T
E
L
K
O
M
N
I
K
A
T
elec
o
m
m
un
ica
t
io
n,
Co
m
pu
t
ing
,
E
lect
ro
nics
a
nd
Co
ntr
o
l
Vo
l.
19
,
No
.
3
,
J
u
n
e
2
0
2
1
,
p
p
.
893
~
90
1
I
SS
N:
1
6
9
3
-
6
9
3
0
,
ac
cr
ed
ited
First Gr
ad
e
b
y
Kem
en
r
is
tek
d
i
k
ti,
Dec
r
ee
No
: 2
1
/E/KPT
/2
0
1
8
DOI
: 1
0
.
1
2
9
2
8
/TE
L
KOM
NI
K
A.
v
1
9
i3
.
1
8
7
7
1
893
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//jo
u
r
n
a
l.u
a
d
.
a
c.
id
/in
d
ex
.
p
h
p
/TELK
OM
N
I
K
A
Deep
fingerprin
t
cla
ss
ificatio
n net
wo
rk
Abdu
l
s
a
t
t
a
r
M
.
I
bra
him
,
Abdu
lra
hm
a
n K
.
E
esee
,
Ra
id R
a
f
i O
ma
r
Al
-
Nim
a
De
p
a
rtme
n
t
o
f
M
e
d
ica
l
In
stru
m
e
n
tatio
n
Tec
h
n
o
lo
g
y
En
g
in
e
e
rin
g
,
No
rth
e
rn
Tec
h
n
ica
l
Un
iv
e
rsity
,
Ir
a
q
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
n
5
,
2
0
2
0
R
ev
is
ed
No
v
1
0
,
2
0
2
0
Acc
ep
ted
No
v
2
5
,
2
0
2
0
F
in
g
e
r
p
rin
t
is
o
n
e
o
f
th
e
m
o
st
w
e
ll
-
k
n
o
wn
b
i
o
m
e
tri
c
s
th
a
t
h
a
s
b
e
e
n
u
se
d
fo
r
p
e
rso
n
a
l
re
c
o
g
n
it
i
o
n
.
Ho
we
v
e
r,
fa
k
e
d
fin
g
e
r
p
rin
ts
h
a
v
e
b
e
c
o
m
e
th
e
m
a
jo
r
e
n
e
m
y
wh
e
re
th
e
y
th
re
a
t
th
e
se
c
u
r
it
y
o
f
t
h
is
b
i
o
m
e
tri
c
.
Th
is
p
a
p
e
r
p
r
o
p
o
se
s
a
n
e
fficie
n
t
d
e
e
p
fi
n
g
e
r
p
rin
t
c
las
sifica
ti
o
n
n
e
two
rk
(DFCN)
m
o
d
e
l
t
o
a
c
h
iev
e
a
c
c
u
ra
te
p
e
rfo
rm
a
n
c
e
s
o
f
c
las
sify
in
g
b
e
twe
e
n
re
a
l
a
n
d
fa
k
e
fi
n
g
e
rp
rin
ts.
Th
is
m
o
d
e
l
h
a
s
e
x
te
n
siv
e
ly
e
v
a
lu
a
ted
o
r
e
x
a
m
in
e
d
p
a
ra
m
e
ters
.
To
tal
o
f
5
1
2
ima
g
e
s
fro
m
th
e
ATVS
-
F
F
p
_
DB
d
a
tas
e
t
a
re
e
m
p
lo
y
e
d
.
T
h
e
p
ro
p
o
se
d
DFC
N
a
c
h
iev
e
d
h
ig
h
c
las
sifica
ti
o
n
p
e
rfo
rm
a
n
c
e
o
f
9
9
.
2
2
%
,
wh
e
re
fin
g
e
rp
ri
n
t
i
m
a
g
e
s
a
re
su
c
c
e
ss
fu
ll
y
c
las
sified
in
to
th
e
ir
t
wo
c
a
teg
o
ries
.
M
o
re
o
v
e
r,
c
o
m
p
a
r
is
o
n
s
wit
h
sta
te
-
of
-
a
rt
a
p
p
ro
a
c
h
e
s a
re
p
r
o
v
id
e
d
.
K
ey
w
o
r
d
s
:
C
las
s
if
icatio
n
Dee
p
lear
n
in
g
Fin
g
er
p
r
in
t
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
:
Ab
d
u
lr
ah
m
a
n
K.
E
esee
Dep
ar
tm
en
t o
f
Me
d
ical
I
n
s
tr
u
m
en
tatio
n
T
ec
h
n
o
lo
g
y
E
n
g
in
e
er
in
g
,
T
ec
h
n
ical
E
n
g
i
n
ee
r
in
g
C
o
lleg
e
/
Mo
s
u
l
,
No
r
th
er
n
T
ec
h
n
ical
Un
iv
e
r
s
ity
I
r
aq
E
m
ail:
ab
d
u
lr
ah
m
an
.
k
h
@
n
tu
.
e
d
u
.
iq
1.
I
NT
RO
D
UCT
I
O
N
B
io
m
etr
ic
r
ec
o
g
n
itio
n
is
th
e
ap
p
r
o
ac
h
th
at
co
m
p
r
is
es
m
o
s
t
o
f
th
e
id
en
tific
atio
n
f
ea
tu
r
es
[
1
]
.
B
io
m
etr
ics
ca
n
b
e
eith
er
a
p
h
y
s
io
lo
g
ical
tr
ait
s
u
ch
as
ir
is
[
2
,
3
]
,
f
ac
e
[
4
-
6]
,
p
alm
[
7
,
8
]
,
ea
r
[
9
]
,
f
in
g
er
tex
tu
r
e
[
10
-
1
2
]
,
f
o
o
tp
r
in
t
[
1
3
]
an
d
f
in
g
er
p
r
in
t
[
1
4
-
1
6
]
,
o
r
a
b
e
h
av
io
u
r
al
tr
ait
as
s
ig
n
atu
r
e
[
1
7
],
h
an
d
wr
itin
g
[
1
8
]
,
g
ait
[
1
9
,
20
]
an
d
v
o
ice
[
21
,
22
]
.
On
th
e
o
t
h
er
h
an
d
,
tr
ad
i
tio
n
al
m
eth
o
d
s
f
o
r
in
d
iv
id
u
al
s
’
r
ec
o
g
n
itio
n
a
n
d
au
th
en
ticatio
n
a
r
e
b
ased
o
n
wh
at
a
p
e
r
s
o
n
h
av
e
o
r
k
n
o
w
s
u
c
h
as c
ar
d
s
,
k
ey
s
,
PIN
c
o
d
es,
p
ass
wo
r
d
s
.
Ho
wev
er
,
th
ese
tr
ad
itio
n
al
m
eth
o
d
s
ca
n
b
e
lo
s
t,
f
o
r
g
o
tten
o
r
s
to
len
wh
ich
m
ay
af
f
ec
t
th
eir
r
eliab
ilit
y
[
1
,
23
]
.
T
h
e
r
ec
e
n
t
d
ev
elo
p
m
e
n
t
in
co
m
p
u
ter
p
r
o
ce
s
s
in
g
h
as
led
to
in
cr
e
ase
th
e
d
ep
en
d
en
cy
o
n
b
io
m
etr
ics
c
o
m
p
ar
in
g
with
th
e
tr
ad
itio
n
al
m
eth
o
d
s
[
1
,
23
].
I
n
th
e
b
io
m
etr
ic
r
ec
o
g
n
itio
n
s
y
s
tem
,
th
er
e
is
n
o
n
ee
d
to
ca
r
r
y
a
k
ey
o
r
ca
r
d
an
d
n
o
n
ee
d
to
r
em
em
b
er
a
PIN
co
d
e
o
r
p
ass
wo
r
d
.
T
h
es
e
f
ac
ilit
ate
th
e
v
alid
atio
n
p
r
o
c
ess
es
o
f
u
s
er
s
[1
,
24]
.
Ho
wev
e
r
,
s
ev
er
al
d
r
awb
ac
k
s
h
av
e
b
ee
n
p
r
esen
ted
i
n
b
io
m
etr
ic
s
y
s
tem
s
wh
ich
in
clu
d
e
b
u
t
n
o
t
lim
ited
to
th
eir
lo
wer
s
ec
r
e
cy
an
d
s
ec
u
r
ity
th
at
m
ak
e
th
em
v
u
l
n
er
ab
le
t
o
ex
t
er
n
al
th
r
ea
ts
[
1
,
2
5
,
2
6
]
.
Fak
e
b
io
m
etr
ics
ar
e
th
e
m
o
s
t
p
o
p
u
lar
th
r
ea
ts
ag
ain
s
t
b
io
m
etr
ic
s
ec
u
r
ity
s
y
s
tem
s
.
Fo
r
in
s
tan
ce
,
f
ak
e
f
in
g
e
r
p
r
in
ts
ca
n
ea
s
ily
b
e
co
n
s
tr
u
cte
d
b
y
u
s
in
g
g
u
m
m
y
f
in
g
er
s
m
ad
e
f
r
o
m
s
p
ec
if
ic
m
ater
ials
lik
e
s
ilico
n
e,
wo
o
d
g
lu
e
an
d
g
elatin
e
[
27
]
.
Fig
u
r
e
1
s
h
o
ws
s
am
p
les
o
f
r
ea
l
an
d
f
ak
e
f
in
g
er
p
r
i
n
ts
f
r
o
m
AT
VS
-
Fak
eFin
g
er
p
r
in
t D
ataBas
e
(
AT
VS
-
FF
p
_
DB
)
[
1
]
wh
ich
ar
e
u
s
ed
in
th
is
s
tu
d
y
.
T
h
ese
ch
allen
g
es
h
a
v
e
in
c
r
ea
s
ed
th
e
d
e
m
an
d
o
f
d
ev
elo
p
in
g
r
o
b
u
s
t
b
io
m
etr
ic
r
ec
o
g
n
itio
n
s
y
s
tem
s
[
2
8
,
2
9
]
.
B
asically
,
f
i
n
g
er
p
r
i
n
t c
lass
if
icatio
n
tech
n
iq
u
es o
f
s
p
o
o
f
in
g
d
etec
tio
n
ca
n
b
e
ca
teg
o
r
ized
in
to
two
class
es b
ased
o
n
ad
d
itio
n
a
l sen
s
o
r
s
th
at
ar
e
u
s
ed
o
r
n
o
t:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6
9
3
0
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
19
,
No
.
3
,
J
u
n
e
2
0
2
1
:
8
9
3
-
90
1
894
(
a)
Har
d
war
e
b
ased
tech
n
iq
u
es:
T
h
ese
ar
e
th
e
f
in
g
er
p
r
i
n
ts
r
ea
d
e
r
s
an
d
class
if
ier
s
wh
ich
ar
e
ab
le
to
s
ca
n
th
e
f
in
g
er
s
with
h
ig
h
r
eso
lu
tio
n
i
m
ag
es
f
o
r
d
etec
tin
g
th
e
s
p
o
o
f
an
d
r
ea
l
f
in
g
er
p
r
in
ts
[
3
0
,
3
1
]
.
T
h
e
aim
o
f
u
s
in
g
th
e
h
ar
d
war
e
class
if
icatio
n
tech
n
iq
u
es
is
to
d
etec
t
t
h
e
m
ain
liv
en
ess
f
ea
tu
r
es
s
u
ch
as
th
e
s
k
in
d
is
to
r
tio
n
[
32
]
an
d
b
lo
o
d
f
l
o
w
[
33
].
(
b
)
So
f
twar
e
b
ased
tech
n
iq
u
es:
T
h
ese
m
eth
o
d
s
ca
n
class
if
y
th
e
f
in
g
er
p
r
i
n
ts
in
to
f
ak
e
an
d
r
e
al
f
in
g
er
p
r
i
n
ts
b
ased
o
n
th
eir
s
ca
n
n
e
d
im
a
g
es
b
y
t
h
e
u
s
ed
s
en
s
o
r
s
with
o
u
t
th
e
n
ee
d
to
u
tili
ze
m
o
r
e
h
ar
d
war
e.
T
h
is
class
if
icatio
n
ca
n
also
b
e
d
iv
id
ed
in
to
two
ca
teg
o
r
ies:
−
F
ea
tu
r
e
ex
tr
ac
tio
n
a
n
d
r
ec
o
g
n
itio
n
m
eth
o
d
s
:
th
e
y
d
ep
e
n
d
o
n
liv
en
ess
d
etec
tio
n
o
f
f
in
g
e
r
p
r
in
ts
.
Ho
wev
er
,
d
u
e
to
th
eir
lo
w
ac
cu
r
ac
ies
f
o
r
s
ev
er
al
s
p
o
o
f
in
g
m
ater
ials
an
d
also
t
h
e
r
e
q
u
ir
e
d
lo
n
g
p
r
o
ce
s
s
in
g
tim
e
f
o
r
ex
tr
ac
t
in
g
f
ea
tu
r
es,
m
an
y
r
esear
ch
er
s
h
av
e
b
ec
o
m
e
less
in
ter
est
in
th
is
ty
p
e
[
2
3
,
3
1
]
.
