
N
e
u
r
a
l
N
e
t
w
o
r
k
2
0
1
7
I
t
i
s t
h
e
f
i
r
st
C
N
N
a
r
c
h
i
t
e
c
t
u
r
e
t
o
i
n
t
r
o
d
u
c
e
a
t
t
e
n
t
i
o
n
m
e
c
h
a
n
i
s
m e
.
g
.
Tr
a
n
sf
o
r
mer
-
b
a
se
d
C
N
N
S
a
n
d
n
o
n
-
l
o
c
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
C
I
F
A
R
-
10,
C
I
F
A
R
-
1
0
0
,
I
mag
e
N
e
t
10
R
e
sN
e
x
t
2
0
1
7
I
t
i
n
st
i
t
u
t
e
d
p
r
i
c
i
p
a
l
i
t
y
,
h
o
m
o
g
e
n
e
o
u
s
t
o
p
o
l
o
g
y
a
n
d
g
r
o
u
p
e
d
c
o
n
v
o
l
u
t
i
o
n
C
I
F
A
R
-
10,
C
I
F
A
R
-
1
0
0
,
I
mag
e
N
e
t
11
S
q
u
e
e
z
e
a
n
d
Ex
c
i
t
a
t
i
o
n
N
e
t
w
o
r
k
(
S
EN
e
t
)
2
0
1
7
I
t
mo
d
e
l
e
d
i
n
t
e
r
d
e
p
e
n
d
e
n
c
i
e
s
b
e
t
w
e
e
n
f
e
a
t
u
r
e
m
a
p
s.
I
n
c
o
r
p
o
r
a
t
e
s
c
h
a
n
n
e
l
-
w
i
se
a
t
t
e
n
t
i
o
n
t
o
a
d
a
p
t
i
v
e
l
y
r
e
c
a
l
i
b
r
a
t
e
f
e
a
t
u
r
e
ma
p
s.
It
e
n
h
a
n
c
e
s fe
a
t
u
r
e
r
e
p
r
e
s
e
n
t
a
t
i
o
n
I
mag
e
N
e
t
12
D
e
n
seN
e
t
2
0
1
7
I
t
h
a
s
b
l
o
c
k
s
o
f
l
a
y
e
r
s,
l
a
y
e
r
s c
o
n
n
e
c
t
e
d
t
o
e
a
c
h
o
t
h
e
r
a
n
d
C
r
o
ss l
a
y
e
r
i
n
f
o
r
mat
i
o
n
f
l
o
w
C
I
F
A
R
-
1
0
,
C
I
F
A
R
-
1
0
0
,
I
mag
e
N
e
t
13
M
o
b
i
l
e
N
e
t
V
1
2
0
1
8
I
t
h
a
s
i
n
v
e
r
t
e
d
r
e
s
i
d
u
a
l
s
t
r
u
c
t
u
r
e
.
D
e
si
g
n
e
d
f
o
r
m
o
b
i
l
e
a
n
d
e
m
b
e
d
d
e
d
d
e
v
i
c
e
s,
d
e
p
t
h
w
i
s
e
s
e
p
a
r
a
b
l
e
c
o
n
v
o
l
u
t
i
o
n
t
o
r
e
d
u
c
e
c
o
m
p
u
t
a
t
i
o
n
I
mag
e
N
e
t
14
H
R
N
e
t
V
2
2
0
2
0
H
i
g
h
-
r
e
so
l
u
t
i
o
n
r
e
p
r
e
s
e
n
t
a
t
i
o
n
I
mag
e
N
e
t
15
D
i
c
e
N
E
T
2
0
2
1
I
n
i
t
i
a
t
e
d
s
i
z
e
-
b
a
s
e
d
C
N
N
,
i
n
c
o
r
p
o
r
a
t
i
n
g
h
e
i
g
h
t
,
w
i
d
t
h
,
a
n
d
d
e
p
t
h
T
h
er
ef
o
r
e,
C
NNs
h
ad
ev
o
lv
ed
r
esu
ltin
g
to
v
a
r
io
u
s
ar
ch
i
tectu
r
es
tailo
r
ed
f
o
r
d
if
f
er
e
n
t
task
s
in
co
m
p
u
ter
v
is
io
n
.
So
m
e
o
f
t
h
e
n
o
tab
le
C
NN
v
ar
ian
ts
an
d
th
eir
d
if
f
e
r
en
ce
s
ar
e
id
en
tifie
d
i
n
th
is
wo
r
k
r
e
f
er
t
o
T
ab
les
4
an
d
5
f
o
r
m
o
r
e
in
f
o
r
m
atio
n
.
T
h
ese
v
a
r
ian
ts
d
if
f
er
in
ter
m
s
o
f
ar
c
h
itectu
r
e,
d
ep
t
h
,
s
k
ip
co
n
n
ec
tio
n
s
an
d
s
p
ec
ialized
f
ea
tu
r
es.
Pre
s
en
tly
,
r
esear
ch
er
s
co
n
tin
u
e
to
ex
p
l
o
r
e
n
ew
C
NN
ar
c
h
itectu
r
es
to
im
p
r
o
v
e
p
er
f
o
r
m
an
ce
an
d
ad
d
r
ess
s
p
ec
if
ic
ch
allen
g
es in
co
m
p
u
ter
v
i
s
io
n
task
s
[
3
4
]
.
T
h
e
ar
ch
itectu
r
al
v
ar
ian
ts
o
f
C
NN
s
u
c
h
as
Alex
Net,
VGG,
Go
o
g
L
eNe
t
an
d
R
esNet
f
o
r
ex
am
p
l
e,
ar
e
m
o
s
tly
e
m
p
l
o
y
e
d
i
n
t
h
e
c
l
a
s
s
i
f
ic
a
t
i
o
n
o
f
PQ
p
r
o
b
l
e
m
s
s
i
g
n
a
l
s
.
N
o
n
e
t
h
e
l
ess
,
t
h
e
s
t
r
u
c
t
u
r
e
s
w
e
r
e
f
ac
i
n
g
h
u
r
d
l
e
i
n
c
o
n
v
e
r
g
e
n
c
e
,
o
v
e
r
f
i
tt
i
n
g
i
s
s
u
e
a
n
d
v
a
n
i
s
h
i
n
g
g
r
a
d
i
e
n
t
p
r
o
b
l
e
m
s
.
A
ls
o
,
t
h
e
i
r
a
c
c
u
r
a
c
i
es
a
n
d
a
d
e
q
u
a
t
e
p
e
r
f
o
r
m
a
n
c
e
s
c
r
i
t
e
r
i
a
s
t
i
ll
r
e
q
u
i
r
e
s
i
m
p
r
o
v
e
m
e
n
t
a
n
d
s
o
m
e
l
it
t
l
e
m
o
d
i
f
i
c
a
ti
o
n
s
i
n
t
h
e
m
o
d
e
l
.
T
h
e
v
e
r
y
g
o
o
d
v
a
r
i
a
n
t
t
o
a
d
d
r
e
s
s
a
ll
t
h
es
e
p
r
o
b
l
em
s
is
D
e
n
s
e
N
et
.
I
t
i
s
r
e
c
o
m
m
e
n
d
e
d
f
o
r
t
h
i
s
w
o
r
k
b
e
c
a
u
s
e
o
f
i
ts
s
i
m
p
l
i
ci
t
y
a
n
d
e
as
e
i
n
a
p
p
l
i
c
a
ti
o
n
.
D
e
n
s
eN
e
t
is
c
o
m
m
o
n
a
r
c
h
i
t
e
c
t
u
r
e
o
f
C
NN
f
o
r
i
m
a
g
e
r
e
c
o
g
n
it
i
o
n
t
h
a
t
h
as
a
c
q
u
i
r
e
d
c
u
r
r
e
n
t
s
t
r
u
ctu
r
e
w
i
t
h
l
o
w
p
a
r
a
m
e
t
e
r
s
h
a
v
i
n
g
d
i
f
f
e
r
e
n
t
n
e
tw
o
r
k
s
s
u
c
h
as
De
n
s
eN
e
t
–
1
2
1
,
D
e
n
s
e
N
e
t
–
160
a
n
d
D
e
n
s
e
N
et
–
2
0
1
.
A
m
o
n
g
t
h
e
t
h
r
e
e
n
e
tw
o
r
k
s
,
D
e
n
s
e
N
e
t
–
121
is
s
p
e
c
i
f
i
ca
l
l
y
r
e
c
o
m
m
e
n
d
e
d
f
o
r
t
h
i
s
r
e
s
e
a
r
c
h
b
e
c
a
u
s
e
i
t
c
a
n
r
e
d
u
c
e
t
h
e
i
s
s
u
e
o
f
v
a
n
i
s
h
i
n
g
g
r
a
d
i
e
n
t
,
a
d
v
o
c
a
t
e
s
f
e
a
t
u
r
e
r
e
u
s
e
a
n
d
l
e
s
s
e
n
s
p
a
r
am
e
t
e
r
i
n
v
o
l
v
e
m
e
n
t
,
w
h
i
c
h
a
r
e
m
a
j
o
r
f
a
c
t
o
r
s
f
o
r
t
r
a
i
n
i
n
g
DL
m
o
d
e
l
s
.
