Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
10
,
No.
3
,
June
2020
,
pp. 3
00
7
~
3013
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v10
i
3
.
pp3007
-
30
13
3007
Journ
al h
om
e
page
:
http:
//
ij
ece.i
aesc
or
e.c
om/i
nd
ex
.ph
p/IJ
ECE
Person
al id
entity
verifi
cation based
ECG biomet
ric
usin
g
non
-
fidu
cial feat
ures
Ma
rw
a A.
El
s
ha
he
d
Ph
y
sics
d
epa
rtm
ent
,
Facu
lty
of
W
om
en
for
Arts,
Scie
n
ce
s
and Ed
uca
t
ion, Ain
Sha
m
s
Univer
sit
y
,
E
g
y
pt
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
A
ug
29
, 201
9
Re
vised
Jan
5
,
2020
Accepte
d
Ja
n
12
, 2
020
Biom
et
ric
s
was
used
as
an
autom
at
ed
and
f
ast
ac
c
eptable
tech
nolog
y
for
hum
an
ide
nt
ifi
c
at
ion
and
it
m
a
y
be
beha
v
ior
al
or
ph
y
sio
log
ic
a
l
tr
ai
ts.
An
y
biometric
s
y
stem
base
d
o
n
ide
nti
f
ic
a
ti
on
or
ver
ifi
c
at
ion
m
odes
for
hum
an
ide
ntit
y
.
The
e
le
c
trocar
diogra
m
(ECG
)
is
conside
red
as
one
of
the
ph
y
sio
logic
al
biometrics
which
impos
sible
to
m
imic
or
stole
.
ECG
fea
tur
e
ex
tracti
o
n
m
et
hods
were
per
form
ed
usin
g
fiduc
i
al
o
r
n
on
-
fiduc
i
al
appr
oac
h
es.
This
rese
arc
h
pre
se
nts
an
aut
hentic
at
ion
ECG
biometric
s
y
stem
using
non
-
fid
ucial
feature
s
obta
i
ned
b
y
Discre
t
e
W
ave
le
t
Dec
om
positi
on
and
the
Euclid
ea
n
Distanc
e
te
chn
i
que
was
used
to
implement
t
he
ide
nt
i
t
y
ver
ifica
t
ion.
From
the
obt
ai
n
ed
result
s,
th
e
pr
oposed
s
y
st
em
ac
cur
acy
is
96.
66%
al
so,
usi
ng
the
ver
ifica
tion
sy
st
em
is
pre
fer
r
ed
for
a
la
rg
e
num
ber
of
indi
viduals a
s
it
ta
kes
le
ss
ti
m
e
t
o
get t
h
e
de
ci
sio
n.
Ke
yw
or
d
s
:
Bi
om
e
tric
ECG
Fidu
ci
al
Non
-
fid
ucial
Ver
ific
at
io
n
Copyright
©
202
0
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Ma
rw
a
A. El
sha
hed
,
Dep
a
rtem
ent o
f
P
hysci
s
,
F
acu
lt
y of
Wo
m
en
f
or Arts, Scie
nc
es an
d
E
ducat
ion
,
Ain Sham
s Un
iversity
,
Ca
iro,
Egy
pt
.
Em
a
il
: l
sn
tl
@c
cu.
e
du.tw
1.
INTROD
U
CTION
Bi
om
e
tric
s
is
a
bio
l
og
ic
al
a
nd
uniq
ue
c
ha
racteri
sti
cs
tha
t
identify
in
di
viduals.
Bi
om
et
rics
m
ay
be
ph
ysi
ologica
l
as
D
NA,
Ir
is,
and
Fin
ge
rprin
t
or
be
ha
vioral
trai
ts
as
keyst
roke,
gait,
a
nd
sig
natur
e
.
All
trai
ts
wh
ic
h
can
be
us
ed
as
biom
et
ric
sh
ou
ld
be
char
act
er
iz
ed
by
Un
i
ve
rsali
ty
,
Exclusivity
,
Perm
anen
c
e
,
Coll
ect
abili
ty
,
Per
form
ance,
A
cce
pta
bili
t
y
and
Ci
rc
um
ven
ti
on
[
1
-
3].