−
Dee
p
lear
n
in
g
m
eth
o
d
s
:
th
ese
o
n
th
e
o
th
er
h
a
n
d
h
a
ve
m
ad
e
g
r
ea
t
s
u
cc
ess
in
th
e
f
ield
o
f
r
ec
o
g
n
itio
n
an
d
class
if
icatio
n
s
u
ch
as
f
in
g
er
p
r
in
ts
[
34
]
,
em
o
tio
n
s
[
3
5
,
3
6
]
,
b
o
n
e
f
is
s
u
r
es
[
3
7
]
an
d
d
is
o
r
d
er
class
if
icatio
n
s
[
3
8
]
.
Dee
p
lear
n
in
g
tech
n
iq
u
e
is
b
ased
o
n
u
s
i
n
g
ad
eq
u
ate
n
u
m
b
er
o
f
tr
ain
in
g
d
ata
wh
ich
lead
to
au
to
m
atica
lly
lear
n
th
eir
s
tr
u
ctu
r
es
an
d
f
ea
tu
r
es
[
3
4
,
3
9
]
.
T
h
e
class
if
icatio
n
p
r
o
ce
s
s
es
in
th
is
f
ield
ar
e
m
o
s
t
lik
ely
c
o
n
ce
n
tr
ate
o
n
eith
e
r
class
if
y
in
g
th
e
f
in
g
er
p
r
in
ts
in
to
th
eir
f
ea
tu
r
es
(
r
ig
h
t lo
o
p
,
lef
t lo
o
p
,
a
r
ch
a
n
d
wh
o
r
l)
o
r
class
if
y
in
g
th
e
f
i
n
g
e
r
p
r
in
ts
in
to
r
ea
l a
n
d
f
a
k
e.
(
a)
(
b
)
Fig
u
r
e
1
.
Sam
p
les o
f
f
in
g
e
r
p
r
i
n
ts
f
r
o
m
AT
VS
-
FF
p
_
DB
[
1
]
:
(
a)
th
e
f
ir
s
t r
o
w
d
em
o
n
s
tr
ates r
ea
l f
in
g
er
p
r
in
ts
,
an
d
(
b
)
th
e
s
ec
o
n
d
r
o
w
s
h
o
ws f
ak
e
f
in
g
er
p
r
i
n
ts
T
h
er
e
ar
e
v
a
r
io
u
s
ap
p
r
o
ac
h
es
th
at
wer
e
s
u
g
g
ested
f
o
r
f
in
g
e
r
p
r
in
t
d
etec
tio
n
s
.
E
x
am
p
les
o
f
th
ese
ar
e:
an
ap
p
r
o
ac
h
o
f
u
s
in
g
s
tatis
tical
weig
h
t
ca
lcu
latio
n
with
n
o
n
-
r
e
f
er
en
ce
f
in
g
er
p
r
in
t
was
p
r
esen
ted
in
[
2
4
]
,
an
alter
n
ativ
e
ap
p
r
o
ac
h
f
o
r
g
en
er
atin
g
ca
n
ce
llab
le
f
in
g
er
p
r
in
t
was
d
escr
ib
ed
b
y
co
n
s
id
er
in
g
o
p
er
atio
n
s
in
m
atr
ices
[
1
5
]
a
n
d
f
in
g
er
p
r
in
ts
ca
n
b
e
u
s
ed
f
o
r
a
d
v
an
ce
d
en
cr
y
p
tio
n
to
g
e
n
er
ate
f
u
zz
y
v
a
u
lt
cr
y
p
to
b
io
m
et
r
ic
k
ey
[
1
6
]
.
T
h
is
p
ap
er
p
r
o
p
o
s
es
an
ef
f
icien
t
m
o
d
el
o
f
d
ee
p
le
ar
n
in
g
ter
m
e
d
th
e
DFC
N
to
ac
h
iev
e
an
ac
cu
r
ate
class
if
icatio
n
o
f
f
in
g
e
r
p
r
in
ts
i
n
to
r
ea
l
a
n
d
f
ak
e.
T
h
e
r
em
ain
i
n
g
s
ec
tio
n
s
in
th
is
p
ap
er
ar
e
o
r
g
an
ized
as
f
o
llo
ws:
s
ec
tio
n
2
p
r
esen
ts
th
e
p
r
i
o
r
wo
r
k
o
f
f
in
g
e
r
p
r
in
t
class
if
icatio
n
s
,
s
ec
tio
n
3
d
is
p
lay
s
th
e
m
eth
o
d
o
l
o
g
y
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el
,
s
ec
tio
n
4
clar
if
ies
th
e
em
p
lo
y
ed
d
ataset
s
p
ec
if
icatio
n
s
,
s
ec
tio
n
5
de
m
o
n
s
tr
ates
th
e
r
esu
lts
an
d
d
is
cu
s
s
io
n
s
an
d
s
ec
tio
n
6
y
iel
d
s
th
e
co
n
clu
s
io
n
.
2.
P
RE
VIOU
S WO
RK
I
n
th
e
liter
atu
r
e
,
d
ee
p
lear
n
in
g
tech
n
iq
u
es
th
at
wer
e
u
s
ed
to
cl
ass
if
y
f
in
g
er
p
r
i
n
ts
in
to
r
ea
l
an
d
f
ak
e
ca
n
r
ec
en
tly
b
e
f
o
u
n
d
.
Pre
v
io
u
s
s
tu
d
ies
ca
n
b
e
r
e
v
iewe
d
as
f
o
llo
ws:
I
n
2
0
1
6
,
No
g
u
ei
r
a
et
a
l.
u
s
ed
f
in
e
-
t
u
n
ed
VGG
-
S
an
d
VGG
-
F
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NNs)
to
class
if
y
f
in
g
er
p
r
in
ts
in
to
f
o
u
r
class
es
(
r
ig
h
t
lo
o
p
,
lef
t
lo
o
p
,
ar
ch
an
d
wh
o
r
l)
u
s
in
g
th
e
NI
ST
s
p
ec
ial
d
atab
ase
4
(
NI
ST
SD4
)
.
T
h
e
ac
h
iev
ed
class
if
y
in
g
ac
cu
r
ac
ies
f
o
r
b
o
t
h
n
etwo
r
k
s
(
th
e
VG
G
-
S
an
d
VGG
-
F)
wer
e
9
5
.
0
5
%
an
d
9
4
.
4
%,
r
esp
ec
tiv
e
ly
[
27
]
.
I
n
2
0
1
8
,
E
l
Ham
d
i
et
a
l.
also
u
s
ed
th
e
NI
ST
SD4
d
atab
ase
to
test
t
h
eir
p
r
o
p
o
s
ed
m
o
d
el.
T
h
e
m
o
d
el
was
co
n
s
tr
u
cted
b
ased
o
n
th
e
co
m
b
i
n
atio
n
o
f
co
n
ic
r
a
d
o
n
tr
an
s
f
o
r
m
(
C
R
T
)
an
d
C
NN.
T
h
e
r
esu
lts
s
h
o
wed
h
ig
h
class
if
icatio
n
ac
cu
r
ac
y
o
f
9
6
.
5
%
[
40
]
.
A
p
atch
-
b
ased
m
o
d
el
u
s
in
g
a
f
u
l
ly
C
NN
with
o
p
tim
al
th
r
esh
o
ld
to
d
etec
t
s
p
o
o
f
f
in
g
er
p
r
in
t
was
u
tili
ze
d
b
y
P
ar
k
et
a
l.
in
2
0
1
8
.
T
h
ey
em
p
lo
y
ed
th
e
liv
en
ess
d
etec
tio
n
(
L
iv
Det)
d
ata
s
ets
(
L
iv
Det
-
2
0
1
1
,
2
0
1
3
,
2
0
1
5
)
in
th
e
ex
p
er
im
en
ts
.
T
h
eir
p
r
o
p
o
s
ed
m
eth
o
d
r
e
v
ea
led
h
ig
h
p
er
f
o
r
m
an
ce
with
1
.
3
5
%
av
er
ag
e
class
if
icatio
n
er
r
o
r
[
41
]
.
I
n
2
0
1
9
,
Uliy
an
et
a
l.
p
r
o
p
o
s
ed
a
d
ee
p
lear
n
in
g
m
o
d
el
to
class
if
y
f
in
g
er
p
r
in
ts
f
r
o
m
L
iv
Det
d
atasets
(
L
iv
Det
-
2013,
2
0
1
5
)
in
t
o
r
ea
l
an
d
f
a
k
e
b
y
u
s
in
g
b
o
th
d
is
cr
im
in
ativ
e
r
estricte
d
b
o
ltzm
an
n
m
ac
h
in
es
(
DR
B
M)
n
etwo
r
k
an
d
d
ee
p
b
o
ltzm
an
n
m
ac
h
in
e
(
D
B
M)
n
etwo
r
k
.
L
in
ea
r
d
is
cr
im
i
n
an
t
an
a
ly
s
is
(
L
DA)
f
o
llo
wed
b
y
th
e
K
-
Nea
r
est
Ne
ig
h
b
o
u
r
was
u
s
ed
f
o
r
t
h
e
class
if
icatio
n
p
u
r
p
o
s
es.
T
h
e
ac
h
iev
ed
ac
cu
r
ac
y
b
y
th
e
p
r
o
p
o
s
ed
m
o
d
el
was
ab
o
u
t
9
6
%
[
31
]
.
I
n
2
0
1
9
,
a
n
o
th
er
d
ee
p
lear
n
in
g
m
et
h
o
d
f
o
r
f
o
r
g
ed
f
in
g
er
p
r
in
ts
d
etec
tio
n
was
p
r
o
p
o
s
ed
b
y
De
So
u
za
an
d
h
is
c
o
lleag
u
es.
T
h
e
s
u
g
g
este
d
m
eth
o
d
d
ep
e
n
d
ed
o
n
d
ee
p
b
o
ltzm
an
n
m
ac
h
in
e
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
Dee
p
fin
g
erp
r
in
t c
la
s
s
ifica
tio
n
n
etw
o
r
k
(
A
b
d
u
ls
a
tta
r
M.
I
b
r
a
h
im
)
895
(
DB
M)
n
etwo
r
k
an
d
u
s
ed
t
h
e
L
iv
Det
-
2
0
1
3
d
ataset.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
was
m
ain
ly
co
n
s
is
ted
o
f
th
r
ee
s
tep
s
:
s
tar
tin
g
with
im
ag
e
n
o
r
m
aliz
atio
n
f
o
llo
wed
b
y
t
r
ain
in
g
p
r
o
ce
s
s
an
d
en
d
e
d
with
th
e
cla
s
s
if
y
in
g
p
r
o
ce
s
s
o
f
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
(
SV
M)
.
T
h
e
ac
c
u
r
ac
y
o
f
th
e
s
u
g
g
ested
s
y
s
tem
was
8
5
.
8
2
%
[
42
]
.
Als
o
in
2
0
1
9
,
Yu
an
et
a
l.
ac
h
iev
e
d
s
atis
f
ac
to
r
y
class
if
y
in
g
r
esu
lts
b
y
u
s
in
g
a
d
ee
p
m
o
d
el
o
f
d
ee
p
r
esid
u
al
n
etwo
r
k
(
DR
N)
f
o
llo
wed
b
y
an
en
h
an
ce
d
an
d
class
if
ied
m
eth
o
d
n
am
ed
th
e
lo
ca
l
g
r
ad
ien
t
p
atter
n
(
L
GP)
[
43
]
.
L
ast
b
u
t
n
o
t
least,
Z
h
an
g
an
d
h
is
r
esear
ch
in
g
t
ea
m
p
r
o
p
o
s
ed
a
s
im
p
le
d
ee
p
s
y
s
tem
d
ep
e
n
d
in
g
o
n
r
esid
u
al
C
NN
ca
lled
th
e
Slim
-
R
es
C
NN
to
clas
s
if
y
th
e
f
in
g
er
p
r
i
n
ts
in
to
r
ea
l
an
d
f
ak
e.
Data
s
ets
f
r
o
m
L
iv
Det
-
2
0
1
3
,
L
iv
Det
-
2
0
1
5
a
n
d
L
iv
Det
-
2
0
1
7
wer
e
u
s
ed
to
c
h
ec
k
th
e
r
eliab
ilit
y
o
f
t
h
eir
p
r
o
p
o
s
ed
s
y
s
tem
.
T
h
ey
m
an
a
g
ed
to
g
et
th
e
h
ig
h
est
class
if
icatio
n
ac
cu
r
ac
y
eq
u
al
t
o
9
5
.
2
5
% f
o
r
L
iv
Det
-
2
0
1
7
d
at
aset [
44
].
I
t
ca
n
b
e
s
ee
n
th
at
class
if
y
in
g
f
in
g
er
p
r
in
ts
h
as
alm
o
s
t
b
ee
n
i
m
p
lem
en
ted
b
y
ap
p
ly
i
n
g
c
o
m
b
in
atio
n
s
o
f
m
u
ltip
le
p
r
o
ce
s
s
es,
wh
ich
m
ay
af
f
ec
t
o
n
t
h
eir
p
er
f
o
r
m
a
n
ce
s
th
r
o
u
g
h
s
p
en
d
in
g
lo
n
g
p
r
o
ce
s
s
in
g
tim
e.