M
o
r
e
s
o
,
D
e
n
s
e
Ne
t
–
1
2
1
h
a
d
p
r
o
v
e
n
t
o
b
e
e
f
f
i
c
i
e
n
t
i
n
p
i
n
p
o
i
n
t
i
n
g
P
Q
p
r
o
b
l
e
m
s
o
n
t
h
e
b
a
s
i
s
o
f
s
i
g
n
a
l
a
n
a
l
y
s
is
.
I
t
h
a
s
a
t
o
t
a
l
o
f
1
2
1
l
a
y
e
r
s
f
o
r
c
o
n
v
o
lu
t
i
o
n
a
l
,
t
r
a
n
s
i
t
i
o
n
,
cl
a
s
s
i
f
ic
a
t
i
o
n
a
n
d
d
e
n
s
e
b
l
o
c
k
l
a
y
e
r
s
a
n
d
i
s
v
e
r
y
p
o
p
u
l
a
r
D
L
a
r
c
h
i
t
e
c
t
u
r
e
i
n
i
m
a
g
e
cl
a
s
s
i
f
ic
a
t
i
o
n
.
T
h
e
D
e
n
s
eN
e
t
C
N
N
a
r
c
h
i
t
e
c
t
u
r
e
u
s
e
s
R
e
L
U
a
ct
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
a
n
d
d
r
o
p
o
u
t
t
o
d
e
m
o
n
s
t
r
a
t
e
e
f
f
e
c
t
iv
e
n
e
s
s
i
n
c
la
s
s
i
f
i
ca
t
i
o
n
p
e
r
f
o
r
m
a
n
c
e
.
T
h
e
v
a
r
i
a
n
t
s
o
f
C
N
N
s
w
e
r
e
c
o
n
s
p
i
r
e
d
w
it
h
n
e
u
r
o
n
s
i
n
h
u
m
a
n
a
n
d
a
n
i
m
a
l
b
r
a
i
n
s
,
v
e
r
y
m
u
c
h
a
l
i
k
e
t
o
AN
N
[
3
5
]
.
I
n
h
u
m
a
n
b
r
a
i
n
,
c
o
m
p
l
e
x
s
e
r
i
es
o
f
c
e
l
ls
f
o
r
m
s
t
h
e
v
i
s
u
a
l
c
o
r
t
e
x
a
n
d
t
h
is
s
e
q
u
e
n
c
e
is
s
i
m
u
la
t
e
d
b
y
t
h
e
C
N
N
.
DL
,
i
n
it
s
r
e
m
a
r
k
a
b
l
e
s
u
c
c
e
s
s
,
p
r
e
s
e
n
t
l
y
i
s
o
n
e
o
f
t
h
e
w
e
l
l
-
k
n
o
w
n
r
e
s
e
a
r
c
h
a
r
e
a
s
i
n
t
h
e
f
i
e
l
d
o
f
ML
[
3
6
]
.
O
u
t
s
t
a
n
d
i
n
g
d
i
f
f
e
r
e
n
c
e
s
b
e
t
w
e
e
n
ML
a
n
d
D
L
a
p
p
r
o
a
c
h
e
s
w
e
r
e
d
e
p
i
c
t
e
d
i
n
F
i
g
u
r
e
s
1
a
n
d
2
r
e
s
p
e
c
t
i
v
e
l
y
.
Fig
u
r
e
1
.
ML
tec
h
n
iq
u
e
[
3
7
]
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
38
,
No
.
1
,
Ap
r
il
20
25
:
1
-
21
6
Fig
u
r
e
2
.
DL
tec
h
n
iq
u
e
[
3
7
]
So
m
e
o
f
th
e
m
ajo
r
b
e
n
ef
it
s
o
f
C
NN
ar
e
eq
u
al
d
escr
i
p
tio
n
s
,
s
ca
tter
ed
i
n
ter
co
n
n
ec
tio
n
s
an
d
p
ar
am
eter
s
s
h
ar
in
g
.
Fu
lly
c
o
n
n
ec
ted
(
FC
)
lay
er
,
s
h
ar
ed
wei
g
h
t
an
d
lo
ca
l
in
te
r
co
n
n
ec
tio
n
o
f
C
NN
s
tr
u
ctu
r
e
was
ap
p
lied
to
m
ak
e
co
m
p
lete
u
s
ag
e
o
f
2
D
in
p
u
t
d
ata
s
tr
u
ctu
r
es
s
u
ch
as
im
ag
e
s
ig
n
als.
Nev
er
th
eless
,
C
NN
ca
n
m
an
ag
e
PQ
b
ig
d
ata
v
e
r
y
f
ast
an
d
p
r
o
d
u
ce
g
o
o
d
r
esu
lts
.
T
h
is
o
p
er
atio
n
ex
tr
e
m
ely
u
tili
ze
s
s
m
all
n
u
m
b
er
s
o
f
p
a
r
am
eter
wh
ich
s
p
ee
d
s
u
p
th
e
tr
ain
i
n
g
p
r
o
ce
s
s
es
an
d
ac
ce
ler
ates
m
o
d
el
p
e
r
f
o
r
m
an
c
e
[
3
8
]
.
A
c
o
m
m
o
n
C
NN
ty
p
e
wh
ich
is
clo
s
e
to
th
e
m
u
lti
-
lay
er
p
er
ce
p
tr
o
n
(
ML
P)
co
n
s
is
ts
o
f
co
n
v
o
lu
tio
n
lay
er
s
,
ac
tiv
atio
n
f
u
n
ctio
n
,
s
u
b
-
s
am
p
lin
g
o
r
p
o
o
lin
g
lay
er
s
an
d
FC
lay
er
s
as ill
u
s
tr
ated
in
Fig
u
r
e
3
.
Fig
u
r
e
3
.
C
NN
ar
ch
itectu
r
e
3
.
1
.
O
pera
t
io
n o
f
CNN
T
h
e
b
asic
o
p
er
atio
n
o
f
C
NN
is
p
r
esen
ted
b
ased
o
n
th
e
wo
r
k
in
g
p
r
in
ci
p
le
o
f
ea
ch
lay
er
in
its
co
m
p
o
s
itio
n
as f
o
llo
ws:
Co
nv
o
lutio
na
l
la
y
er
:
T
h
e
m
o
s
t
s
ig
n
if
ican
t
lay
er
in
C
NN
t
r
ain
in
g
p
r
o
ce
s
s
is
th
e
co
n
v
o
l
u
tio
n
al
lay
er
w
h
ich
co
n
tain
s
co
llectio
n
o
f
f
ilter
s
ca
lled
k
er
n
els
[
3
9
]
.
I
n
p
u
t
s
ig
n
al
to
th
e
C
NN
is
ex
p
r
ess
ed
as
N
-
d
im
en
s
io
n
al
m
atr
ix
,
co
n
v
o
l
v
ed
with
th
e
f
il
ter
s
to
p
r
o
d
u
ce
th
e
o
u
tp
u
t
f
ea
tu
r
e
m
ap
.
Usu
ally
,
a
g
r
id
o
f
d
is
cr
ete
n
u
m
b
er
s
o
r
v
alu
es
r
ep
r
esen
ts
th
e
k
er
n
el
k
n
o
wn
as
th
e
k
er
n
el
weig
h
t.
As
s
ig
n
ed
ar
b
itra
r
ily
v
al
u
es
ac
t
as
th
e
weig
h
ts
o
f
th
e
k
er
n
el
at
th
e
in
itial st
ag
e
o
f
t
h
e
C
NN
tr
ain
in
g
ex
er
cise.
Als
o
,
th
er
e
wer
e
m
a
n
y
way
s
em
p
l
o
y
ed
to
in
itialize
th
e
weig
h
ts
[
4
0
]
.
C
o
n
s
eq
u
e
n
tly
,
th
e
k
er
n
el
weig
h
ts
ar
e
ad
ju
s
t
ed
at
ev
er
y
tr
ain
in
g
tim
e
to
ex
tr
ac
t
s
ig
n
if
ican
t
f
ea
tu
r
es.
T
h
is
f
il
ter
is
al
s
o
k
n
o
wn
as
f
ea
tu
r
e
d
etec
to
r
.
I
n
(
1
)
d
escr
ib
es
co
n
v
o
lu
tio
n
o
p
e
r
atio
n
in
s
im
p
lifie
d
f
o
r
m
.
Fig
u
r
e
4
d
escr
ib
es
an
ex
am
p
le
o
f
co
n
v
o
lu
tio
n
o
p
e
r
atio
n
o
f
5
b
y
5
g
r
ay
–
s
ca
le
im
ag
e
with
3
b
y
3
r
an
d
o
m
in
itialized
weig
h
t
k
e
r
n
el
th
at
s
lid
es
with
t
h
e
in
p
u
t
im
ag
e
h
o
r
izo
n
tally
an
d
v
e
r
tically
to
p
r
o
d
u
ce
t
h
e
f
ea
tu
r
e
m
ap
.