T
he
el
ect
ro
ca
rd
i
ogram
(EC
G)
is
bio
lo
gical
bi
om
et
ric
i
t
m
ea
su
res
the
c
hange
of
el
ect
rical
act
ivit
y
of
the
hu
m
an
hea
rt
over
ti
m
e
as
s
hown
i
n
F
igure
1
[
4].
The
EC
G
sig
nal
co
ns
ist
s
of
three
wa
ves
sta
rt
with
th
e
el
ect
rical
P
-
wav
e
ge
ner
at
e
s
from
the
up
per
cha
m
ber
s
fo
ll
owe
d
by
strai
gh
t
-
li
ne
generate
s
wh
e
n
t
he
bloo
d
flo
ws
from
t
op
cham
ber
s
to
lo
w
e
r
cham
ber
s.
T
he
QRS
com
plex
is
the
ne
xt
wa
ve
wh
ic
h
gene
rates
f
ro
m
bo
tt
om
cha
m
ber
s
wh
il
e
the
la
st
wav
e
is
the T
wa
ve
f
r
om
e
le
ct
rical
reco
ve
ry [2,
5,
6].
ECG
featu
res
are
fid
ucial
and
non
-
fid
uci
al
featu
res.
Fi
gure
2
sho
ws
al
l
fid
ucial
f
eat
ur
es
f
ro
m
the
wa
ves
w
hich
are
in
t
he
ti
m
e
do
m
ai
n.
N
on
-
fi
du
ci
al
fea
tures
a
re
ob
ta
i
ned
by
us
in
g
s
ever
al
m
et
ho
ds
from
the
trans
form
e
d
dom
ai
n.
The
se
feat
ur
e
s
nee
dless
com
pu
ta
ti
on
ti
m
e
than
the
fid
ucial
fea
tures
al
so
it
do
esn'
t
require
t
o
fin
d
boun
dar
ie
s
of
the
EC
G
waveform
s
wh
ic
h
change
c
onsta
ntly
[7
-
9].
Tw
o
EC
G
a
ppr
oa
ches
a
re
avail
able
base
d
on
the
us
e
d
featur
e
s
wh
ic
h
are
fidu
ci
al
and
non
-
fid
ucia
l
a
pp
r
oac
hes
as
sh
ow
n
in
F
igure
3.
Ther
e
is
no
ge
ner
al
r
ule
to
ge
t
the
exact
locat
ion
s
of
ECG
sign
al
bo
undar
ie
s
beca
us
e
it
has
con
ti
nuou
sly
var
ia
ti
ons, so
usi
ng no
n
-
fid
ucial
f
eat
ures is s
olv
e
d
this
prob
lem
[
1].
A
bio
m
et
ric
s
yst
e
m
m
ay
op
erate
as
identif
ic
at
ion
or
ve
ri
ficat
ion
schem
es.
In
al
l
syst
e
m
s,
the
first
sta
ge
is
data
colle
ct
ion
a
nd
prepa
rin
g
the
databa
se
f
or
the
syst
em
thi
s
proces
s
is
c
al
le
d
the
e
nro
llm
ent
process as
s
ho
wn
i
n
Fig
ur
e
4. The
e
nroll
m
ent p
rocess sta
rts
w
it
h
ECG si
gnal
s co
ll
ect
ing
from
ind
ividu
a
ls
an
d
PI
N
for
eac
h
on
e
.
ECS
sig
nals
prep
ro
c
es
sing
is
pe
rfo
r
m
ed
to
get
m
or
e
s
uitable
sign
al
s
the
n
a
pply
ing
on
e
of
feat
ur
e
extracti
on
te
c
hn
i
qu
e
s
the
n
s
toring
the
obta
ined
featu
res
in
the
syst
em
database
with
a
PIN
for
eac
h
in
div
i
du
al
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
3
,
J
une
2020
:
30
0
7
-
30
1
3
3008
Figure
1. O
ne
hear
t
beat cy
cl
e
Figure
2. ECG
fid
ucial
f
eat
ure
s [10
]
Figure
3. ECG
bio
m
et
ric appr
oach
e
s
Figure
4. En
r
oll
m
ent p
r
ocess
In
a
n
i
de
ntific
at
ion
syst
em
,
i
t
sta
rts
with
th
e
sign
al
ac
qu
is
it
ion
from
ind
ivid
ual
an
d
the
sam
e
s
te
ps
us
e
d
be
fore
in
the
en
ro
ll
m
ent
sta
ge
re
peated
exactl
y
in
prep
ro
ce
ssin
g
a
nd
featu
re
e
xtr
act
ion
process
then
com
par
ing
between
the
ob
ta
ined
feat
ur
es
with
that
store
s
in
the
datab
ase
to
get
the
ind
ivid
ual
id
entit
y.