I
t
ca
n
also
b
e
in
v
esti
g
ated
f
r
o
m
th
e
liter
atu
r
e
th
at
th
e
ac
c
u
r
ac
ies
o
f
o
f
d
etec
tin
g
s
p
o
o
f
f
in
g
er
p
r
in
ts
ar
e
n
o
t
s
u
f
f
icien
t
.
W
e
ar
e
aim
in
g
to
co
n
tr
ib
u
te
to
t
h
is
im
p
o
r
tan
t
ar
ea
b
y
p
r
esen
tin
g
o
n
e
in
tellig
en
t
m
o
d
el
ca
lled
th
e
DFC
N,
wh
ich
h
as
th
e
ab
ilit
y
to
d
etec
t th
e
r
ea
l a
n
d
f
ak
e
f
in
g
er
p
r
in
ts
with
h
ig
h
p
er
f
o
r
m
a
n
ce
an
d
ac
cu
r
ac
y
.
3.
M
E
T
H
O
DO
L
O
G
Y
T
h
is
p
ap
er
ap
p
r
o
a
h
ce
d
a
n
ef
f
icien
t
d
ee
p
lear
n
in
g
m
o
d
el
n
a
m
ed
th
e
DFC
N.
T
h
is
n
etwo
r
k
m
o
d
el
is
b
ased
o
n
th
e
c
o
n
v
o
lu
tio
n
al
n
e
u
r
al
n
etwo
r
k
(
C
NN)
.
I
t
aim
s
t
o
class
if
y
f
in
g
e
r
p
r
in
ts
in
to
r
ea
l
(
o
r
ig
in
al
)
an
d
f
a
k
e
f
in
g
er
p
r
in
ts
.
Fig
u
r
e
2
s
h
o
ws
th
e
DFC
N
ar
ch
itectu
r
e.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
co
n
s
is
ts
o
f
m
u
ltip
le
lay
er
s
.
T
h
e
f
o
llo
win
g
p
o
in
ts
s
u
m
m
ar
iz
e
th
e
lay
er
s
o
f
th
e
DFC
N
m
o
d
el
s
tar
tin
g
f
r
o
m
a
f
in
g
e
r
p
r
i
n
t
in
p
u
t
to
th
e
last
class
if
icatio
n
lay
er
:
−
R
ea
d
in
g
f
in
g
er
p
r
in
t
im
ag
e.
T
h
e
DFC
N
i
s
ad
ap
ted
to
ac
ce
p
t
a
n
im
ag
e
o
f
r
ea
l
o
r
f
a
k
e
f
in
g
er
p
r
in
t
in
its
in
p
u
t.
An
y
g
r
ay
s
ca
le
f
in
g
e
r
p
r
in
t
im
a
g
e
o
f
s
ize
3
0
0
×3
0
0
p
i
x
els ca
n
b
e
u
s
ed
as
an
in
p
u
t t
o
o
u
r
DF
C
N
m
o
d
el.
E
ac
h
p
ix
el
v
alu
e
in
a
f
in
g
e
r
p
r
in
t
im
ag
e
co
u
ld
b
e
with
in
th
e
r
an
g
e
[
0
-
2
5
5
]
.
−
T
h
e
co
n
v
o
lu
ti
o
n
al
p
r
o
ce
s
s
is
th
e
f
ir
s
t
lay
er
in
th
e
DFC
N.
T
wo
-
Dim
en
s
io
n
al
(
2
D)
c
o
n
v
o
lu
tio
n
s
ar
e
h
er
e
im
p
lem
en
ted
f
o
r
ea
c
h
in
p
u
t
f
i
n
g
er
p
r
i
n
t
im
ag
e.
W
e
u
s
ed
8
f
il
ter
s
in
th
is
lay
er
with
t
h
e
s
ize
5
×5
p
i
x
els.
E
ac
h
o
f
th
e
eig
h
t
f
ilter
s
was
s
p
ec
if
ied
to
ex
tr
ac
t
u
s
ef
u
l
f
ea
tu
r
es
f
r
o
m
a
f
in
g
e
r
p
r
in
t
in
p
u
t.
L
et
l
is
th
e
cu
r
r
en
t
h
i
d
d
e
n
lay
er
,
l
-
1
is
th
e
p
r
ev
io
u
s
h
id
d
en
lay
er
a
n
d
n
is
th
e
n
u
m
b
er
o
f
n
o
d
es
in
t
h
e
cu
r
r
en
t
lay
e
r
.
T
h
e
f
o
llo
win
g
m
ath
em
atica
l e
x
p
r
ess
io
n
s
h
o
w
s
th
e
g
en
er
al
f
o
r
m
u
la
o
f
th
e
c
o
n
v
o
lu
tio
n
al
lay
er
:
=
∑
,
∗
−
1
+
(
1
)
wh
er
e:
is
th
e
n
-
th
o
u
tp
u
t
o
f
l
lay
er
,
is
th
e
m
-
th
i
n
p
u
t
to
l
lay
e
r
,
,
is
th
e
c
o
n
v
o
lu
tio
n
k
er
n
el
b
et
wee
n
th
e
in
p
u
t
a
n
d
o
u
tp
u
t
(
m
-
th
a
n
d
n
-
th
,
r
esp
ec
tiv
el
y
)
,
(
*
)
is
th
e
o
p
er
atio
n
o
f
th
e
co
n
v
o
l
u
tio
n
,
−
1
is
th
e
n
u
m
b
er
o
f
in
p
u
t c
h
an
n
el
s
,
an
d
r
ep
r
ese
n
ts
th
e
b
ias o
f
th
e
n
-
th
o
u
tp
u
t
[
4
5
]
.
−
Activ
atio
n
f
u
n
ctio
n
is
u
s
ed
af
t
er
th
e
co
n
v
o
lu
tio
n
lay
e
r
.
T
h
e
R
ec
tifie
d
L
in
ea
r
Un
ite
(
R
eL
U
)
is
em
p
lo
y
ed
as
th
e
ac
tiv
atio
n
f
u
n
ctio
n
.
T
h
e
f
o
llo
win
g
f
o
r
m
u
la
r
ep
r
esen
ts
th
e
R
eL
U
o
p
er
atio
n
:
=
R
e
L
U
(
−
1
)
=
{
−
1
,
−
1
≥
0
0
,
−
1
<
0
(
2
)
wh
er
e:
is
an
o
u
t
p
u
t o
f
th
e
R
eL
U
ac
tiv
atio
n
f
u
n
ctio
n
[
4
6
]
.
−
Max
-
p
o
o
lin
g
lay
e
r
is
ap
p
lied
to
s
h
r
in
k
t
h
e
s
ize
o
f
r
esu
lted
d
ata
f
r
o
m
th
e
p
r
ev
io
u
s
lay
er
.
T
h
e
s
ize
o
f
u
s
ed
f
ilter
in
th
is
lay
er
was
5
×5
p
ix
els
with
th
e
s
tr
id
e
o
f
5
p
ix
els.
T
h
e
m
ax
-
p
o
o
lin
g
lay
er
to
o
k
th
e
h
ig
h
est
v
alu
e
o
f
ea
ch
5
×5
b
lo
ck
o
f
p
ix
els.
T
h
is
v
alu
e
was
co
n
s
id
er
ed
as
th
e
m
o
s
t
ac
tiv
ated
p
ix
el
o
n
b
eh
al
f
o
f
o
th
er
p
ix
els.
T
h
is
lay
er
lead
to
d
ec
r
ea
s
e
th
e
co
m
p
u
tatio
n
al
lo
ad
a
n
d
th
e
r
e
q
u
ir
ed
p
r
o
ce
s
s
in
g
tim
e.
I
t
also
p
lay
s
a
r
o
le
in
d
ec
r
ea
s
in
g
th
e
o
v
er
f
itti
n
g
p
r
o
c
ess
.
T
h
e
f
o
llo
win
g
f
o
r
m
u
la
r
e
p
r
esen
ts
th
e
f
u
n
ctio
n
o
f
th
is
lay
er
:
=
ma
x
(
−
1
)
(
3
)
wh
er
e:
is
an
o
u
t
p
u
t o
f
th
e
m
a
x
-
p
o
o
lin
g
lay
e
r
an
d
Z
is
a
b
lo
ck
o
f
5
×5
p
i
x
els.
−
Fu
lly
co
n
n
ec
ted
la
y
er
is
p
r
o
v
i
d
ed
to
p
e
r
f
o
r
m
m
atch
in
g
b
etw
ee
n
th
e
p
r
e
v
io
u
s
la
y
er
(
th
e
m
a
x
-
p
o
o
lin
g
lay
er
i
n
o
u
r
c
a
s
e
)
a
n
d
t
h
e
n
e
x
t
l
a
y
er
.
B
as
e
d
o
n
t
h
e
f
u
n
c
t
i
o
n
o
f
t
h
i
s
l
a
y
e
r
,
t
h
e
n
u
m
b
e
r
o
f
o
u
t
p
u
t
c
l
as
s
es
ca
n
b
e
s
p
e
c
i
f
i
e
d
h
e
r
e
.
I
n
t
h
i
s
s
t
u
d
y
,
2
o
u
t
p
u
t
s
a
r
e
r
e
q
u
i
r
e
d
f
o
r
t
h
e
r
e
al
a
n
d
f
a
k
e
cl
as
s
es
.
T
h
e
n
a
s
m
en
t
i
o
n
e
d
,
t
h
e
f
u
ll
y
c
o
n
n
e
c
t
e
d
l
a
y
e
r
w
il
l
m
a
tc
h
b
e
tw
e
e
n
t
h
e
n
u
m
b
e
r
o
f
r
e
q
u
i
r
e
d
c
la
s
s
e
s
a
n
d
t
h
e
n
o
d
es
o
f
t
h
e
m
a
x
-
p
o
o
l
i
n
g
l
a
y
e
r
.
−
So
f
tm
ax
lay
e
r
is
s
u
b
s
eq
u
en
tly
g
iv
en
to
p
r
e
p
ar
e
t
h
e
d
ata
f
o
r
th
e
class
if
icatio
n
p
u
r
p
o
s
es.
T
h
e
o
u
tp
u
t
o
f
th
is
lay
er
is
r
ep
r
esen
ted
b
y
v
al
u
es
in
th
e
r
an
g
e
o
f
0
to
1
.
E
ac
h
v
alu
e
g
iv
es
th
e
p
r
o
b
ab
ilit
y
o
f
th
e
r
elatio
n
s
h
ip
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6
9
3
0
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
19
,
No
.
3
,
J
u
n
e
2
0
2
1
:
8
9
3
-
90
1
896
b
etwe
en
th
e
cu
r
r
en
t
f
in
g
er
p
r
in
t
in
p
u
t
an
d
a
n
o
u
tp
u
t
class
.
T
h
e
m
ath
em
atica
l
f
o
r
m
u
la
o
f
th
e
s
o
f
tm
ax
f
u
n
ctio
n
is
r
ep
r
esen
ted
as f
o
llo
ws:
=
−
1
∑
−
1
=
1
(
4
)
wh
er
e:
is
an
o
u
tp
u
t
o
f
th
e
s
o
f
t
m
ax
lay
er
an
d
−
1
is
an
in
p
u
t
to
th
is
lay
er
an
d
C
is
th
e
n
u
m
b
e
r
o
f
class
es
.
T
h
e
d
en
o
m
in
ato
r
is
th
e
s
u
m
m
atio
n
o
f
th
e
ex
p
o
n
en
tials
o
f
all
th
e
v
alu
es in
th
e
v
ec
to
r
[
4
5
]
.
−
T
h
e
f
in
al
lay
er
i
n
th
e
p
r
o
p
o
s
e
d
DFC
N
m
o
d
el
is
th
e
class
if
ica
tio
n
lay
er
.
I
t ta
k
es its
in
p
u
ts
f
r
o
m
th
e
s
o
f
tm
ax
lay
er
as
n
u
m
b
er
s
b
etwe
en
0
an
d
1
,
two
v
alu
es
f
o
r
ea
ch
i
n
p
u
t
f
in
g
e
r
p
r
in
t
im
ag
e
.
T
h
e
f
u
n
ctio
n
o
f
th
is
class
if
icatio
n
lay
er
is
to
d
ec
id
e
wh
eth
er
th
e
f
in
g
e
r
p
r
in
t
is
r
e
al
o
r
f
ak
e
d
e
p
en
d
in
g
o
n
th
e
p
r
o
v
id
ed
v
al
u
es
f
r
o
m
th
e
s
o
f
tm
ax
lay
er
.
T
h
e
cl
o
s
er
th
e
in
p
u
t
to
1
lead
to
s
elec
t
its
co
r
r
esp
o
n
d
i
n
g
o
u
tp
u
t
class
an
d
v
ice
v
er
s
a.
Fig
u
r
e
2
.