=
∑
(
∗
+
)
(
1
)
w
h
er
e
is
o
u
tp
u
t
f
ea
tu
r
e
m
a
p
,
in
p
u
t
f
ea
tu
r
e
m
ap
,
is
s
e
t
o
f
2
D
f
ilter
s
an
d
is
tr
ain
ab
le
b
ias
p
ar
am
eter
.
Ho
wev
er
,
in
o
r
d
er
n
o
t
m
is
s
o
r
lo
s
es
s
o
m
e
v
ital
in
f
o
r
m
atio
n
at
th
e
ex
tr
em
e
ed
g
e
o
f
th
e
i
n
p
u
t
im
ag
e,
a
p
ad
d
in
g
tech
n
iq
u
e
is
ap
p
lied
.
T
h
is
is
h
ig
h
ly
im
p
o
r
tan
t
in
d
eter
m
in
in
g
b
o
r
d
e
r
s
ize
d
ata
in
r
elatio
n
to
in
p
u
t
s
ig
n
al
wh
ich
wh
en
em
p
lo
y
ed
,
d
im
en
s
io
n
o
f
th
e
in
p
u
t
im
ag
e
will
in
cr
ea
s
e
an
d
co
n
s
eq
u
en
tly
,
th
e
d
im
en
s
io
n
o
f
th
e
o
u
tp
u
t f
ea
tu
r
e
m
ap
will a
ls
o
in
cr
ea
s
e
as d
escr
ib
ed
in
Fig
u
r
e
5
.
C
la
s
s
if
ic
a
t
io
n
N
e
t
w
o
r
k
C
o
n
v
o
l
u
t
i
o
n
l
a
y
er
P
o
o
l
i
n
g
l
a
y
er
R
e
L
U
F
u
l
l
y
c
o
n
n
e
c
te
d
l
a
y
e
r
F
e
a
tu
r
e
E
x
tr
a
c
ti
o
n
N
e
tw
o
r
k
I
n
p
u
t
I
m
a
ge
Cl
a
s
s
i
f
i
c
a
ti
o
n
N
e
tw
o
r
k
Sa
g
Sw
e
ll
I
n
t
e
r
H
a
r
m
T
r
a
n
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
A
r
ev
iew
o
f c
o
n
vo
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
ks fo
r
cla
s
s
i
fyin
g
p
o
w
er q
u
a
lity p
r
o
b
lems u
s
in
g
…
(
A
d
a
mu
S
a
’
id
u
)
7
Fig
u
r
e
4
.
C
o
n
v
o
lu
tio
n
o
p
e
r
atio
n
o
f
in
p
u
t im
a
g
e
with
f
ilter
Fig
u
r
e
5
.
C
o
n
v
o
lu
tio
n
o
p
e
r
atio
n
u
s
in
g
p
ad
d
i
n
g
tech
n
iq
u
e
Act
iv
a
t
io
n
f
un
ct
io
n:
Ma
p
p
i
n
g
in
p
u
t
d
ata
to
o
u
tp
u
t
is
th
e
p
r
in
cip
al
o
p
e
r
atio
n
o
f
all
k
in
d
s
o
f
ac
tiv
atio
n
f
u
n
ctio
n
in
n
eu
r
al
n
etwo
r
k
.
T
h
at’
s
to
s
ay
ac
tiv
atio
n
f
u
n
c
tio
n
m
ak
es
d
ec
is
io
n
as
to
eith
er
o
r
n
o
t
to
f
ir
e
a
n
eu
r
o
n
in
r
e
f
er
en
ce
to
s
p
ec
if
ic
in
p
u
t
s
ig
n
al
b
y
p
r
o
d
u
ci
n
g
th
e
co
r
r
ec
t
o
u
tp
u
t.
T
h
e
A
ctiv
atio
n
f
u
n
ctio
n
in
tr
o
d
u
ce
s
n
o
n
lin
ea
r
ity
to
th
e
o
u
tp
u
t
m
atr
ix
[
4
1
]
.
I
t
also
h
as
th
e
ab
ilit
y
to
d
is
tin
g
u
is
h
d
is
tin
ctly
d
o
m
in
an
t
f
ea
tu
r
e
wh
ich
allo
ws
er
r
o
r
b
a
ck
-
p
r
o
p
ag
atio
n
to
b
e
a
p
p
lied
to
tr
ain
th
e
n
etwo
r
k
.
Sig
m
o
i
d
,
T
an
h
an
d
R
eL
U
to
g
eth
er
with
its
v
ar
ian
ts
lik
e
leak
y
R
eL
U,
n
o
is
y
R
eL
U
an
d
p
ar
am
etr
ic
lin
e
ar
u
n
it
ar
e
co
m
m
o
n
ly
u
s
e
d
ac
tiv
atio
n
f
u
n
ctio
n
s
i
n
C
NN
a
n
d
o
th
er
d
ee
p
n
e
u
r
al
n
etwo
r
k
s
.
T
h
e
m
o
s
t
c
o
m
m
o
n
ly
ap
p
lied
ac
tiv
atio
n
f
u
n
ctio
n
is
R
eL
U
b
ec
au
s
e
it
ch
a
n
g
es
al
l
th
e
v
al
u
es
o
f
in
p
u
ts
in
to
p
o
s
i
tiv
e
n
u
m
b
er
s
an
d
h
as
lo
wer
c
o
m
p
u
tatio
n
al
lo
ad
s
wh
en
co
m
p
ar
ed
with
o
th
er
ac
t
iv
atio
n
f
u
n
ctio
n
s
.
(
)
=
ma
x
(
0
,
)
(
2
)
P
o
o
lin
g
la
y
er
:
T
h
is
lay
er
in
d
icate
s
m
o
v
em
en
t
o
f
two
–
d
i
m
en
s
io
n
al
f
ilter
with
ea
ch
c
h
an
n
el
o
f
f
ea
tu
r
e
m
ap
th
er
eb
y
s
u
m
m
ar
izin
g
th
e
f
ea
t
u
r
es
ly
in
g
with
in
th
e
r
an
g
e
o
p
er
ated
b
y
th
e
f
ilter
[
4
2
]
.
Ma
j
o
r
o
p
er
atio
n
o
f
th
is
lay
er
is
th
e
s
u
b
s
am
p
lin
g
o
f
f
e
atu
r
e
m
ap
s
[
4
3
]
.
I
t
also
d
o
wn
s
izes
o
r
s
h
r
in
k
s
lar
g
e
s
ize
f
ea
tu
r
e
m
ap
to
c
r
ea
te
s
m
aller
m
ap
s
wh
ile
m
ain
tain
in
g
m
ajo
r
p
a
r
t
o
f
s
u
p
er
io
r
in
f
o
r
m
atio
n
(
o
r
f
ea
tu
r
es)
in
ea
c
h
an
d
ev
er
y
s
tep
s
o
f
th
e
p
o
o
lin
g
o
p
er
atio
n
[
4
4
]
.
M
an
y
p
o
o
lin
g
tec
h
n
iq
u
es
ar
e
av
ailab
le
f
o
r
u
tili
za
tio
n
i
n
th
e
p
o
o
lin
g
lay
e
r
s
.
T
h
e
m
o
s
t
co
m
m
o
n
ly
u
s
ed
m
eth
o
d
s
ar
e
tr
ee
p
o
o
lin
g
,
g
ated
p
o
o
l
in
g
,
av
er
a
g
e
p
o
o
lin
g
,
m
i
n
p
o
o
lin
g
,
m
ax
p
o
o
lin
g
,
g
lo
b
al
av
e
r
ag
e
p
o
o
lin
g
(
GAP
)
,
an
d
g
lo
b
al
m
ax
p
o
o
lin
g
.
T
h
e
f
am
iliar
ized
an
d
r
ep
ea
ted
l
y
ap
p
lied
p
o
o
lin
g
tech
n
iq
u
es
ar
e
m
ax
a
n
d
a
v
er
a
g
e
p
o
o
lin
g
ag
ain
in
[
4
4
]
.
Fo
r
th
e
PQ
p
r
o
b
lem
s
wav
ef
o
r
m
s
av
er
ag
e
p
o
o
lin
g
is
ap
p
lied
b
ec
au
s
e
it
is
m
o
r
e
s
en
s
itiv
e
to
n
o
is
e
s
ig
n
al
[
4
5
]
.
Ma
x
p
o
o
lin
g
ap
p
r
o
ac
h
h
as
b
etter
p
er
f
o
r
m
an
ce
ca
p
ab
ilit
y
th
an
av
er
ag
e
p
o
o
li
n
g
m
eth
o
d
.
Hen
ce
,
it
i
s
u
tili
ze
d
in
th
is
wo
r
k
.