it
do
es
n'
t
nee
d
ide
ntit
y
cl
ai
m
ed
an
d
c
ons
idere
d
as
one
to
m
any
m
a
t
chin
g
beca
us
e
it
searc
hes
on
al
l
the
datab
ase
a
nd
gets
who
t
he
use
r'
s
ident
it
y
or
non
-
ide
ntifie
d
as
s
ho
wn
i
n
F
ig
ure
5.
In
t
he
ve
rif
ic
at
ion
syst
e
m
,
the
sa
m
e
as
in
the
enr
ollm
ent
pr
oce
ss
in
ECG
sig
na
l
pr
ep
ro
c
essin
g
an
d
featu
re
e
xtracti
on
te
chni
qu
es
is
exactl
y
rep
e
at
ed.
T
he
ve
rif
ic
at
ion
syst
e
m
need
s
i
den
ti
ty
cl
aim
ed
and
c
on
si
der
e
d
as
one
to
on
e
m
atch
in
g
then
getti
ng
th
e
decis
io
n
ye
s
if
the
featu
res
are
m
at
ching
a
nd
no
f
or
not
m
at
ching
s
o
it
s
res
ult
is
pe
rfor
m
ed
qu
ic
kly
an
d
ac
cur
at
e as
s
how
n
in
F
ig
ure
6.
Most
of
the
bio
m
et
ric
syst
e
m
s
us
ed
identi
ficat
ion
sc
hem
es
with
fi
duci
al
and
non
-
fi
duci
al
featur
es
,
wh
il
e
m
os
t
of
ver
ific
at
io
n
bi
om
et
ric
s
yst
e
ms
us
e
d
fid
ucial
featur
e
s
[
11
-
13].
In
[
2],
the
Mult
i
m
od
al
Bi
om
et
ri
c
Au
t
hen
ti
cat
io
n
syst
e
m
is
pr
op
ose
d
us
in
g
a
com
bin
at
ion
of
ECG
a
nd
Fing
e
r
pr
int
to
get
her.
Fid
ucial
f
eat
ur
es
wer
e
extra
ct
ed
from
ECG
sig
nals
afte
r
prep
ro
ce
ssin
g
al
s
o
c
onve
rti
n
g
the
e
xtr
act
ed
fi
ng
e
r
pri
nt
to
the
trans
f
or
m
e
d
dom
ai
n
us
in
g
D
WT
t
hen
e
xtract
only
tw
o
feat
ur
es
.
The
decisi
on
was
getti
ng
from
th
e
fina
l
scor
e
w
hic
h
c
al
culat
ed
f
r
om
m
at
ching
EC
G
a
nd
fi
ng
e
r
pri
nt
trai
ts.
A
l
ot
of
pa
ram
et
ers
we
re
cal
c
ulate
d
f
or
the syst
em
eff
i
ci
ency m
or
eover,
t
he nu
m
ber
of sub
j
ect
s is
not cl
ear.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Perso
nal i
de
ntit
y veri
fi
cation ba
s
ed
ECG
b
i
omet
ric
u
si
ng non
-
fi
duci
al feat
ur
es
(
M
ar
wa
A
. Elsha
hed
)
3009
Figure
5. I
den
t
ific
at
ion
bio
m
e
tric
syst
e
m
Figure
6. Ve
rif
ic
at
ion
b
i
om
et
r
ic
syst
e
m
The
autho
rs
in
[1
4],
Pr
op
os
ed
an
Au
thentic
at
ion
Bi
om
et
ric
syst
em
wh
ic
h
based
on
the
fidu
ci
al
featur
es
directl
y
after
pr
epr
ocessing
,
they
us
ed
73
su
bj
ect
s
in
their
exp
erim
ent.