T
h
e
ar
ch
itectu
r
e
o
f
t
h
e
p
r
o
p
o
s
ed
DFC
N
m
o
d
el
4.
E
M
P
L
O
YE
D
DAT
ASE
T
T
o
ass
ess
o
u
r
m
o
d
el,
we
h
a
v
e
u
s
ed
two
g
r
o
u
p
s
o
f
r
ea
l
an
d
f
a
k
e
f
in
g
er
p
r
i
n
t
im
ag
es
f
r
o
m
th
e
AT
VS
-
Fak
eFin
g
er
p
r
in
t
Data
B
ase
(
AT
VS
-
FF
p
D
B
)
,
s
p
ec
if
ic
ally
,
th
e
with
o
u
t
co
o
p
er
atio
n
d
ataset.
T
h
e
m
id
d
le
an
d
in
d
ex
f
in
g
er
s
o
f
b
o
th
h
an
d
s
wer
e
u
s
ed
to
ca
p
tu
r
e
th
eir
f
in
g
er
p
r
in
ts
b
y
a
ca
p
ac
itiv
e
s
en
s
o
r
.
Fo
u
r
f
in
g
er
p
r
i
n
t
im
ag
es
f
o
r
ea
ch
f
in
g
er
wer
e
ac
q
u
ir
ed
[
1
,
47
]
.
T
h
e
to
tal
n
u
m
b
er
s
o
f
f
in
g
er
p
r
i
n
ts
th
at
ar
e
u
tili
ze
d
in
th
is
s
tu
d
y
is
5
1
2
im
ag
es.
T
h
e
y
ar
e
eq
u
ally
d
iv
id
ed
in
to
two
g
r
o
u
p
s
o
f
r
ea
l a
n
d
f
ak
e
f
in
g
e
r
p
r
in
ts
.
W
e
h
av
e
ex
p
lo
ited
h
alf
o
f
th
e
d
ataset
(
5
0
%
o
f
r
ea
l
f
in
g
er
p
r
in
ts
p
lu
s
5
0
%
o
f
f
ak
e
f
in
g
e
r
p
r
in
ts
)
f
o
r
th
e
lear
n
i
n
g
p
u
r
p
o
s
es.
W
h
ils
t,
th
e
s
ec
o
n
d
h
alf
h
as b
ee
n
ex
p
lo
ited
f
o
r
th
e
DFC
N
ef
f
icien
cy
test
in
g
.
5.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
h
as
b
ee
n
th
o
r
o
u
g
h
ly
ev
alu
ated
b
y
ch
an
g
in
g
d
if
f
er
en
t
d
ee
p
lear
n
in
g
p
ar
am
eter
s
.
Ma
n
y
ex
p
er
im
e
n
t
s
wer
e
im
p
lem
en
ted
to
e
x
am
i
n
e
th
e
ac
cu
r
ac
ies
o
f
tu
n
i
n
g
v
a
r
io
u
s
d
ee
p
lear
n
in
g
p
ar
am
eter
s
.
I
t
is
wo
r
th
m
en
tio
n
in
g
th
at
th
e
s
tar
tin
g
p
ar
am
ete
r
s
wer
e
in
itialized
ac
co
r
d
in
g
to
th
e
n
o
v
e
l
ap
p
r
o
ac
h
in
[
48
]
.
T
ab
le
1
s
h
o
ws
th
e
8
c
ases
th
at
ar
e
co
n
s
id
er
ed
in
th
i
s
p
ap
er
,
wh
er
e
ea
c
h
ca
s
e
r
ep
r
esen
ts
tu
n
in
g
v
alu
es
f
o
r
a
ce
r
tain
p
ar
am
ete
r
.
T
o
r
ea
ch
th
e
m
o
s
t
ef
f
icien
t
p
e
r
f
o
r
m
a
n
ce
o
f
th
e
s
u
g
g
ested
D
FC
N
m
o
d
el,
th
e
tu
n
i
n
g
p
r
o
ce
s
s
es
o
f
th
e
d
ee
p
lear
n
in
g
p
ar
a
m
eter
s
ar
e
d
iv
id
ed
in
t
o
eig
h
t
ca
s
es.
E
a
ch
o
f
th
ese
eig
h
t
ca
s
es
in
clu
d
es
ad
ju
s
tin
g
o
n
e
p
ar
am
eter
wh
ile
leav
i
n
g
th
e
o
th
er
s
with
o
u
t
alter
atio
n
.
T
h
e
n
u
m
b
e
r
o
f
ep
o
c
h
s
was
ass
ig
n
ed
to
5
0
f
o
r
al
l
ex
p
er
im
en
ts
as
th
is
wo
u
ld
p
r
o
v
id
e
f
air
co
m
p
ar
is
o
n
s
b
et
we
en
all
ex
p
er
im
en
ts
an
d
ad
d
r
e
s
s
th
e
p
r
o
b
lem
o
f
im
p
lem
en
tin
g
v
ar
iatio
n
s
.
Star
tin
g
with
th
e
in
itializatio
n
ca
s
e,
th
e
p
r
o
p
o
s
ed
p
ar
am
eter
s
o
f
th
e
n
o
v
el
ap
p
r
o
ac
h
i
n
[
4
8
]
wer
e
u
s
ed
with
o
u
t
ad
ju
s
tm
en
t.
T
h
ese
v
ar
iab
les
y
ield
ed
t
h
e
f
ir
s
t
ac
c
u
r
ac
y
o
f
8
7
.
8
9
%
in
T
a
b
l
e
1
.
T
h
e
f
i
lter
s
ize
in
th
is
ca
s
e
was
m
o
d
if
ied
in
d
escen
d
in
g
s
tep
s
f
r
o
m
15
15
p
ix
els
to
3
3
p
ix
els
wh
ile
th
e
o
th
er
p
a
r
am
eter
s
wer
e
lef
t
u
n
ch
an
g
ed
.
T
h
e
h
ig
h
est
ac
cu
r
ac
y
in
th
is
ca
s
e
was
9
6
.
8
8
%
wh
en
u
s
in
g
th
e
f
ilter
s
ize
o
f
5
5
p
ix
els.
T
h
e
f
ilter
s
ize
(
5
5
p
ix
els)
t
h
at
g
av
e
t
h
e
h
ig
h
est
ac
cu
r
ac
y
in
th
e
f
ir
s
t
ca
s
e
was
f
ix
ed
in
th
e
s
ec
o
n
d
ca
s
e
wh
er
ea
s
th
e
n
u
m
b
er
o
f
f
ilter
s
was
co
n
s
id
er
ed
as
th
e
ch
an
g
ea
b
le
v
ar
iab
le
f
r
o
m
1
6
to
4
f
ilter
s
.
T
h
e
b
est
n
u
m
b
er
o
f
f
ilter
s
was
8
f
ilter
s
,
wh
e
r
e
it
ac
h
iev
ed
th
e
h
ig
h
est
ac
c
u
r
ac
y
o
f
9
9
.
2
2
%
a
n
d
th
is
is
t
h
e
r
atio
n
ale
b
e
h
in
d
u
s
in
g
th
is
n
u
m
b
e
r
o
f
f
ilter
s
.
I
n
cr
e
asin
g
o
r
d
ec
r
ea
s
in
g
th
is
n
u
m
b
er
lead
s
t
o
r
ed
u
ce
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
ap
p
l
ied
ap
p
licatio
n
as it
ca
n
clea
r
ly
b
e
s
ee
n
in
th
e
s
am
e
tab
le.
T
h
e
r
est
ca
s
es
o
f
3
,
4
,
5
,
6
,
7
an
d
8
in
th
e
s
am
e
tab
le
in
clu
d
ed
ch
an
g
in
g
o
n
ly
o
n
e
p
a
r
am
eter
in
ea
ch
ca
s
e.
Ho
wev
er
,
th
e
ex
p
er
im
en
ts
s
h
o
wed
th
at
th
e
alter
atio
n
o
f
th
e
o
th
er
p
ar
am
ete
r
s
in
th
ese
s
tep
s
d
id
n
o
t
en
h
an
c
e
th
e
ac
c
u
r
ac
y
o
f
th
e
p
r
o
p
o
s
e
d
m
o
d
el.
C
ase
3
in
v
o
lv
ed
ch
an
g
in
g
th
e
s
tr
id
e
o
f
f
ilt
er
s
in
th
e
co
n
v
o
l
u
tio
n
lay
er
b
etwe
en
[
1
1
]
p
ix
els
an
d
[
4
4
]
p
ix
els.
C
ase
4
f
o
cu
s
ed
o
n
c
h
an
g
in
g
th
e
p
ad
d
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
Dee
p
fin
g
erp
r
in
t c
la
s
s
ifica
tio
n
n
etw
o
r
k
(
A
b
d
u
ls
a
tta
r
M.
I
b
r
a
h
im
)
897
p
r
o
ce
s
s
f
r
o
m
th
e
0
p
ad
d
i
n
g
,
wh
ich
g
av
e
t
h
e
b
est
r
esu
lts
,
t
o
th
e
1
,
2
,
3
an
d
(
s
am
e)
p
ad
d
in
g
v
al
u
es.
C
ase
5
ex
am
in
ed
th
e
c
h
an
g
in
g
o
f
p
o
o
lin
g
f
ilter
s
ize,
f
o
llo
wed
b
y
ca
s
e
6
wh
ich
co
m
p
r
is
ed
o
f
c
h
an
g
in
g
th
e
p
o
o
lin
g
s
tr
id
e.
T
h
e
f
in
al
two
ca
s
es
7
a
n
d
8
in
clu
d
ed
th
e
ad
ju
s
tin
g
o
f
p
ad
d
i
n
g
v
alu
e
a
n
d
t
y
p
e
o
f
p
o
o
lin
g
,
r
esp
ec
tiv
ely
.
At
th
e
en
d
o
f
th
e
tu
n
in
g
ex
p
er
im
en
ts
,
th
e
p
ar
am
eter
s
th
at
h
av
e
r
esu
lted
with
th
e
h
ig
h
est
ac
cu
r
ac
y
ar
e
b
en
ch
m
ar
k
ed
f
o
r
th
e
DFC
N
m
o
d
el.
T
ab
le
1
.
I
m
p
lem
en
ted
ex
p
e
r
im
en
ts
to
ex
am
in
e
th
e
ac
cu
r
ac
i
es o
f
d
etec
tin
g
r
ea
l a
n
d
f
ak
e
f
i
n
g
er
p
r
i
n
ts
b
y
tu
n
in
g
v
ar
io
u
s
d
ee
p
lear
n
in
g
p
ar
a
m
eter
s
(
5
0
ep
o
ch
s
ea
ch
)
C
a
ses
C
o
n
v
.
(
N
o
.
o
f
r
e
q
u
i
r
e
d
p
a
r
a
met
e
r
s
4
)
P
o
o
l
i
n
g
(
N
o
.
o
f
r
e
q
u
i
r
e
d
p
a
r
a
m
e
t
e
r
s 4
)
A
c
c
u
r
a
c
y
(
%)
F
i
l
t
e
r
si
z
e
N
o
.
o
f
f
i
l
t
e
r
s
S
t
r
i
d
e
P
a
d
d
i
n
g
Ty
p
e
S
i
z
e
S
t
r
i
d
e
P
a
d
d
i
n
g
1
15
15
10
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
8
7
.
8
9
13
13
10
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
8
4
.
3
8
11
11
10
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
9
1
.
8
0
9
9
10
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
7
8
.
1
3
7
7
10
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
9
2
.
9
7
5
5
10
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
9
6
.
8
8
3
3
10
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
9
3
.
3
6
2
5
5
1
6
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
9
1
.
4
1
5
5
1
4
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
9
5
.
3
1
5
5
1
2
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
9
2
.
9
7
5
5
10
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
9
6
.
8
8
5
5
8
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
9
9
.
2
2
5
5
6
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
9
3
.
3
6
5
5
4
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
9
2
.
9
7
3
5
5
8
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
9
9
.
2
2
5
5
8
[
2
2
]
0
M
a
x
5
5
[
5
5
]
0
9
2
.
9
7
5
5
8
[
3
3
]
0
M
a
x
5
5
[
5
5
]
0
9
2
.
1
9
5
5
8
[
4
4
]
0
M
a
x
5
5
[
5
5
]
0
9
1
.
8
0
4
5
5
8
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
9
9
.
2
2
5
5
8
[
1
1
]
1
M
a
x
5
5
[
5
5
]
0
9
1
.
4
1
5
5
8
[
1
1
]
2
M
a
x
5
5
[
5
5
]
0
9
4
.
9
2
5
5
8
[
1
1
]
3
M
a
x
5
5
[
5
5
]
0
9
7
.
6
6
5
5
8
[
1
1
]
same
M
a
x
5
5
[
5
5
]
0
9
4
.
9
2
5
5
5
8
[
1
1
]
0
M
a
x
15
15
[
5
5
]
0
9
6
.
0
9
5
5
8
[
1
1
]
0
M
a
x
13
13
[
5
5
]
0
9
2
.
5
8
5
5
8
[
1
1
]
0
M
a
x
11
11
[
5
5
]
0
9
8
.