Fig
u
r
e
6
ill
u
s
tr
ates
th
ese
t
wo
p
o
o
lin
g
o
p
er
atio
n
s
.
Fig
u
r
e
6
.
Ma
x
an
d
a
v
er
ag
e
p
o
o
lin
g
o
p
er
atio
n
Ma
n
y
at
tim
es,
th
e
o
v
er
all
p
er
f
o
r
m
an
ce
o
f
C
NN
is
r
ed
u
ce
d
d
u
e
to
p
o
o
lin
g
o
p
e
r
atio
n
.
T
h
is
in
d
icate
s
th
e
m
ajo
r
s
h
o
r
tf
all
o
f
th
is
lay
er
as
it
h
elp
s
to
e
s
tim
ate
av
aila
b
il
ity
o
f
ce
r
tain
f
ea
tu
r
e
in
th
e
in
p
u
t
d
ata
an
d
also
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
38
,
No
.
1
,
Ap
r
il
20
25
:
1
-
21
8
p
ay
s
m
o
r
e
atten
tio
n
e
n
tire
ly
o
n
f
in
d
i
n
g
c
o
r
r
ec
t
p
o
s
itio
n
o
f
t
h
at
f
ea
tu
r
e.
Hen
ce
,
C
NN
m
o
d
el
m
is
s
es
o
u
t
v
er
y
im
p
o
r
tan
t
i
n
f
o
r
m
atio
n
wh
ic
h
lead
s
to
th
e
a
p
p
licatio
n
o
f
p
ad
d
in
g
tech
n
iq
u
e
to
r
etain
th
e
in
f
o
r
m
atio
n
.
T
h
e
wh
o
le
s
er
ies o
f
co
n
v
o
l
u
tio
n
o
p
er
atio
n
,
n
o
n
lin
ea
r
ity
f
u
n
ctio
n
an
d
p
o
o
lin
g
ex
er
cise is r
ep
ea
t
ed
n
u
m
b
e
r
o
f
tim
es
to
o
b
tain
th
e
f
latten
v
ec
to
r
b
ef
o
r
e
m
o
v
in
g
to
th
e
f
in
al
la
y
er
wh
ich
is
FC
lay
er
f
o
r
class
if
icatio
n
.
F
ull
y
co
nn
ec
t
ed
la
y
er
:
Usu
al
ly
FC
lay
er
is
p
o
s
itio
n
ed
at
t
h
e
en
d
o
f
C
NN
ar
ch
itectu
r
e.
W
ith
in
th
e
lay
e
r
,
ea
c
h
n
eu
r
o
n
h
as a
lin
k
ag
e
to
all
n
eu
r
o
n
s
o
f
th
e
s
u
b
s
eq
u
en
t la
y
er
(
h
en
ce
n
am
e
FC
lay
er
)
.
I
t
is
d
escr
ib
ed
as th
e
C
N
N
class
if
ier
wh
ich
m
im
ic
th
e
b
a
s
ic
m
eth
o
d
o
f
m
u
ltip
le
-
lay
er
p
er
ce
p
tr
o
n
n
eu
r
al
n
etwo
r
k
in
f
ee
d
-
f
o
r
war
d
ANN.
I
n
p
u
t
to
FC
lay
er
co
m
es
f
r
o
m
th
e
last
p
o
o
lin
g
tech
n
i
q
u
e
o
r
co
n
v
o
l
u
tio
n
al
lay
er
.
T
h
e
in
p
u
t
to
th
e
FC
lay
er
is
in
v
ec
to
r
f
o
r
m
wh
ich
is
g
en
e
r
ated
f
r
o
m
th
e
f
ea
tu
r
e
m
ap
af
te
r
f
latten
in
g
.
T
h
e
o
u
tp
u
t
o
f
th
e
FC
lay
er
g
iv
es
th
e
f
in
al
C
NN
o
u
tp
u
t
as
il
lu
s
tr
ate
d
in
Fig
u
r
e
3
.
L
o
s
s
f
u
n
ctio
n
s
ar
e
em
p
lo
y
ed
in
th
e
o
u
tp
u
t
la
y
er
to
esti
m
ate
th
e
p
r
ed
icted
er
r
o
r
p
r
o
d
u
ce
d
ac
r
o
s
s
tr
ain
in
g
s
am
p
les in
th
e
C
NN
m
o
d
el.
I
n
(
3
)
clea
r
ly
d
escr
ib
e
s
th
is
lay
er
.
=
(
+
)
(
3
)
w
h
er
e
an
d
Z
ar
e
in
p
u
t
an
d
o
u
tp
u
t
r
esp
ec
tiv
ely
,
d
escr
ib
e
s
m
atr
ix
with
co
n
n
ec
tio
n
s
wei
g
h
ts
an
d
d
en
o
tes
b
ias ter
m
v
ec
to
r
.
3
.
2
.
L
o
s
s
f
un
ct
io
n
So
f
ar
,
th
e
p
r
ev
io
u
s
s
ec
tio
n
s
h
av
e
d
escr
ib
ed
d
is
tin
g
u
is
h
ed
la
y
er
ty
p
es
in
C
NN
ar
ch
itectu
r
e.
Mo
r
eso
,
class
if
icatio
n
is
s
u
cc
ess
f
u
l
wh
ich
r
ep
r
esen
ts
th
e
f
i
n
al
l
ay
er
o
f
th
e
C
NN
ar
ch
itectu
r
e.
T
h
e
lo
s
s
o
r
co
s
t
f
u
n
ctio
n
s
ar
e
ap
p
lied
in
th
e
o
u
tp
u
t
lay
er
to
ap
p
r
o
x
im
ate
th
e
p
r
ed
icted
er
r
o
r
p
r
o
d
u
ce
d
d
u
r
in
g
tr
ain
in
g
in
t
h
e
C
NN
m
o
d
el.
T
h
e
p
r
o
d
u
c
ed
er
r
o
r
r
e
v
ea
ls
v
ar
iatio
n
b
etwe
en
ac
tu
al
o
u
tp
u
ts
an
d
p
r
ed
ict
ed
o
n
e.
Ma
n
y
lo
s
s
f
u
n
ctio
n
s
s
u
ch
as
cr
o
s
s
–
en
tr
o
p
y
o
r
So
f
tMa
x
f
u
n
ctio
n
,
E
u
c
lid
ea
n
lo
s
s
f
u
n
ctio
n
an
d
h
in
g
e
lo
s
s
f
u
n
ctio
n
wer
e
em
p
lo
y
ed
in
d
if
f
er
en
t
C
NN
ap
p
licatio
n
s
as
ex
p
lain
e
d
in
[
4
5
]
.
So
f
tMa
x
ac
tiv
atio
n
f
u
n
cti
o
n
is
th
e
c
o
m
m
o
n
ly
u
tili
ze
d
f
u
n
ctio
n
f
o
r
m
ea
s
u
r
i
n
g
C
NN
p
er
f
o
r
m
a
n
ce
.
I
t
is
also
k
n
o
wn
as
lo
g
lo
s
s
f
u
n
ct
io
n
an
d
h
as
o
u
tp
u
t
p
r
o
b
a
b
ilit
y
o
f
p
∈
{0
,
1
}.
T
h
is
o
u
tp
u
t
lay
er
em
p
lo
y
s
th
e
So
f
tMa
x
ac
tiv
atio
n
to
g
en
er
ate
th
e
o
u
tp
u
t
with
in
th
e
p
r
o
b
a
b
il
ity
d
is
tr
ib
u
tio
n
.
3
.
3
.
CNN
re
g
ula
riza
t
io
n t
ec
hn
iq
ue
I
n
th
e
C
NN
m
o
d
el
an
aly
s
is
,
o
v
er
-
f
itti
n
g
s
ig
n
if
ies
th
e
k
e
y
is
s
u
e
in
v
o
lv
ed
in
d
ev
elo
p
in
g
g
o
o
d
g
en
er
aliza
tio
n
.
A
m
o
d
el
is
co
n
s
id
er
ed
b
e
o
v
er
-
f
itted
,
u
n
d
er
-
f
itted
o
r
ju
s
t
-
f
itted
as
illu
s
tr
ated
in
Fig
u
r
e
7
.
I
t
is
ju
s
t
-
f
itted
if
it
o
p
er
ates
g
o
o
d
o
n
tr
ain
in
g
an
d
test
in
g
d
ata.
Dif
f
er
en
t
in
tu
itiv
e
asp
ec
ts
s
u
ch
as
d
r
o
p
-
o
u
t,
d
r
o
p
-
weig
h
ts
,
d
ata
au
g
m
en
tatio
n
an
d
b
atch
n
o
r
m
aliza
tio
n
wer
e
o
c
ca
s
io
n
ally
ap
p
lied
to
h
elp
r
eg
u
lar
izatio
n
in
o
r
d
er
to
av
o
id
o
v
er
-
fi
ttin
g
.