4
4
5
5
8
[
1
1
]
0
M
a
x
9
9
[
5
5
]
0
9
4
.
1
4
5
5
8
[
1
1
]
0
M
a
x
7
7
[
5
5
]
0
9
4
.
1
4
5
5
8
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
9
9
.
2
2
5
5
8
[
1
1]
0
M
a
x
3
3
[
5
5
]
0
9
1
.
8
0
6
5
5
8
[
1
1
]
0
M
a
x
5
5
[
2
2
]
0
8
6
.
3
3
5
5
8
[
1
1
]
0
M
a
x
5
5
[
3
3
]
0
9
2
.
9
7
5
5
8
[
1
1
]
0
M
a
x
5
5
[
4
4
]
0
9
6
.
0
9
5
5
8
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
9
9
.
2
2
7
5
5
8
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
9
9
.
2
2
5
5
8
[
1
1
]
0
M
a
x
5
5
[
5
5
]
1
9
6
.
8
8
5
5
8
[
1
1
]
0
M
a
x
5
5
[
5
5
]
2
9
8
.
8
3
5
5
8
[
1
1
]
0
M
a
x
5
5
[
5
5
]
3
9
8
.
8
3
5
5
8
[
1
1
]
0
M
a
x
5
5
[
5
5
]
same
9
8
.
8
3
8
5
5
8
[
1
1
]
0
M
a
x
5
5
[
5
5
]
0
9
9
.
2
2
5
5
8
[
1
1
]
0
A
v
e
r
a
g
e
5
5
[
5
5
]
0
9
1
.
8
0
Fig
u
r
e
3
s
h
o
ws
th
e
ac
cu
r
ac
y
o
f
ea
ch
ca
s
e
with
r
esp
ec
t
t
o
its
ch
an
g
ea
b
le
p
ar
am
eter
.
Ad
d
itio
n
al
ex
ec
u
ted
ex
p
er
im
en
ts
f
o
r
ass
ess
in
g
d
if
f
er
en
t
d
ee
p
lear
n
in
g
o
p
tim
izer
s
.
T
h
e
th
r
ee
tr
ain
in
g
o
p
tim
izer
s
:
s
to
ch
asti
c
g
r
ad
ien
t
d
escen
t
with
m
o
m
e
n
tu
m
(
SGDM)
,
ad
ap
tiv
e
m
o
m
en
t
esti
m
atio
n
(
Ad
am
)
an
d
r
o
o
t
m
ea
n
s
q
u
ar
e
p
r
o
p
a
g
atio
n
(
R
MSPr
o
p
)
wer
e
ev
alu
ated
ac
co
r
d
in
g
to
th
e
class
if
icatio
n
ac
cu
r
ac
y
.
T
ab
le
2
s
u
m
m
ar
izes
th
e
ef
f
ec
ts
o
f
ex
am
i
n
in
g
th
ese
o
p
t
im
izer
s
o
n
th
e
DFC
N
ac
cu
r
ac
y
.
T
ab
le
2
.
Ad
d
itio
n
al
ex
ec
u
te
d
ex
p
er
im
en
ts
f
o
r
ass
ess
in
g
d
if
f
er
en
t d
ee
p
lea
r
n
in
g
o
p
tim
izer
s
(
5
0
e
p
o
ch
s
ea
ch
)
O
p
t
i
mi
z
e
r
A
c
c
u
r
a
c
y
(
%)
S
G
D
M
9
9
.
2
2
A
d
a
m
9
6
.
4
8
R
M
S
P
r
o
p
9
7
.
2
7
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6
9
3
0
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
19
,
No
.
3
,
J
u
n
e
2
0
2
1
:
8
9
3
-
90
1
898
Fig
u
r
e
3
.
T
h
e
ac
cu
r
ac
ies o
f
ea
ch
ca
s
e
with
r
esp
ec
t to
its
ch
a
n
g
ea
b
le
p
a
r
am
eter
(
th
e
n
u
m
b
er
o
f
ep
o
ch
s
was a
s
s
ig
n
ed
to
5
0
)
Fig
u
r
e
4
s
h
o
ws
th
e
tr
ain
in
g
c
u
r
v
es
o
f
o
u
r
s
u
g
g
ested
a
p
p
r
o
ac
h
.
T
h
e
p
ar
am
eter
s
th
at
ar
e
u
tili
ze
d
f
o
r
tr
ain
in
g
s
tag
e
wer
e
s
et
u
p
as
f
o
llo
ws:
SG
DM
o
p
tim
izer
,
m
o
m
en
tu
m
=
0
.
9
,
weig
h
t
d
ec
a
y
=0
.
0
0
0
1
,
lear
n
in
g
r
ate=
0
.
0
3
,
m
i
n
im
u
m
-
b
atch
s
ize=
1
2
8
an
d
m
a
x
im
u
m
ep
o
ch
=
5
0
.
T
h
e
cu
r
v
es
o
f
th
e
ac
cu
r
ac
y
an
d
th
e
lo
s
s
s
h
o
w
th
at
th
e
s
u
g
g
ested
m
o
d
el
s
tar
t
ed
with
h
i
g
h
l
o
s
s
an
d
lo
w
ac
c
u
r
ac
y
d
u
r
in
g
th
e
f
ir
s
t
iter
atio
n
s
.
Ho
w
ev
er
,
t
h
e
lo
s
s
was
th
en
d
r
am
atica
lly
d
ec
r
ea
s
ed
an
d
th
e
ac
cu
r
ac
y
was
s
ig
n
if
ican
tly
in
cr
ea
s
ed
at
t
h
e
s
am
e
tim
e.
T
h
e
c
u
r
v
es
s
h
o
w
th
at
th
e
lo
s
s
an
d
ac
cu
r
ac
y
r
ea
ch
ed
th
eir
b
est p
e
r
f
o
r
m
an
ce
s
an
d
co
n
tin
u
ed
u
n
til th
e
en
d
o
f
tr
ai
n
in
g
.
Fo
r
estab
lis
h
in
g
f
air
co
m
p
ar
is
o
n
s
with
s
tate
-
of
-
th
e
-
ar
t
n
etwo
r
k
s
,
p
r
e
v
io
u
s
p
r
o
p
o
s
ed
m
o
d
els
h
av
e
b
ee
n
s
im
u
lated
an
d
ev
alu
ate
d
f
o
r
o
u
r
em
p
lo
y
ed
d
ataset
an
d
o
u
tp
u
ts
.
T
h
r
ee
d
ee
p
lear
n
in
g
n
etw
o
r
k
s
wer
e
u
s
ed
in
th
ese
co
m
p
ar
is
o
n
s
.
T
h
ese
ar
e
th
e
C
NN
[
4
8
]
,
DFTL
[
4
9
]
an
d
DFL
[
5
0
]
.
T
ab
le
3
s
h
o
ws
th
ese
co
m
p
ar
is
o
n
s
.
T
h
is
tab
le
clea
r
ly
d
em
o
n
s
tr
ates
th
at
th
e
s
u
g
g
ested
DF
C
N
s
u
r
p
ass
ed
o
th
er
s
tate
-
of
-
th
e
-
ar
t
m
o
d
els
an
d
it
atta
in
ed
Ca
se
1
Ca
se
2
Ca
se
3
Ca
se
4
Ca
se
5
Ca
se
6
Ca
se
7
Ca
se
8
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
Dee
p
fin
g
erp
r
in
t c
la
s
s
ifica
tio
n
n
etw
o
r
k
(
A
b
d
u
ls
a
tta
r
M.
I
b
r
a
h
im
)
899
s
u
p
er
io
r
p
er
f
o
r
m
a
n
ce
o
f
9
9
.
2
2
%.
Fu
r
th
er
m
o
r
e
,
th
e
im
p
r
o
v
em
en
ts
in
th
e
ac
cu
r
ac
ies
y
ield
9
.
0
6
%,
7
.
0
8
%
an
d
2
.
7
6
% c
o
m
p
ar
in
g
to
th
e
C
NN,
DFTL
an
d
DFL
n
etwo
r
k
s
,
r
esp
ec
tiv
ely
.
Fig
u
r
e
4
.
T
h
e
tr
ain
in
g
cu
r
v
es o
f
o
u
r
s
u
g
g
ested
DFC
N
m
o
d
el
T
ab
le
3
.
C
o
m
p
a
r
is
o
n
s
with
s
tate
-
of
-
th
e
-
a
r
t a
p
p
r
o
ac
h
es
A
p
p
r
o
a
c
h
A
c
c
u
r
a
c
y
(
%)
C
N
N
[
4
8
]
9
0
.
2
3
D
F
TL
[
4
9
]
9
2
.
1
9
D
F
L
[
5
0
]
9
6
.
4
8
Ou
r
a
p
p
r
o
a
c
h
9
9
.
2
2
6.
CO
NCLU
SI
O
N
Du
e
to
th
e
im
p
o
r
tan
ce
o
f
f
in
g
er
p
r
in
t
as
a
p
h
y
s
io
lo
g
ical
tr
ait
f
o
r
in
d
iv
id
u
al
r
ec
o
g
n
itio
n
an
d
d
u
e
to
th
e
r
ea
lity
o
f
its
v
u
ln
er
ab
ilit
y
to
b
e
b
r
ea
c
h
ed
b
y
f
ak
e
s
am
p
les,
we
p
r
o
p
o
s
ed
a
n
ef
f
icien
t
d
ee
p
lea
r
n
in
g
m
o
d
el
n
am
e
d
th
e
DFC
N
to
clas
s
if
y
f
in
g
er
p
r
in
ts
in
to
two
ca
teg
o
r
ies:
r
ea
l
an
d
f
ak
e.
T
h
e
s
u
g
g
ested
m
o
d
el
was
ap
p
r
o
ac
h
ed
af
ter
ap
p
ly
i
n
g
m
a
n
y
e
x
p
er
i
m
en
ts
th
at
ex
ten
s
iv
ely
e
v
alu
ate
v
ar
io
u
s
d
ee
p
lear
n
in
g
p
a
r
am
eter
s
.
W
e
u
s
ed
AT
VS
-
FF
p
DB
f
in
g
er
p
r
in
t
im
ag
es
f
o
r
o
u
r
p
r
o
p
o
s
ed
m
o
d
el.
T
h
e
DFC
N
ac
h
iev
ed
g
r
ea
t
s
u
cc
ess
with
a
class
if
icatio
n
ac
cu
r
ac
y
r
ea
ch
ed
to
9
9
.
2
2
%.
T
h
e
ex
p
e
r
im
en
ts
also
s
h
o
wed
th
at
th
e
D
FC
N
o
u
tp
er
f
o
r
m
ed
s
tate
-
of
-
ar
t a
p
p
r
o
ac
h
es w
h
en
u
tili
zin
g
th
e
em
p
lo
y
e
d
d
ataset.
I
n
f
u
tu
r
e
wo
r
k
,
d
i
f
f
er
en
t
ty
p
e
s
o
f
f
in
g
er
p
r
in
t f
a
k
e
s
am
p
les
ca
n
b
e
co
n
s
id
er
e
d
.
I
t a
p
p
ea
r
s
th
at
it
is
im
p
o
r
tan
t
to
estab
lis
h
ad
d
itio
n
al
class
if
icatio
n
ap
p
r
o
ac
h
wh
ich
ca
n
r
ec
o
g
n
ize
th
e
ty
p
e
o
f
s
p
o
o
f
in
g
.
RE
F
E
R
E
NC
E
S
[1
]
J.
G
a
lb
a
ll
y
,
F
.
Alo
n
so
-
F
e
rn
a
n
d
e
z
,
J.
F
ierre
z
a
n
d
J.
Orte
g
a
-
G
a
rc
ia,
“
A
Hig
h
P
e
rfo
rm
a
n
c
e
F
i
n
g
e
rp
rin
t
Li
v
e
n
e
ss
De
tec
ti
o
n
M
e
t
h
o
d
Ba
se
d
o
n
Q
u
a
li
ty
Re
late
d
F
e
a
tu
re
s
,
”
Fu
t
u
re
Ge
n
e
ra
ti
o
n
C
o
mp
u
ter
S
y
ste
ms
,
v
o
l.
2
8
,
n
o
.
1
,
p
p
.
3
1
1
-
3
2
1
,
2
0
1
2
.
[2
]
M
.
A.
M
.
A
b
d
u
ll
a
h
,
R.
R.
Al
-
Ni
m
a
,
S
.
S
.
Dla
y
,
W
.
L
.
Wo
o
a
n
d
J
.
A.
Ch
a
m
b
e
rs
,
“
Cro
ss
-
sp
e
c
tral
Ir
is
M
a
tch
i
n
g
f
o
r
S
u
rv
e
il
la
n
c
e
Ap
p
li
c
a
ti
o
n
s
,
”
S
p
rin
g
e
r,
S
u
rv
e
il
la
n
c
e
in
Acti
o
n
T
e
c
h
n
o
l
o
g
ies
fo
r
Civili
a
n
,
M
il
i
ta
ry
a
n
d
Cy
b
e
r
S
u
rv
e
il
la
n
c
e
,
Ch
a
p
ter
5
,
2
0
1
7
.