Fig
u
r
e
7
.
Descr
ip
tio
n
o
f
o
v
er
–
f
itti
n
g
,
u
n
d
er
–
f
itti
n
g
a
n
d
ju
s
t
-
f
itted
is
s
u
es
Ap
p
licatio
n
o
f
b
atch
n
o
r
m
ali
za
tio
n
m
eth
o
d
g
u
ar
an
tees
g
o
o
d
p
er
f
o
r
m
an
ce
o
f
th
e
o
u
tp
u
t
ac
tiv
atio
n
f
u
n
ctio
n
d
u
e
to
th
e
u
n
it
Gau
s
s
ian
d
is
tr
ib
u
tio
n
b
e
h
av
io
u
r
s
.
Als
o
,
s
u
b
tr
ac
tin
g
m
ea
n
f
r
o
m
a
g
i
v
en
in
p
u
t
an
d
d
iv
id
in
g
b
y
s
tan
d
ar
d
d
ev
iatio
n
n
o
r
m
alize
s
th
e
o
u
tp
u
t
at
e
ac
h
s
tep
wh
ich
m
o
s
tly
p
r
ev
e
n
ts
th
e
p
r
o
b
lem
o
f
v
an
is
h
in
g
g
r
ad
ien
t f
r
o
m
r
is
in
g
.
3
.
4
.
CNN
o
pti
m
iza
t
io
n
t
ec
h
niq
ue
s
Op
tim
izatio
n
tech
n
iq
u
es a
r
e
C
NN
lear
n
in
g
p
r
o
ce
s
s
es.
T
wo
m
ajo
r
th
in
g
s
th
at
in
v
o
lv
ed
i
n
t
h
e
lear
n
in
g
p
r
o
ce
s
s
ar
e
th
e
lear
n
i
n
g
alg
o
r
i
th
m
s
elec
tio
n
(
o
p
tim
izer
)
an
d
th
e
ap
p
licatio
n
o
f
m
an
y
en
h
an
ce
m
en
t
tech
n
iq
u
es
(
s
u
ch
as
Ad
aDe
lta,
Ad
a
g
r
ad
,
an
d
m
o
m
en
t
u
m
)
.
T
h
e
la
tter
alo
n
g
with
t
h
e
lear
n
i
n
g
al
g
o
r
ith
m
im
p
r
o
v
es
th
e
o
u
tp
u
t
o
f
th
e
tr
ain
i
n
g
r
esu
lt.
T
h
e
n
etwo
r
k
p
ar
a
m
eter
s
s
h
all
alwa
y
s
b
e
u
p
d
ated
th
r
o
u
g
h
e
v
er
y
ep
o
c
h
s
s
o
as
to
less
en
s
th
e
er
r
o
r
[
4
6
]
,
[
4
7
]
.
I
n
o
r
d
er
to
r
e
d
u
ce
th
e
im
p
ac
t
o
f
th
e
lear
n
in
g
er
r
o
r
,
th
e
al
g
o
r
ith
m
r
ep
ea
te
d
ly
u
p
d
ates
th
e
n
etwo
r
k
p
ar
am
e
ter
s
at
ea
ch
an
d
ev
er
y
iter
a
tio
n
[
4
8
]
.
Mo
r
eo
v
e
r
,
in
o
r
d
er
to
u
p
g
r
a
d
e
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
A
r
ev
iew
o
f c
o
n
vo
lu
tio
n
a
l n
e
u
r
a
l n
etw
o
r
ks fo
r
cla
s
s
i
fyin
g
p
o
w
er q
u
a
lity p
r
o
b
lems u
s
in
g
…
(
A
d
a
mu
S
a
’
id
u
)
9
p
ar
am
eter
s
r
ig
h
tly
,
th
e
r
e
is
n
ee
d
to
esti
m
ate
th
e
g
r
ad
ien
t
f
u
n
ctio
n
(
s
lo
p
e)
b
y
u
tili
zin
g
f
i
r
s
t
-
o
r
d
er
d
er
iv
ativ
e
with
r
esp
ec
t
to
t
h
e
n
etwo
r
k
ch
ar
a
cter
is
tics
.
Ag
ain
,
p
a
r
am
eter
s
ar
e
u
p
g
r
ad
e
d
in
th
e
o
th
er
d
ir
ec
tio
n
o
f
th
e
g
r
ad
ien
t
to
less
en
th
e
lear
n
i
n
g
er
r
o
r
.
Var
io
u
s
g
r
ad
ien
t
b
ased
lear
n
in
g
alg
o
r
ith
m
s
s
u
c
h
as
b
atch
g
r
a
d
ien
t
d
escen
t
(
B
GD)
,
s
to
ch
asti
c
g
r
a
d
ien
t
d
escen
t
(
SGD)
,
m
in
i
–
b
atch
g
r
ad
ien
t
d
escen
t
(
MBGD)
,
m
o
m
en
tu
m
a
n
d
ad
ap
tiv
e
m
o
m
en
t
esti
m
atio
n
(
Ad
am
)
ar
e
av
ailab
le
an
d
co
m
m
o
n
ly
em
p
lo
y
ed
as h
ig
h
lig
h
te
d
in
[
4
9
]
.
Ad
a
m
is
a
lear
n
in
g
ap
p
r
o
ac
h
d
esig
n
e
d
p
a
r
ticu
lar
ly
f
o
r
tr
ain
in
g
d
ee
p
n
eu
r
al
n
etwo
r
k
s
.
3
.
5
.
CNN
m
o
del
a
rc
hite
ct
ur
e
I
n
o
r
d
er
to
o
b
tain
g
o
o
d
r
esu
lt,
s
elec
tio
n
o
f
C
NN
ar
ch
ite
ctu
r
e
is
an
o
th
er
v
er
y
cr
itical
is
s
u
e
f
o
r
en
h
an
cin
g
p
e
r
f
o
r
m
an
ce
o
f
v
ar
io
u
s
C
NN
d
esig
n
s
.
Ma
n
y
ad
ju
s
tm
en
ts
h
av
e
b
ee
n
p
u
t
in
p
lace
f
o
r
C
NN
ar
ch
itectu
r
es
r
ec
e
n
tly
.
Sp
ec
if
i
ca
lly
,
th
e
n
ew
m
o
d
if
icatio
n
s
i
n
C
NN
ar
ch
itectu
r
es
wer
e
ac
h
iev
ed
b
ased
o
n
th
e
u
tili
za
tio
n
o
f
n
etwo
r
k
d
ep
th
.
T
ab
le
5
p
r
esen
ts
b
r
ief
o
v
er
v
i
ew
o
f
m
o
s
t
p
o
p
u
lar
C
NN
ar
ch
itectu
r
es,
s
tar
tin
g
f
r
o
m
Alex
Net
m
o
d
el
in
2
0
1
2
an
d
e
n
d
in
g
at
th
e
h
ig
h
r
e
s
o
lu
tio
n
(
HR
)
m
o
d
el
in
2
0
2
0
.
C
o
n
s
id
er
in
g
th
e
ar
ch
itectu
r
al
f
ea
tu
r
es
(
s
u
ch
as
i
n
p
u
t
s
ize,
d
ep
th
,
a
n
d
r
o
b
u
s
t
n
ess
)
is
th
e
m
ain
asp
ec
t
in
as
s
is
tin
g
en
g
in
ee
r
s
o
r
r
esear
ch
er
s
to
ch
o
o
s
e
th
e
m
o
s
t
ap
p
r
o
p
r
iate
ar
ch
itectu
r
e
f
o
r
th
eir
p
r
o
p
o
s
e
ap
p
licatio
n
.
T
ab
le
5
.
Ov
e
r
v
iew
o
f
m
o
s
t p
o
p
u
lar
C
NN
ar
ch
itectu
r
es
[
1
7
]
M
o
d
e
l
D
e
p
t
h
D
a
t
a
s
e
t
Er
r
o
r
r
a
t
e
I
n
p
u
t
si
z
e
Y
e
a
r
A
l
e
x
N
e
t
8
I
mag
e
N
e
t
1
6
.
4
2
2
7
x
2
2
7
x
3
2
0
1
2
VGG
1
6
,
1
9
I
mag
e
N
e
t
7
.
3
2
2
4
x
2
2
4
x
3
2
0
1
4
G
o
o
g
L
e
N
e
t
22
I
mag
e
N
e
t
6
.
7
2
2
4
x
2
2
4
x
3
2
0
1
5
R
e
sN
e
t
1
5
2
I
mag
e
N
e
t
3
.
5
7
2
2
4
x
2
2
4
x
3
2
0
1
6
D
e
n
seN
e
t
1
2
1
,
1
6
0
,
2
0
1
C
I
F
A
R
1
0
,
C
I
F
A
R
1
0
0
,
I
mag
e
N
e
t
3
.
4
6
,
1
7
.
1
8
,
5
.
5
4
2
2
4
X
2
2
4
X
3
2
0
1
7
M
o
b
i
l
e
N
e
t
‑
v
2
53
I
mag
e
N
e
t
-
2
2
4
x
2
2
4
x
3
2
0
1
8
H
R
N
e
t
V
2
-
I
mag
e
N
e
t
5
.
4
2
2
4
x
2
2
4
x
3
2
0
2
0
4.
AP
P
L
I
CA
T
I
O
N
O
F
CNN
I
N
CL
AS
SI
F
Y
I
NG
P
Q
P
RO
B
L
E
M
S
T
h
is
s
ec
tio
n
p
o
r
tr
ay
s
a
liter
atu
r
e
s
u
r
v
ey
o
n
th
e
class
if
icatio
n
o
f
PQ
p
r
o
b
lem
s
u
s
in
g
DL
a
p
p
r
o
ac
h
es.
I
n
a
s
tu
d
y
o
f
DL
ap
p
r
o
ac
h
f
o
r
d
etec
tio
n
an
d
ca
te
g
o
r
izatio
n
o
f
PQDs
with
win
d
o
wed
s
ig
n
als
d
escr
ib
ed
u
s
in
g
v
o
ltag
e
s
ig
n
als
o
n
ly
is
co
n
d
u
c
ted
in
[
1
]
.
T
h
e
a
p
p
r
o
ac
h
u
s
es
o
v
er
lap
p
e
d
win
d
o
wed
s
ig
n
als with
d
if
f
er
en
t
SNR
wh
ich
p
er
f
o
r
m
s
s
atis
f
ac
to
r
ily
with
a
v
alu
e
a
b
o
v
e
9
7
%
ac
c
u
r
ac
y
,
ev
e
n
th
o
u
g
h
th
e
o
p
er
ati
o
n
o
f
th
e
class
if
ier
r
ed
u
ce
s
as
th
e
n
o
is
e
in
ter
f
er
en
ce
in
th
e
in
p
u
t
im
p
r
o
v
es.
T
h
e
u
s
e
o
f
W
ig
n
er
–
Ville
d
is
tr
ib
u
tio
n
v
ia
d
ee
p
co
n
v
o
l
u
tio
n
al
n
e
u
r
al
n
etwo
r
k
s
(
W
VD
–
C
NN)
f
o
r
id
en
tific
atio
n
o
f
PQDs
is
p
r
o
p
o
s
ed
b
y
[
9
]
wh
er
e
b
y
th
e
W
VD
wa
s
em
p
lo
y
ed
to
tr
an
s
p
o
r
t
1
D
v
o
ltag
e
d
is
tu
r
b
a
n
ce
s
ig
n
als
in
to
2
D
v
o
ltag
e
im
ag
e
f
iles
an
d
C
NN
b
ased
m
o
d
el
was
d
ev
elo
p
e
d
an
d
p
r
o
ce
s
s
ed
u
s
in
g
im
ag
e
d
a
ta
to
o
b
tain
o
p
tim
ized
p
ar
am
eter
s
f
o
r
PQDs
class
if
icatio
n
.
Nin
e
ty
p
es
o
f
s
y
n
th
etic
s
ig
n
als
an
d
th
r
ee
r
ea
l
wo
r
ld
m
ea
s
u
r
em
en
ts
wer
e
e
m
p
lo
y
ed
to
test
th
e
m
o
d
el
wh
ic
h
g
i
v
es
b
est
class
if
icatio
n
ac
cu
r
ac
y
o
f
9
9
.
6
7
%.
T
h
e
r
esu
lt
o
b
tai
n
ed
en
s
u
r
es
t
h
e
ef
f
ec
tiv
e
n
ess
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el.
R
esear
c
h
o
n
n
o
v
el
tech
n
iq
u
e
f
o
r
m
u
ltip
les
PQ
d
is
tu
r
b
an
ce
s
class
if
icatio
n
em
p
lo
y
in
g
m
u
lti
–
task
C
NN
(
MT
-
C
N
N)
ap
p
r
o
ac
h
was
d
ev
elo
p
ed
t
o
ac
tu
alize
m
u
lti
–
lab
el
class
if
icatio
n
o
f
m
u
ltip
le
PQDs
[
1
1
]
.
T
h
e
m
eth
o
d
ex
tr
a
cts
m
o
r
e
s
ig
n
if
ican
t
f
ea
tu
r
es
an
d
y
ield
s
b
etter
r
ec
o
g
n
itio
n
r
a
te
o
f
9
4
.
6
3
%
an
d
it
h
as
v
er
y
s
tr
o
n
g
ca
p
ab
ilit
y
to
r
esis
t
o
v
er
f
itti
n
g
is
s
u
e.
T
h
e
tech
n
iq
u
e
c
o
u
ld
lar
g
el
y
im
p
r
o
v
e
ac
cu
r
ac
y
r
ate
f
o
r
r
ep
r
esen
tin
g
m
u
ltip
le
PQDs
u
n
d
er
d
i
f
f
er
en
t sig
n
al
t
o
n
o
is
e
r
atio
co
n
d
itio
n
s
.
An
o
th
er
r
esear
ch
o
n
PQDs
m
o
n
ito
r
in
g
a
n
d
class
if
icatio
n
em
p
lo
y
in
g
im
p
r
o
v
ed
p
r
in
cip
al
co
m
p
o
n
en
t
an
aly
s
is
(
PC
A)
an
d
C
NN
f
o
r
win
d
g
r
id
d
is
tr
ib
u
tio
n
s
y
s
tem
s
was
p
r
esen
ted
wh
er
eb
y
th
e
s
tatis
tica
l
f
ea
tu
r
es
wer
e
ex
tr
ac
ted
v
ia
im
p
r
o
v
ed
P
C
A
(
I
PC
A)
wh
er
ea
s
f
ea
tu
r
es
lik
e
m
ea
n
,
s
tan
d
ar
d
d
e
v
i
atio
n
,
en
e
r
g
y
wer
e
ex
tr
ac
ted
u
s
in
g
1
D
–
C
NN.
T
h
e
m
eth
o
d
class
if
ies
PQD
s
w
ith
m
ax
im
u
m
class
if
icatio
n
a
cc
u
r
ac
y
o
f
9
9
.
9
2
%
wh
ich
was
test
ed
with
n
o
is
e
an
d
n
o
is
eless
en
v
ir
o
n
m
e
n
t
[
1
2
]
.
Als
o
,
a
s
tu
d
y
o
f
PQ
d
is
tu
r
b
an
ce
class
if
icatio
n
in
co
r
p
o
r
atin
g
co
m
p
r
ess
ed
s
en
s
in
g
an
d
d
ee
p
co
n
v
o
l
u
tio
n
a
l
n
eu
r
al
n
etwo
r
k
(
C
S
–
DC
NN)
was
s
u
g
g
ested
wh
er
eb
y
d
ata
is
co
m
p
r
ess
ed
to
r
ed
u
ce
t
h
e
r
eq
u
ir
em
en
t
f
o
r
ac
q
u
is
itio
n
d
ev
ice
m
em
o
r
y
wh
ich
in
cr
ea
s
e
th
e
tr
an
s
m
is
s
io
n
r
ate
[
1
5
]
.
T
h
e
d
ee
p
C
NN
was
u
s
ed
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
class
if
icatio
n
with
o
u
t
an
y
d
elay
an
d
d
ata
p
r
e
–
p
r
o
ce
s
s
in
g
o
p
e
r
atio
n
.
T
h
e
m
o
d
el
in
d
icate
d
g
o
o
d
class
if
icatio
n
p
er
f
o
r
m
a
n
c
e
in
n
o
is
e
d
ata
wit
h
ac
cu
r
ac
y
o
f
9
9
.
5
0
% a
n
d
a
lo
s
s
o
f
0
.
0
2
.
An
o
th
er
w
o
r
k
in
th
e
s
am
e
y
e
ar
b
y
[
1
7
]
p
r
o
p
o
s
ed
id
en
tific
atio
n
o
f
PQ
d
is
tu
r
b
a
n
ce
s
(
PQ
Ds)
u
s
in
g
p
h
ase
s
p
ac
e
r
ec
o
g
n
itio
n
with
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
PS
R
–
C
NN)
.
T
h
e
PS
R
tr
a
n
s
f
o
r
m
1
D
v
o
ltag
e
s
ig
n
al
in
to
2
D
v
o
ltag
e
im
a
g
e
f
ile
an
d
th
e
C
NN
d
o
es
th
e
class
if
icatio
n
au
to
m
atica
lly
wit
h
h
ig
h
ac
c
u
r
ac
y
o
f
9
9
.
8
0
%.
T
h
e
p
er
f
o
r
m
an
c
e
s
h
o
wed
th
at
th
e
m
o
d
el
is
ab
le
t
o
p
r
o
d
u
ce
m
o
r
e
im
p
r
o
v
es
r
esu
lts
with
o
u
t
h
u
m
an
in
v
o
lv
em
e
n
t
wh
en
c
o
m
p
ar
e
d
to
ex
is
tin
g
m
eth
o
d
s
.