[3
]
M
.
R.
Kh
a
li
l,
M
.
S
.
M
a
jee
d
,
a
n
d
R.
R.
Om
a
r,
“
P
e
rso
n
a
l
id
e
n
ti
fic
a
ti
o
n
wit
h
iri
s
p
a
t
tern
s
,
”
AL
-
Ra
f
i
d
a
i
n
J
o
u
r
n
a
l
o
f
Co
mp
u
ter
S
c
ien
c
e
s a
n
d
M
a
t
h
e
ma
ti
c
s
,
v
o
l
.
6
,
n
o
.
1
,
p
p
.
1
3
-
2
6
,
2
0
0
9
.
[4
]
S
.
A.
M
.
Al
-
S
u
m
a
i
d
a
e
e
,
M
.
A.
M
.
Ab
d
u
ll
a
h
,
R.
R
.
O.
Al
-
Nim
a
,
S
.
S
.
Dla
y
a
n
d
J.
A.
C
h
a
m
b
e
rs,
“
M
u
lt
i
-
g
ra
d
ien
t
F
e
a
tu
re
s
a
n
d
El
o
n
g
a
ted
Q
u
in
a
r
y
P
a
tt
e
rn
En
c
o
d
i
n
g
f
o
r
Im
a
g
e
-
b
a
se
d
F
a
c
ial
Ex
p
re
ss
io
n
Re
c
o
g
n
it
i
o
n
,
”
El
se
v
ier
,
Pa
tt
e
rn
Rec
o
g
n
it
io
n
,
v
o
l.
7
1
,
p
p
.
2
4
9
-
2
6
3
,
2
0
1
7
.
[5
]
R.
R.
Al
-
Nim
a
,
F
.
S
.
M
u
sta
fa
,
“
F
a
c
e
Re
c
o
g
n
it
i
o
n
Us
in
g
In
v
a
rian
t
M
o
m
e
n
ts
F
e
a
tu
re
s
,
”
T
h
e
S
c
ien
ti
fi
c
J
o
u
rn
a
l,
Co
ll
e
g
e
o
f
S
c
ien
c
e
,
T
ikr
it
Un
ive
rs
it
y
,
v
o
l.
1
4
,
n
o
.
2
,
p
p
.
2
5
3
-
2
5
9
,
2
0
0
9
.
[6
]
R.
Alimu
in
,
E.
Da
d
io
s,
J.
Da
y
a
o
,
a
n
d
S
.
Are
n
a
s,
“
De
e
p
h
y
p
e
rsp
h
e
re
e
m
b
e
d
d
in
g
fo
r
re
a
l
-
ti
m
e
fa
c
e
re
c
o
g
n
it
io
n
,
”
T
EL
KOM
NIKA
T
e
lec
o
mm
u
n
ic
a
ti
o
n
C
o
mp
u
ti
n
g
El
e
c
tro
n
ics
a
n
d
Co
n
tro
l
,
v
o
l
.
1
8
,
n
o
.
3
,
p
p
.
1
6
7
1
-
1
6
7
7
,
2
0
2
0
.
[7
]
R.
R.
O.
Al
-
Nim
a
,
M
.
Y.
Al
-
Ri
d
h
a
,
F
.
H.
Ab
d
u
lrah
e
e
m
,
“
Re
g
e
n
e
ra
ti
n
g
fa
c
e
ima
g
e
s
fr
o
m
m
u
lt
i
-
sp
e
c
tral
p
a
lm
ima
g
e
s
u
sin
g
m
u
lt
i
p
le
f
u
sio
n
m
e
th
o
d
s
,
”
T
EL
KOM
NIKA
T
e
lec
o
mm
u
n
ic
a
ti
o
n
C
o
mp
u
ti
n
g
E
lec
tro
n
ics
a
n
d
C
o
n
tro
l
,
v
o
l
.
1
7
,
n
o
.
6
,
p
p
.
3
1
1
0
-
3
1
1
9
,
2
0
1
9
.
I
t
e
ra
t
i
o
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6
9
3
0
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
19
,
No
.
3
,
J
u
n
e
2
0
2
1
:
8
9
3
-
90
1
900
[8
]
R.
R.
O.
Al
-
Nim
a
,
N
.
A.
Al
-
Ob
a
i
d
y
a
n
d
L.
A.
Al
-
Hb
e
ti
,
“
S
e
g
m
e
n
ti
n
g
F
i
n
g
e
r
I
n
n
e
r
S
u
rfa
c
e
f
o
r
t
h
e
P
u
rp
o
se
o
f
H
u
m
a
n
Re
c
o
g
n
it
i
o
n
,
”
2
n
d
I
n
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
En
g
i
n
e
e
rin
g
T
e
c
h
n
o
l
o
g
y
a
n
d
it
s
A
p
p
l
ica
ti
o
n
s
(IICE
T
A),
IEE
E
,
2
0
1
9
,
p
p
.
1
0
5
-
110
.
[9
]
R.
R.
Al
-
Nim
a
,
“
Hu
m
a
n
a
u
th
e
n
ti
c
a
ti
o
n
wit
h
e
a
rp
ri
n
t
fo
r
se
c
u
re
tele
p
h
o
n
e
sy
ste
m
,
”
Ira
q
i
J
o
u
rn
a
l
o
f
Co
mp
u
ter
s,
Co
mm
u
n
ica
ti
o
n
s,
C
o
n
tr
o
l
a
n
d
S
y
ste
ms
En
g
in
e
e
rin
g
IJ
CCCE
,
v
o
l.
1
2
,
n
o
.
2
,
p
p
.
4
7
-
5
6
,
2
0
1
2
.
[1
0
]
R.
R.
O.
Al
-
Nim
a
,
S
.
S
.
Dla
y
,
W.
L.
W
o
o
a
n
d
J.
A.
Ch
a
m
b
e
rs,
“
Eff
icie
n
t
F
in
g
e
r
S
e
g
m
e
n
tatio
n
Ro
b
u
st
to
Ha
n
d
Alig
n
m
e
n
t
i
n
Im
a
g
i
n
g
wit
h
A
p
p
l
ica
ti
o
n
to
Hu
m
a
n
Ve
rifi
c
a
ti
o
n
,
”
5
th
IEE
E
I
n
ter
n
a
t
io
n
a
l
W
o
rk
sh
o
p
o
n
Bi
o
me
trics
a
n
d
Fo
re
n
sic
s (IW
BF
)
,
2
0
1
7
.
[1
1
]
R.
R.
O.
Al
-
Nim
a
,
T.
Ha
n
,
T.
Ch
e
n
,
S
.
Dla
y
a
n
d
J.
Ch
a
m
b
e
rs,
"
F
i
n
g
e
r
Tex
t
u
re
Bio
m
e
tri
c
Ch
a
ra
c
teristic:
a
S
u
rv
e
y
,
"
a
rXiv p
re
p
ri
n
t
a
rX
iv
:2
0
0
6
.
0
4
1
9
3
,
2
0
2
0
.
[1
2
]
R.
R.
O.
Al
-
Nim
a
,
M
.
Al
-
Ka
lt
a
k
c
h
i,
S
.
Al
-
S
u
m
a
id
a
e
e
,
S
.
Dla
y
,
W.
Wo
o
,
T.
Ha
n
a
n
d
J.
Ch
a
m
b
e
rs,
“
P
e
rso
n
a
l
v
e
r
i
f
i
c
a
t
i
o
n
b
a
s
e
d
o
n
m
u
l
t
i
-
s
p
e
c
t
r
a
l
f
i
n
g
e
r
t
e
x
t
u
r
e
l
i
g
h
t
i
n
g
i
m
a
g
e
s
,
”
I
ET
S
i
g
n
a
l
P
r
o
c
e
s
s
i
n
g
,
v
o
l
.
1
2
,
no
.
9
,
p
p
.
1
-
1
1
,
2
0
1
8
.
[1
3
]
M
.
M
.
A.
Ab
u
q
a
d
u
m
a
h
,
M
.
A.
M
.
Ali
a
n
d
R.
R.
O.
Al
-
Nim
a
,
“
P
e
rso
n
a
l
Au
th
e
n
ti
c
a
ti
o
n
Ap
p
li
c
a
ti
o
n
Us
in
g
De
e
p
Lea
rn
in
g
Ne
u
ra
l
Ne
two
rk
,
”
in
1
6
t
h
IEE
E
In
ter
n
a
ti
o
n
a
l
Co
ll
o
q
u
i
u
m
o
n
S
ig
n
a
l
Pro
c
e
ss
in
g
&
it
s
A
p
p
li
c
a
ti
o
n
s
(CS
PA
)
,
La
n
g
k
a
wi,
M
a
lay
sia
,
2
0
2
0
,
p
p
.
1
8
6
-
1
9
0
.
[1
4
]
R.
M
u
k
h
a
i
y
a
r,
“
Ca
n
c
e
ll
a
b
le
b
i
o
m
e
tri
c
u
sin
g
m
a
tri
x
a
p
p
ro
a
c
h
e
s
,
”
P
h
D
t
h
e
sis,
S
c
h
o
o
l
o
f
El
e
c
tri
c
a
l
a
n
d
E
lec
tro
n
i
c
En
g
i
n
e
e
rin
g
,
Ne
wc
a
stle Un
iv
e
rsit
y
,
UK
,
2
0
1
5
.
[1
5
]
R.
M
u
k
h
a
iy
a
r,
S
.
S
.
Dla
y
,
a
n
d
W
.
L.
Wo
o
,
“
Altern
a
ti
v
e
Ap
p
r
o
a
c
h
in
G
e
n
e
ra
ti
n
g
Ca
n
c
e
ll
a
b
le
F
in
g
e
rp
rin
t
b
y
Us
i
n
g
M
a
tri
c
e
s Op
e
ra
ti
o
n
s,”
5
6
t
h
In
ter
n
a
ti
o
n
a
l
S
y
mp
o
siu
m E
L
M
A
R
-
2
0
1
4
,
1
0
-
1
2
S
e
p
t
2
0
1
4
,
Za
d
a
r,
Cro
a
ti
a
,
p
p
.
1
0
-
1
2
.
[1
6
]
M
.
S
.
Altara
wn
e
h
,
W.
L.
W
o
o
,
a
n
d
S
.
S
.
Dla
y
,
“
F
u
z
z
y
Va
u
lt
Cr
y
p
to
Bio
m
e
tri
c
Ke
y
Ba
se
d
o
n
F
in
g
e
rp
rin
t
Ve
c
to
r
F
e
a
tu
re
s,”
in
Pro
c
o
f
6
t
h
In
t
.
S
y
mp
.
o
n
Co
mm
u
n
ica
ti
o
n
S
y
ste
ms
Ne
two
rk
a
n
d
Di
g
it
a
l
S
i
g
n
a
l
Pro
c
e
ss
in
g
2
0
0
8
,
IEE
E
,
2
0
0
8
,
p
p
.
4
5
2
-
4
5
6
.
[1
7
]
R.
R.
Al
-
Nim
a
,
S
.
Dla
y
,
a
n
d
W.
Wo
o
,
“
A
n
e
w
a
p
p
ro
a
c
h
t
o
p
re
d
ictin
g
p
h
y
sic
a
l
b
io
m
e
tri
c
s
f
ro
m
b
e
h
a
v
i
o
u
ra
l
b
io
m
e
tri
c
s
,
”
In
:
1
6
t
h
In
ter
n
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
n
Im
a
g
e
An
a
ly
sis
a
n
d
Pr
o
c
e
ss
in
g
.
2
0
1
4
,
L
o
n
d
o
n
,
UK:
W
o
rld
Aca
d
e
my
o
f
S
c
ien
c
e
,
En
g
i
n
e
e
rin
g
a
n
d
T
e
c
h
n
o
l
o
g
y
,
2
0
1
4
.
[1
8
]
O
.
S
u
d
a
n
a
,
I
.
W
.
G
u
n
a
y
a
,
I
.
K
.
G
.
D
.
P
u
t
r
a
,
“
H
a
n
d
w
r
i
t
i
n
g
i
d
e
n
t
i
f
i
c
a
t
i
o
n
u
s
i
n
g
d
e
e
p
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
m
e
t
h
o
d
,
”
T
E
L
K
O
M
N
I
K
A
T
e
l
e
c
o
m
m
u
n
i
c
a
t
i
o
n
C
o
m
p
u
t
i
n
g
E
l
e
c
t
r
o
n
i
c
s
a
n
d
C
o
n
t
r
o
l
,
v
o
l
.
1
8
,
n
o
.
4
,
p
p
.
1
9
3
4
-
1
9
4
1
,
2
0
2
0
.
[1
9
]
Y.
M
a
k
ih
a
ra
,
D.
S
.
M
a
to
v
s
k
i,
M
.
S
.
Nix
o
n
,
J.
N.
Ca
rter,
a
n
d
Y.
Ya
g
i,
“
G
a
it
re
c
o
g
n
it
io
n
:
Da
tab
a
se
s,
re
p
re
se
n
tatio
n
s
,
a
n
d
a
p
p
li
c
a
ti
o
n
s
,
”
W
il
e
y
En
c
y
c
lo
p
e
d
ia
o
f
El
e
c
trica
l
a
n
d
El
e
c
tro
n
ics
En
g
i
n
e
e
rin
g
,
2
0
1
5
.
[2
0
]
D.
M
u
ra
m
a
tsu
,
Y.
M
a
k
i
h
a
ra
,
H.
I
wa
m
a
,
T.
Tan
o
u
e
,
a
n
d
Y.
Ya
g
i
,
“
G
a
it
v
e
rifi
c
a
ti
o
n
sy
ste
m
fo
r
s
u
p
p
o
rti
n
g
c
rimin
a
l
in
v
e
stig
a
t
io
n
,
"
in
2
n
d
IAP
R
Asi
a
n
Co
n
fer
e
n
c
e
o
n
P
a
tt
e
rn
Rec
o
g
n
it
i
o
n
,
2
0
1
3
.
[2
1
]
M
.
T.
S
.
Al
‑
Ka
lt
a
k
c
h
i,
R.
R.
O.
Al
‑
Nim
a
,
M
.
A.
M
.
Ab
d
u
ll
a
h
,
H
.
N.
Ab
d
u
ll
a
h
,
“
Th
o
ro
u
g
h
e
v
a
lu
a
ti
o
n
o
f
TIM
IT
d
a
tab
a
se
sp
e
a
k
e
r
id
e
n
ti
f
ica
ti
o
n
p
e
rfo
rm
a
n
c
e
u
n
d
e
r
n
o
ise
wit
h
a
n
d
with
o
u
t
t
h
e
G
.
7
1
2
ty
p
e
h
a
n
d
se
t
,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
S
p
e
e
c
h
T
e
c
h
n
o
l
o
g
y
,
v
o
l.
2
2
,
n
o
.
3
,
p
p
.
8
5
1
-
8
6
3
,
2
0
1
9
.
[2
2
]
M
.
T.
,
Al
-
Ka
lt
a
k
c
h
i
e
t
a
l.
,
“
C
o
m
p
a
riso
n
o
f
Ex
trem
e
Lea
rn
in
g
M
a
c
h
in
e
a
n
d
Ba
c
k
P
ro
p
a
g
a
ti
o
n
Ba
se
d
i
-
Ve
c
to
r
,
”
T
u
rk
ish
J
o
u
rn
a
l
o
f
El
e
c
trica
l
En
g
in
e
e
rin
g
&
C
o
mp
u
ter
S
c
ien
c
e
s
,
v
o
l.
2
8
,
p
p
.
1
2
3
6
-
1
2
4
3
,
2
0
2
0
.
[2
3
]
D.
M
ich
e
lsa
n
ti
,
A.
D
.
En
e
,
Y.
G
u
ich
i,
R
.
S
tef,
K.
Na
sro
ll
a
h
i
a
n
d
T.
B.
M
o
e
slu
n
d
,
“
F
a
st
F
i
n
g
e
rp
r
in
t
Clas
sifica
ti
o
n
with
De
e
p
Ne
u
ra
l
Ne
two
r
k
s
,
”
In
t
e
rn
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
C
o
mp
u
ter
Vi
sio
n
T
h
e
o
ry
a
n
d
A
p
p
li
c
a
ti
o
n
s
,
2
0
1
7
.
[2
4
]
M
.
S
.
Altara
wn
e
h
,
L.
C.
K
h
o
r,
W.
L.
Wo
o
,
a
n
d
S
.
S
.
Dla
y
,
“
A
NO
N
Re
fe
r
e
n
c
e
F
in
g
e
rp
rin
t
Im
a
g
e
Va
li
d
it
y
v
i
a
S
tatisti
c
a
l
Weig
h
t
Ca
lcu
lati
o
n
,
”
J
o
u
rn
a
l
o
f
Di
g
it
a
l
In
fo
rm
a
ti
o
n
M
a
n
a
g
e
me
n
t
,
”
v
o
l.
5
,
n
o
.
4
,
p
p
.
2
2
0
-
2
2
4
,
2
0
0
7
.
[2
5
]
S
.
M
a
rc
e
l,
M
.
Nix
o
n
a
n
d
S
.
L
i,
H
a
n
d
b
o
o
k
o
f
b
io
me
tric a
n
ti
-
sp
o
o
f
in
g
,
1
st
Ed
.
,
S
p
ri
n
g
e
r
,
C
h
a
m
,
2
0
1
4
.
[2
6
]
D.
M
e
n
o
tt
i
e
t
a
l.
,
“
De
e
p
re
p
re
se
n
tatio
n
s
fo
r
iri
s,
fa
c
e
,
a
n
d
fi
n
g
e
r
p
rin
t
sp
o
o
fi
n
g
d
e
tec
ti
o
n
,
”
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
In
fo
rm
a
t
io
n
F
o
re
n
sic
s a
n
d
S
e
c
u
rity
,
v
o
l.
1
0
,
n
o
.
4
,
p
p
.
8
6
4
-
8
7
9
,
2
0
1
5
.
[2
7
]
R.
F
.
No
g
u
e
ira,
R.
d
e
A.
Lo
tu
f
o
a
n
d
R.
C.
M
a
c
h
a
d
o
,
“
F
in
g
e
rp
ri
n
t
li
v
e
n
e
ss
d
e
tec
ti
o
n
u
sin
g
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
two
rk
s
,
”
IE
EE
tr
a
n
sa
c
ti
o
n
s
o
n
in
fo
rm
a
ti
o
n
f
o
re
n
sic
s a
n
d
se
c
u
rit
y
,
v
o
l
.
1
1
,
n
o
.
6
,
p
p
.
1
2
0
6
-
1
2
1
3
,
2
0
1
6
.
[2
8
]
T.
M
a
tsu
m
o
to
,
H.
M
a
tsu
m
o
to
,
K.
Ya
m
a
d
a
a
n
d
S
.
Ho
sh
i
n
o
"
Im
p
a
c
t
o
f
a
rti
ficia
l
"
g
u
m
m
y
"
f
in
g
e
rs
o
n
f
in
g
e
rp
ri
n
t
sy
ste
m
s
,
"
Pro
c
.
S
PIE
4
6
7
7
,
Op
ti
c
a
l
S
e
c
u
rity
a
n
d
C
o
u
n
ter
feit
De
ter
re
n
c
e
T
e
c
h
n
iq
u
e
s IV
,
2
0
0
2
.
[2
9
]
V.
R.
Alb
a
c
e
te
e
t
a
l.
,
“
Dire
c
t
a
tt
a
c
k
s
u
si
n
g
fa
k
e
ima
g
e
s
in
ir
is
v
e
rifi
c
a
ti
o
n
,
”
In
Eu
r
o
p
e
a
n
W
o
rk
sh
o
p
o
n
B
io
me
trics
a
n
d
Id
e
n
ti
ty M
a
n
a
g
e
me
n
t
,
S
p
rin
g
e
r,
Be
rli
n
,
He
id
e
l
b
e
rg
,
M
a
y
2
0
0
8
.
[3
0
]
J.
J.
En
g
e
lsm
a
,
K.
Ca
o
,
a
n
d
A.
K.
Ja
in
,
“
Ra
sp
irea
d
e
r:
Op
e
n
s
o
u
rc
e
fin
g
e
rp
r
in
t
re
a
d
e
r
,
”
IE
EE
tra
n
s
a
c
ti
o
n
s
o
n
p
a
tt
e
r
n
a
n
a
lys
is a
n
d
ma
c
h
in
e
i
n
telli
g
e
n
c
e
,
v
o
l.
4
1
,
n
o
.
1
0
,
p
p
.
2
5
1
1
-
2
5
2
4
,
2
0
1
8
.
[3
1
]
D.
Uliy
a
n
,
S
.
S
a
d
e
g
h
i
a
n
d
H.
J
a
lab
,
“
An
ti
-
sp
o
o
fi
n
g
m
e
th
o
d
fo
r
fin
g
e
rp
r
in
t
re
c
o
g
n
it
io
n
u
sin
g
p
a
tch
b
a
se
d
d
e
e
p
lea
rn
in
g
m
a
c
h
i
n
e
,
”
En
g
in
e
e
rin
g
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
a
n
In
ter
n
a
ti
o
n
a
l
J
o
u
r
n
a
l
,
v
o
l.
2
3
,
n
o
.
2
,
p
p
.
2
6
4
-
2
7
3
,
2
0
1
9
.
[3
2
]
C
.
C
h
a
m
p
o
d
,
C
.
L
e
n
n
a
r
d
,
P
.
M
a
r
g
o
t
a
n
d
M
.
S
t
o
i
l
o
v
i
c
,
“
F
i
n
g
e
r
p
r
i
n
t
s
a
n
d
o
t
h
e
r
r
i
d
g
e
s
k
i
n
i
m
p
r
e
s
s
i
o
n
s
,
”
C
R
C
p
r
e
s
s
,
2
0
1
7
.
[3
3
]
M
.
Dra
h
a
n
sk
y
,
R.
N
o
tze
l
a
n
d
W.
F
u
n
k
,
“
Li
v
e
n
e
ss
De
tec
ti
o
n
b
a
se
d
o
n
F
i
n
e
M
o
v
e
m
e
n
ts o
f
t
h
e
F
in
g
e
r
ti
p
S
u
rfa
c
e
,
”
in
2
0
0
6
IE
EE
I
n
fo
rm
a
ti
o
n
Assu
r
a
n
c
e
W
o
rk
sh
o
p
,
Wes
t
P
o
i
n
t,
NY
,
US
A,
2
0
0
6
.
[3
4
]
M
.
A.
Wa
n
i,
F
.
A.
Bh
a
t,
S
.
Afz
a
l
a
n
d
A.
I
.
Kh
a
n
,
“
S
u
p
e
rv
ise
d
De
e
p
Lea
rn
i
n
g
in
F
in
g
e
rp
ri
n
t
Re
c
o
g
n
it
io
n
,
”
In
b
o
o
k
:
Ad
v
a
n
c
e
s i
n
De
e
p
L
e
a
r
n
in
g
,
p
p
.
1
1
1
-
1
3
2
,
2
0
2
0
.
[3
5
]
P
.
S
re
e
v
i
d
y
a
,
O.
V.
Ra
m
a
n
a
M
u
r
th
y
,
S
.
Ve
n
i,
“
S
e
n
ti
m
e
n
t
a
n
a
ly
sis
b
y
d
e
e
p
lea
r
n
in
g
a
p
p
ro
a
c
h
e
s
,”
T
EL
KOM
NIKA
T
e
lec
o
mm
u
n
ica
ti
o
n
Co
mp
u
ti
n
g
E
lec
tro
n
ics
a
n
d
C
o
n
tr
o
l
,
v
o
l
.
1
8
,
n
o
.
2
,
p
p
.
7
5
2
-
7
6
0
,
2
0
2
0
.
[3
6
]
O.
J.
Yin
g
,
M
.
M
.
A.
Zab
i
d
i,
N.
Ra
m
li
,
a
n
d
U.
U.
S
h
e
ik
h
,
“
S
e
n
t
i
m
e
n
t
a
n
a
ly
sis
o
f
in
f
o
rm
a
l
M
a
lay
twe
e
ts
with
d
e
e
p
lea
rn
in
g
,
”
IAE
S
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Arti
fi
c
ia
l
In
telli
g
e
n
c
e
,
v
o
l.
9
,
n
o
.
2
,
p
p
.
2
1
2
-
2
2
0
,
2
0
2
0
.
[3
7
]
F
.
M
a
r
t
í
n
e
z
,
C
.
H
e
r
n
á
n
d
e
z
,
a
n
d
F
.
M
a
r
t
í
n
e
z
,
“
E
v
a
l
u
a
t
i
o
n
o
f
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
a
r
c
h
i
t
e
c
t
u
r
e
s
i
n
t
h
e
i
d
e
n
t
i
f
i
c
a
t
i
o
n
o
f
b
o
n
e
f
i
s
s
u
r
e
s
,
”
T
E
L
K
O
M
N
I
K
A
T
e
l
e
c
o
m
m
u
n
i
c
a
t
i
o
n
C
o
m
p
u
t
i
n
g
E
l
e
c
t
r
o
n
i
c
s
a
n
d
C
o
n
t
r
o
l
,
,
v
o
l
.
1
8
,
n
o
.
2
,
p
p
.
8
0
7
-
8
1
4
,
2
0
2
0
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
Dee
p
fin
g
erp
r
in
t c
la
s
s
ifica
tio
n
n
etw
o
r
k
(
A
b
d
u
ls
a
tta
r
M.
I
b
r
a
h
im
)
901
[3
8
]
N.
A.
Ali
e
t
a
l.
,
“
A
u
ti
sm
sp
e
c
tr
u
m
d
is
o
rd
e
r
c
las
sifica
ti
o
n
o
n
e
l
e
c
tro
e
n
c
e
p
h
a
lo
g
ra
m
sig
n
a
l
u
si
n
g
d
e
e
p
lea
rn
in
g
a
lg
o
rit
h
m
”
IAE
S
I
n
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Arti
fi
c
i
a
l
I
n
telli
g
e
n
c
e
,
v
o
l.