E
k
ici
et
a
l.
[
2
0
]
in
tr
o
d
u
ce
d
o
p
tim
ized
B
ay
esian
C
NN
f
o
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
38
,
No
.
1
,
Ap
r
il
20
25
:
1
-
21
10
PQ
p
r
o
b
lem
s
class
if
icatio
n
.
T
h
e
r
esu
lt
p
r
o
v
ed
th
at
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
s
u
p
er
s
ed
ed
s
o
m
e
ML
alg
o
r
ith
m
s
s
u
ch
as
d
ec
is
io
n
t
r
ee
s
an
d
r
a
n
d
o
m
f
o
r
ests
r
eg
a
r
d
in
g
ac
cu
r
ac
y
.
T
h
e
wh
o
le
ac
cu
r
ac
y
atta
in
ed
b
y
th
e
m
eth
o
d
was
r
ep
o
r
ted
to
b
e
9
9
.
8
0
%.
T
h
e
ap
p
r
o
ac
h
was
ev
alu
ate
d
u
s
in
g
p
u
b
licly
av
ailab
le
PQ d
ataset.
R
o
d
r
ig
u
ez
et
a
l.
[
3
8
]
p
r
esen
ted
an
o
th
er
s
tu
d
y
o
n
PQ
d
is
tu
r
b
an
ce
class
if
icat
io
n
v
ia
d
ee
p
co
n
v
o
lu
ti
o
n
al
a
u
to
–
en
c
o
d
er
s
a
n
d
s
tack
ed
L
STM
R
NNs.
I
n
t
h
is
r
esear
ch
,
a
s
tr
o
n
g
al
g
o
r
ith
m
t
h
at
co
n
tain
s
d
ee
p
C
AE
an
d
s
tack
ed
L
STM
R
NNs
was
u
s
ed
.
Hig
h
class
if
icatio
n
ac
cu
r
ac
y
o
f
9
8
.
7
%
was
r
ea
c
h
ed
an
d
tr
ain
in
g
tim
e
was
r
ed
u
ce
d
f
r
o
m
5
s
ec
o
n
d
s
to
3
s
ec
o
n
d
s
p
er
tr
ain
in
g
iter
a
tio
n
.
T
h
e
u
s
ag
e
o
f
DL
with
2
D
wav
elet
s
ca
lo
g
r
am
s
f
o
r
PQD
class
if
ica
tio
n
was
d
escr
ib
ed
b
y
[
4
5
]
.
C
o
n
tin
u
o
u
s
wav
elet
tr
an
s
f
o
r
m
(
C
W
T
)
was
u
s
ed
to
p
r
o
d
u
ce
s
ca
lo
g
r
am
o
f
2
D
th
at
ex
p
r
ess
s
ig
n
al
p
atter
n
o
f
PQ
ev
en
t
th
r
o
u
g
h
tim
e
–
f
r
eq
u
e
n
cy
r
ep
r
es
en
tatio
n
.
C
NN
is
em
p
lo
y
ed
to
ca
teg
o
r
ize
th
e
d
ata
in
ac
co
r
d
a
n
ce
to
th
e
im
a
g
e
d
is
tu
r
b
an
c
e
with
h
ig
h
ac
cu
r
ac
y
o
f
9
7
.
6
7
%
in
n
o
is
eless
s
ig
n
als.
Xu
e
et
a
l.
[
4
7
]
d
ev
elo
p
e
d
a
d
ee
p
C
NN
with
s
p
ec
tr
o
g
r
am
u
s
in
g
m
icr
o
g
r
id
f
o
r
PQ
d
is
tu
r
b
a
n
ce
class
if
icatio
n
.
Sp
ec
tr
o
g
r
am
tech
n
iq
u
e
was
u
s
ed
to
r
estru
ctu
r
e
th
e
PQDs
s
ig
n
als
an
d
th
e
C
NN
u
s
ed
f
o
r
clas
s
if
icatio
n
.
T
h
e
m
eth
o
d
d
iv
id
es
th
e
PQDs
in
to
d
if
f
er
en
t
s
em
an
tic
f
ea
tu
r
es
f
o
r
d
etec
tin
g
s
in
g
le
an
d
m
u
ltip
les
s
ig
n
als
o
v
er
co
n
tr
ar
y
tim
e
s
ca
les
in
a
s
am
p
le.
T
h
e
m
eth
o
d
p
r
o
v
ed
to
b
e
r
o
b
u
s
t
to
n
o
is
e
with
h
ig
h
class
if
icatio
n
ac
cu
r
ac
y
o
f
9
9
.
6
0
%
b
u
t
m
er
g
in
g
o
f
C
NN
an
d
R
NN
will im
p
r
o
v
e
th
e
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
.
R
esear
ch
co
n
d
u
cte
d
b
y
[
4
8
]
d
escr
ib
ed
h
o
w
s
ig
n
al
p
r
o
ce
s
s
in
g
an
d
DL
tech
n
iq
u
es
wer
e
u
s
ed
f
o
r
PQ
p
r
o
b
lem
s
m
an
a
g
in
g
a
n
d
class
if
icatio
n
.
T
h
e
s
tu
d
y
in
tr
o
d
u
ce
s
in
n
o
v
ativ
e
m
eth
o
d
em
p
l
o
y
in
g
co
m
p
r
ess
iv
e
s
en
s
in
g
(
C
S),
s
in
g
u
lar
s
p
ec
tr
u
m
an
aly
s
is
(
SS
A)
,
WT
an
d
DNN
f
o
r
m
o
n
ito
r
in
g
class
if
icatio
n
o
f
s
in
g
le
an
d
co
m
b
in
ed
PQ
d
is
tu
r
b
a
n
ce
s
(
PQDs
)
.
T
h
e
SS
A
–
C
S
–
D
NN
alg
o
r
ith
m
p
r
o
v
es
to
b
e
b
et
way
o
f
PQDs
class
if
icatio
n
with
h
ig
h
ac
c
u
r
ac
y
o
f
9
9
.
8
5
%.
Ah
ajjam
et
a
l.
[
5
0
]
p
r
esen
ted
r
esear
ch
o
n
elec
tr
ical
PQDs
class
if
icatio
n
u
s
in
g
tem
p
o
r
al
–
s
p
ec
tr
al
im
ag
es
with
d
ee
p
C
NNs.
T
h
e
s
tu
d
y
d
escr
ib
ed
n
ew
ap
p
r
o
ac
h
o
f
PQD
d
etec
tio
n
an
d
class
if
icatio
n
te
ch
n
iq
u
e
in
v
o
l
v
in
g
f
u
s
in
g
tem
p
o
r
al
with
s
p
ec
tr
al
im
ag
es
an
d
d
ee
p
C
NNs
(
FTSI
–
C
NN)
.
T
h
e
tech
n
iq
u
e
r
ed
u
ce
s
f
ea
tu
r
e
d
im
en
s
io
n
wh
ile
r
etain
in
g
tim
e
–
f
r
eq
u
e
n
cy
in
f
o
r
m
atio
n
t
o
attain
g
o
o
d
p
er
f
o
r
m
a
n
ce
ac
c
u
r
ac
y
o
f
9
9
.
6
7
%.
Go
o
d
FTSI
d
esig
n
will
f
u
r
th
er
im
p
r
o
v
e
ac
cu
r
ac
y
wh
ile
m
ain
tain
i
n
g
its
lo
w
co
m
p
lex
ity
.
Mo
h
am
m
ad
i
et
a
l.
[
5
1
]
co
n
d
u
cted
a
r
es
ea
r
ch
o
f
PQ
d
is
tu
r
b
an
ce
s
class
if
icatio
n
v
ia
f
u
ll
–
co
n
v
o
l
u
tio
n
al
Siam
ese
n
etwo
r
k
an
d
k
–
n
ea
r
est
n
eig
h
b
o
r
.
I
n
th
e
s
tu
d
y
,
co
m
b
in
ed
alg
o
r
it
h
m
o
f
k
–
NN
an
d
f
u
lly
–
co
n
v
o
l
u
tio
n
al
Siam
ese
wer
e
p
u
t
in
p
lace
t
o
class
if
y
PQDs
b
y
lear
n
in
g
s
m
all
s
am
p
les
with
h
ig
h
er
th
a
n
8
0
%
ac
cu
r
ac
y
.
Mu
ltit
u
d
e
c
o
n
v
o
lu
tio
n
al
lay
e
r
s
an
d
co
n
n
ec
tio
n
lay
er
s
ar
e
th
er
e
to
d
ev
el
o
p
Siam
ese
n
etwo
r
k
an
d
o
u
tp
u
t
r
ea
cti
o
n
ju
d
g
e’
s
ca
teg
o
r
y
o
f
th
e
s
ig
n
al.
T
o
en
s
u
r
e
h
ig
h
en
o
u
g
h
ac
c
u
r
ac
y
,
t
h
e
s
am
p
lin
g
f
r
e
q
u
en
c
y
s
h
o
u
ld
b
e
m
o
r
e
th
a
n
1
2
7
5
Hz.
L
iu
et
a
l.
[
5
2
]
p
r
esen
ted
a
co
m
p
lex
PQ
d
is
tu
r
b
an
ce
class
if
icatio
n
u
s
in
g
cu
r
v
elet
tr
an
s
f
o
r
m
a
n
d
DL
tech
n
iq
u
e.
I
n
th
is
n
o
v
el
ap
p
r
o
ac
h
,
SS
A,
cu
r
v
elet
tr
an
s
f
o
r
m
(
C
T
)
an
d
d
ee
p
C
NNs
wer
e
ap
p
lied
to
s
en
s
e
an
d
ca
t
eg
o
r
ize
PQDs
.
Go
o
d
class
if
ic
atio
n
r
esu
lt
was
o
b
tain
ed
an
d
co
m
p
ar
ed
t
o
SVM
an
d
o
th
e
r
class
if
ier
s
in
wh
ich
th
e
cu
r
r
e
n
t te
ch
n
iq
u
e
s
u
p
er
s
ed
ed
in
ter
m
s
o
f
ac
cu
r
ac
y
with
9
9
.
5
2
%.
An
o
th
er
n
ew
ap
p
r
o
ac
h
f
o
r
PQDs
cla
s
s
if
icatio
n
th
r
o
u
g
h
s
p
ar
s
e
au
to
en
co
d
er
s
(
SAE)
b
ased
o
n
DNN
was p
r
esen
ted
to
ex
tr
ac
t f
ea
tu
r
es a
n
d
class
if
y
P
QDs [
5
3
]
.
A
g
o
o
d
f
ea
t
u
r
e
ex
tr
ac
tio
n
p
er
f
o
r
m
an
ce
was r
ea
lized
v
ia
SAE
an
d
h
ig
h
class
if
icati
o
n
ac
cu
r
ac
y
o
f
9
3
%
was
attain
ed
u
s
in
g
DNN
.
Ma
n
an
et
a
l.
[
5
4
]
d
ev
el
o
p
e
d
an
o
th
er
DL
a
p
p
r
o
ac
h
in
th
e
f
ield
o
f
PQ
d
is
tu
r
b
a
n
ce
class
if
icatio
n
wh
er
e
CWT
was
u
s
ed
to
g
e
n
er
ate
th
e
co
ef
f
icien
t
m
atr
ix
a
n
d
later
t
h
e
co
ef
f
icien
ts
wer
e
ch
a
n
g
ed
to
im
ag
e
f
ile
u
s
in
g
f
ea
tu
r
e
m
atr
ix
an
d
g
iv
en
to
C
NN
as
in
p
u
t
f
o
r
c
lass
if
icatio
n
.
Hig
h
er
class
if
icatio
n
ac
cu
r
a
cy
o
f
9
9
.
6
0
%
an
d
n
o
is
e
im
m
u
n
ity
wer
e
r
ec
o
r
d
ed
in
th
is
tech
n
iq
u
e
.
Mish
r
a
et
a
l.
[
5
5
]
d
escr
ib
ed
th
e
p
o
w
er
o
f
t
h
eir
ap
p
r
o
ac
h
in
c
o
r
r
ec
tly
d
etec
tin
g
a
n
d
class
if
y
in
g
m
u
ltip
le
PQDs
u
s
i
n
g
tem
p
o
r
al
DL
.
E
n
co
d
e
–
d
e
co
d
e
tem
p
o
r
al
C
NN
(
E
DT
-
C
NN)
tech
n
iq
u
e
th
at
m
er
g
es
f
ea
tu
r
e
s
elec
tio
n
wi
th
class
if
icatio
n
in
a
s
in
g
le
b
lo
ck
was
em
p
lo
y
ed
.
A
g
o
o
d
class
if
icatio
n
p
er
f
o
r
m
an
ce
o
f
9
9
.
5
2
%
was
r
ea
lized
in
n
o
is
y
d
ata.
R
am
alin
g
ap
p
a
an
d
Ma
n
ju
n
at
h
a
[
5
6
]
d
ev
elo
p
ed
h
y
b
r
id
ap
p
r
o
ac
h
with
c
o
m
p
lex
wav
e
let
p
h
aso
r
m
o
d
el
a
n
d
cu
s
to
m
ized
C
NN
to
r
ep
r
esen
t
PQ
is
s
u
es.
T
h
e
d
ataset
in
v
o
lv
ed
in
th
e
s
tu
d
y
wer
e
c
o
llected
f
r
o
m
t
h
e
p
o
wer
g
r
id
in
I
n
d
ia.
T
h
e
m
eth
o
d
attain
e
d
g
o
o
d
ac
c
u
r
ac
y
o
f
9
9
.
3
3
%
in
class
if
y
in
g
PQ
p
r
o
b
lem
s
.
Mo
h
an
et
a
l.
[
5
7
]
p
r
o
p
o
s
ed
an
o
th
e
r
ar
ch
itectu
r
e
ca
lled
Dee
p
Po
wer
f
o
r
PQ
d
is
tu
r
b
an
ce
s
class
if
icatio
n
.
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
was
e
v
alu
ated
o
n
t
h
e
p
u
b
licly
av
ail
ab
l
e
UC
I
E
lectr
ic
PQ
d
ataset
an
d
th
e
r
esu
lt
s
h
o
wed
r
em
ar
k
ab
le
ac
h
iev
e
m
en
t
with
h
ig
h
ac
cu
r
ac
y
o
f
9
9
.
7
1
%.
Sah
an
i
an
d
Dash
[
5
8
]
r
ec
o
m
m
e
n
d
ed
th
e
ap
p
lic
atio
n
o
f
FP
GA
b
ased
d
ee
p
C
NN
f
o
r
PQ
ev
en
t
id
en
tific
atio
n
.
T
h
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
was
ev
alu
ated
u
s
in
g
s
y
n
th
esized
an
d
ex
p
e
r
im
en
tal
d
ata
co
llected
f
r
o
m
p
r
o
ce
s
s
ad
ap
tiv
e
VM
D
d
ata.
Ov
er
all
ac
cu
r
ac
y
o
b
tain
e
d
f
r
o
m
th
is
m
eth
o
d
is
r
ep
o
r
ted
a
s
9
6
.
7
5
%.
Qiu
et
a
l.
[
5
9
]
d
e
v
elo
p
ed
a
d
if
f
e
r
en
t
m
eth
o
d
u
s
in
g
m
u
ltifu
s
io
n
C
NN
b
ased
au
t
o
m
atic
class
if
icatio
n
f
r
am
ewo
r
k
o
f
c
o
m
p
lex
PQ
d
is
t
u
r
b
an
ce
s
.
T
h
e
m
eth
o
d
f
u
s
ed
in
f
o
r
m
atio
n
b
as
ed
o
n
th
e
PQ
e
v
en
t’
s
tim
e
d
o
m
ain
an
d
f
r
eq
u
en
cy
d
o
m
ain
f
ea
tu
r
es.
T
h
e
m
eth
o
d
is
ev
alu
ated
o
n
a
d
ataset
c
o
llected
f
r
o
m
PQ
m
o
n
ito
r
in
g
s
y
s
tem
an
d
th
e
r
esu
lt
s
h
o
wed
r
em
ar
k
ab
le
ac
h
iev
em
en
t
with
o
v
er
all
ac
c
u
r
ac
y
o
f
9
8
.
4
6
%.
Yig
it
et
a
l.
[
6
0
]
p
r
esen
te
d
an
au
to
m
atic
d
etec
tio
n
m
o
d
el
o
f
PQD
u
s
in
g
C
N
N
s
tr
u
ctu
r
e
w
ith
g
ated
r
ec
u
r
r
e
n
t
u
n
it.
A
m
atr
ix
o
f
PQD
s
ig
n
als
wa
s
o
b
tain
ed
u
s
in
g
STFT
wh
ich
co
n
tain
s
r
ec
o
n
s
tr
u
ctio
n
o
f
s
ig
n
al
in
tim
e
an
d
f
r
e
q
u
en
cy
d
o
m
ain
w
h
ich
ar
e
g
o
o
d
in
p
u
t
f
o
r
C
N
N.
C
h
ar
ac
ter
is
tics
ar
e
au
to
m
atica
lly
d
is
till
ed
u
s
in
g
C
NN
with
o
u
t
an
y
p
r
e
p
r
o
ce
s
s
in
g
tec
h
n
iq
u
e
an
d
th
e
GR
U
is
em
p
lo
y
ed
to
class
if
y
f
ea
tu
r
es.
T
h
e
f
u
n
ctio
n
o
f
th
e
m
eth
o
d
is
te
s
ted
o
n
a
d
ataset
wh
ich
co
n
tain
s
a
to
tal
o
f
Evaluation Warning : The document was created with Spire.PDF for Python.