9
,
n
o
.
1
,
p
p
.
91
-
9
9
,
2
0
2
0
.
[3
9
]
R.
R.
O.
Al
-
Nim
a
,
T.
Ha
n
a
n
d
T.
Ch
e
n
,
“
Ro
a
d
Trac
k
i
n
g
Us
i
n
g
De
e
p
Re
in
fo
rc
e
m
e
n
t
Lea
rn
in
g
fo
r
S
e
lf
-
d
riv
i
n
g
Ca
r
Ap
p
li
c
a
ti
o
n
s
,
”
In
:
R.
B
u
rd
u
k
,
M
.
Ku
rz
y
n
sk
i
a
n
d
M
.
W
o
z
n
iak
(e
d
s),
Pro
g
re
ss
in
C
o
mp
u
ter
Rec
o
g
n
it
i
o
n
S
y
ste
ms
,
CORES
2
0
1
9
,
Ad
v
a
n
c
e
s in
I
n
te
ll
i
g
e
n
t
S
y
ste
ms
a
n
d
Co
mp
u
ti
n
g
,
S
p
r
in
g
e
r,
C
h
a
m
,
2
0
2
0
.
[4
0
]
D.
El
H
a
m
d
i
e
t
a
l
.
,
“
C
o
m
b
i
n
i
n
g
F
i
n
g
e
r
p
r
i
n
t
s
a
n
d
t
h
e
i
r
R
a
d
o
n
T
r
a
n
s
f
o
r
m
a
s
I
n
p
u
t
t
o
D
e
e
p
L
e
a
r
n
i
n
g
f
o
r
a
F
i
n
g
e
r
p
r
i
n
t
C
l
a
s
s
i
f
i
c
a
t
i
o
n
T
a
s
k
,
”
1
5
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
C
o
n
t
r
o
l
,
A
u
t
o
m
a
t
i
o
n
,
R
o
b
o
t
i
c
s
a
n
d
V
i
s
i
o
n
(
I
C
A
R
C
V
)
,
I
E
E
E
,
2
0
1
8
.
[4
1
]
E.
P
a
rk
,
X.
Cu
i,
W.
Kim
,
J.
L
iu
a
n
d
H.
Kim
,
“
P
a
tch
-
b
a
se
d
fa
k
e
fin
g
e
rp
r
in
t
d
e
tec
ti
o
n
u
sin
g
a
fu
ll
y
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
with
a
sm
a
ll
n
u
m
b
e
r
o
f
p
a
ra
m
e
ters
a
n
d
a
n
o
p
ti
m
a
l
th
re
sh
o
l
d
,
”
a
rXiv
p
re
p
ri
n
t
a
rXiv
:1
8
0
3
.
0
7
8
1
7
,
Co
rn
e
ll
U
n
iv
e
rsit
y
,
2
0
1
8
.
[4
2
]
G
.
d
e
S
o
u
z
a
,
D.
S
a
n
t
o
s,
R.
P
ires
,
A.
M
a
ra
n
a
a
n
d
J.
P
a
p
a
,
"
De
e
p
F
e
a
tu
re
s
Ex
trac
ti
o
n
f
o
r
Ro
b
u
st
F
i
n
g
e
rp
rin
t
S
p
o
o
fi
n
g
Attac
k
De
tec
ti
o
n
,
"
J
o
u
r
n
a
l
o
f
Art
if
icia
l
I
n
telli
g
e
n
c
e
a
n
d
S
o
f
t
Co
m
p
u
ti
n
g
Res
e
a
rc
h
,
v
o
l.
9
,
n
o
.
1
,
p
p
.
4
1
-
4
9
,
2
0
1
9
.
[4
3
]
C.
Yu
a
n
e
t
a
l
.
,
“
De
e
p
Re
sid
u
a
l
Ne
two
rk
wit
h
Ad
a
p
ti
v
e
Lea
rn
in
g
F
ra
m
e
wo
rk
fo
r
F
i
n
g
e
rp
r
in
t
Li
v
e
n
e
ss
De
tec
ti
o
n
,
”
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
Co
g
n
it
ive
a
n
d
De
v
e
lo
p
me
n
t
a
l
S
y
ste
ms
,
v
o
l.
1
2
,
n
o
.
3
,
p
p
.
4
1
-
4
7
3
,
2
0
1
9
.
[4
4
]
Y.
Zh
a
n
g
,
D.
S
h
i,
X.
Zh
a
n
,
D.
Ca
o
,
K.
Z
h
u
a
n
d
Z.
Li
,
“
S
li
m
-
R
e
sCN
N:
A
De
e
p
Re
sid
u
a
l
Co
n
v
o
lu
ti
o
n
a
l
Ne
u
ra
l
Ne
two
rk
fo
r
F
i
n
g
e
rp
ri
n
t
Li
v
e
n
e
ss
De
tec
ti
o
n
,
”
IEE
E
Acc
e
ss
,
v
o
l
.
7
,
p
p
.
9
1
4
7
6
-
9
1
4
8
7
,
2
0
1
9
.
[4
5
]
H.
Qin
a
n
d
M
.
A.
El
-
Ya
c
o
u
b
i,
“
De
e
p
Re
p
re
se
n
tatio
n
-
Ba
se
d
F
e
a
tu
re
Ex
trac
ti
o
n
a
n
d
Re
c
o
v
e
rin
g
f
o
r
F
i
n
g
e
r
-
Ve
i
n
Ve
rifi
c
a
ti
o
n
,
”
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
In
f
o
rm
a
ti
o
n
Fo
re
n
sic
s a
n
d
S
e
c
u
rity
,
v
o
l.
1
2
,
n
o
.
8
,
p
p
.
1
8
1
6
-
1
8
2
9
,
2
0
1
7
.
[4
6
]
M
.
Ta
h
ir,
H.
Tay
a
ra
a
n
d
K.T
.
Ch
o
n
g
,
“
iRNA
-
P
se
KN
C
(
2
m
e
th
y
l):
Id
e
n
ti
f
y
RNA
2
'
-
O
-
m
e
th
y
latio
n
sites
b
y
c
o
n
v
o
lu
ti
o
n
n
e
u
ra
l
n
e
tw
o
rk
a
n
d
C
h
o
u
'
s
p
se
u
d
o
c
o
m
p
o
n
e
n
ts
,
”
J
o
u
r
n
a
l
o
f
t
h
e
o
re
ti
c
a
l
b
i
o
lo
g
y
,
v
o
l.
4
6
5
,
p
p
.
1
-
6
,
2
0
1
9
.
[4
7
]
“
ATVS
-
F
a
k
e
F
in
g
e
rp
rin
t
DA
TAB
ASE
(ATVS
-
F
F
p
DB)”
.
[On
li
n
e
].
Av
a
il
a
b
le:
h
tt
p
:/
/atv
s.
ii
.
u
a
m
.
e
s/a
t
v
s/ffp
_
d
b
.
h
tml
.
[4
8
]
M
.
Ka
ra
d
u
m
a
n
a
n
d
H.
Eren
,
“
De
e
p
lea
rn
i
n
g
b
a
se
d
traffic
d
irec
ti
o
n
sig
n
d
e
tec
ti
o
n
a
n
d
d
e
term
in
in
g
d
riv
in
g
st
y
le,”
i
n
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
C
o
mp
u
ter
S
c
ien
c
e
a
n
d
E
n
g
i
n
e
e
rin
g
(
UBM
K)
,
2
0
1
7
,
p
p
.
1
0
4
6
-
1
0
5
0
.
[4
9
]
R.
R.
Om
a
r,
T.
Ha
n
,
S
.
A.
M
.
Al
-
S
u
m
a
id
a
e
e
a
n
d
T.
Ch
e
n
,
“
De
e
p
fin
g
e
r
tex
t
u
re
lea
rn
in
g
fo
r
v
e
rify
i
n
g
p
e
o
p
le
,
”
IET
Bi
o
me
trics
,
Vo
l.
8
,
Iss
u
e
.
1
,
p
p
.
1
-
1
0
,
2
0
1
9
.
[
5
0
]
R
.
R
.
O
.
A
l
-
N
i
m
a
,
S
.
Q
.
H
a
s
a
n
a
n
d
S
.
E
s
m
a
i
l
,
“
E
x
p
l
o
i
t
i
n
g
t
h
e
D
e
e
p
L
e
a
r
n
i
n
g
w
i
t
h
F
i
n
g
e
r
p
h
o
t
o
s
t
o
R
e
c
o
g
n
i
z
e
P
e
o
p
l
e
,
”
A
c
c
e
p
t
e
d
i
n
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
A
d
v
a
n
c
e
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
2
9
,
n
o
.
7
,
p
p
.
1
3
0
3
5
-
1
3
0
4
6
,
2
0
2
0
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Abd
u
lsa
tt
a
r
M.
Ibra
h
i
m
wo
r
k
s
a
s
a
lec
tu
re
r
in
th
e
N
o
rth
e
r
n
Te
c
h
n
ica
l
Un
i
v
e
rsity
(NTU)
,
Tec
h
n
ica
l
En
g
in
e
e
rin
g
Co
ll
e
g
e
,
M
o
su
l
.
Ho
ld
in
g
M
a
ste
r
d
e
g
re
e
M
S
c
in
t
h
e
El
e
c
tri
c
a
l
a
n
d
El
e
c
tro
n
ic
E
n
g
i
n
e
e
rin
g
fro
m
U
n
iv
e
rsity
o
f
Be
lg
ra
d
e
.
Th
e
m
a
in
re
se
a
rc
h
in
g
field
s
a
re
t
h
e
sig
n
a
l
p
ro
c
e
ss
in
g
a
n
d
d
e
e
p
lea
rn
i
n
g
.
Abd
u
lra
h
m
a
n
K
.
Ee
se
e
is
a
Lec
t
u
re
r
in
t
h
e
Tec
h
n
ica
l
En
g
i
n
e
e
rin
g
Co
ll
e
g
e
o
f
M
o
su
l/
No
rt
h
e
r
n
Tec
h
n
ica
l
Un
i
v
e
rsity
(NTU),
Ir
a
q
.
He
re
c
e
iv
e
d
M
a
ste
r
d
e
g
re
e
M
S
c
i
n
th
e
Bi
o
m
e
d
ica
l
En
g
i
n
e
e
rin
g
fro
m
th
e
Un
iv
e
rsity
o
f
S
trat
h
c
ly
d
e
/Un
it
e
d
Kin
g
d
o
m
i
n
2
0
1
5
.
His
Ba
c
h
e
lo
r
B
.
En
g
wa
s
in
th
e
M
e
d
i
c
a
l
In
str
u
m
e
n
tatio
n
Tec
h
n
o
lo
g
y
E
n
g
in
e
e
rin
g
fro
m
t
h
e
NTU
in
2
0
1
1
.
T
h
e
m
a
in
re
se
a
rc
h
in
g
field
s
a
re
th
e
Bi
o
-
sig
n
a
l
p
r
o
c
e
ss
in
g
,
De
e
p
Lea
rn
i
n
g
,
m
e
d
ica
l
se
n
so
rs
a
n
d
ima
g
e
p
ro
c
e
ss
in
g
.
Ra
id
Ra
fi
O
m
a
r
Al
-
Nim
a
re
c
e
iv
e
d
th
e
BS
c
a
n
d
M
S
c
d
e
g
re
e
s
in
Tec
h
n
ica
l
Co
m
p
u
ter
En
g
i
n
e
e
rin
g
in
2
0
0
0
a
n
d
2
0
0
6
,
re
sp
e
c
ti
v
e
ly
.
D
u
rin
g
2
0
0
6
,
h
e
wo
rk
e
d
a
s a
n
As
sista
n
t
Lec
tu
re
r
in
t
h
e
Tec
h
n
ica
l
Co
ll
e
g
e
o
f
M
o
su
l
,
Ira
q
.
In
2
0
1
1
,
h
e
o
b
tai
n
e
d
th
e
Lec
tu
re
r
sc
ien
ti
fic
t
it
le
i
n
t
h
e
sa
m
e
c
o
ll
e
g
e
.
In
2
0
1
7
,
h
e
a
c
c
o
m
p
li
sh
e
d
h
is
P
h
D
i
n
th
e
S
c
h
o
o
l
o
f
El
e
c
tri
c
a
l
a
n
d
E
lec
tro
n
i
c
En
g
i
n
e
e
rin
g
a
t
Ne
wc
a
stle
Un
iv
e
rsity
,
UK
.
His
re
se
a
rc
h
in
tere
sts
a
re
in
th
e
field
s
o
f
p
a
tt
e
r
n
re
c
o
g
n
it
i
o
n
,
se
c
u
rit
y
,
a
rti
ficia
l
i
n
t
e
ll
ig
e
n
c
e
a
n
d
ima
g
e
p
r
o
c
e
ss
in
g
.
Evaluation Warning : The document was created with Spire.PDF for Python.