The
Eucli
dean
distance
was
us
ed
fo
r
m
at
ching
.
Then
they
extend
ed
their
un
im
od
al
f
ram
ewo
rk
based
on
ECG
trai
t
on
ly
to
be
a m
ulti
m
od
al
bio
m
et
ric
syst
em
based
on
ECG
and
Face
or
ECG an
d
fing
erp
rint
us
ing
scor
e
fu
sion
te
chn
iqu
e.
A
ver
ific
at
ion
bio
m
et
ric
syst
em
was
pr
op
os
ed
in
[1
5],
they
us
ed
on
ly
15
su
bj
ect
s,
RR
intervals
wer
e
detect
ed
fr
om
ECG
sign
al
s
then
app
ly
ing
DCT
to
get
the
req
uired
featur
es.
The
ob
ta
ined
accuracy
is
97
.7
8%
after
us
ing
the co
rr
el
at
ion
to
get the d
eci
sion
.
In
[1
6],
ECG
Bi
om
et
ric
Ver
ific
at
ion
syst
em
was
pr
op
os
ed
by
us
ing
PCA,
they
us
ed
8
su
bj
ect
s
on
ly
in
their
exp
eri
m
ent.
They
us
ed
fidu
ci
al
featur
es
us
ing
PCA
and
DW
T
fo
r
no
n
-
fidu
ci
al
featur
es.
The
autho
rs
in
[1
7],
they
get
the
sign
al
featur
es
us
ing
the
Hj
or
th
Descr
iptor
and
Sam
ple
Entro
py
(S
am
pEn
).
Su
pp
or
t
Vector
Ma
chine
(S
VM)
cl
assifi
er
was
us
ed
fo
r
authen
ti
cat
ion
.
The
ob
ta
ined
accuracy
is
93
.8
%
fr
om
Hj
or
th D
escriptor
Com
par
ed
to
Sam
pEn
. I
n
[1
8], ECG
based
b
iom
et
ric
syst
em
w
as
design
ed.
ECG
dataset
of
55
us
ers
was
create
d
du
ring
fo
ur
m
on
ths
to
stud
y
the
ECG
-
based
bio
m
et
rics
sta
bili
ty
.
He
stud
ie
d
the
per
fo
rm
ance
of
E
CG
as
a
bio
m
et
ric
trai
t
in
sh
or
t
and
lon
g
te
rm
s
and
fo
un
d
that
m
or
e
need
ing
stud
ie
s
towar
ds
im
pr
ov
ing
the lon
g
-
te
rm
ECG b
iom
et
ric syst
em
s p
erf
or
m
ance.
The
autho
rs
in
[1
9]
pr
op
os
ed
ECG
bio
m
et
ric
al
go
rithm
us
ing
18
4
su
bj
ect
s
fr
om
diff
eren
t
dataset
s.
They
ob
t
ai
ned
98
.3
3%
accuracy
us
ing
a
ran
do
m
fo
rest
cl
assifi
er
and
96
.3
1%
accuracy
us
ing
the
wav
el
et
distance
m
easur
e
al
go
rithm
wh
il
e
the
two
cl
assifi
ers
tog
et
her
giv
e
accuracy
99
.5
2%.
In
[2
0],
a
no
n
-
fidu
ci
al
ECG
bio
m
et
ric v
erifica
ti
on
syst
em
was pro
po
sed
us
ing
k
e
rn
el
m
et
ho
ds
f
or
5
2
su
bj
ect
s w
it
h
Discrete W
avelet
Tran
sfo
rm
(D
W
T)
fo
r
den
oising
.
They
stud
ie
d
the
per
fo
rm
ance
of
us
ing
Su
pp
or
t
Vector
Ma
chine
(S
VM)
and
the
Linear
Discrim
inant
An
al
ysi
s
(LD
A)
.
A
two
-
le
ad
ECG
sign
al
s
bio
m
et
ric
ver
ific
at
ion
was
pr
op
os
ed
in
[2
1],
autoco
rr
el
at
ion
featur
e
extracti
on
m
et
ho
d
was
us
ed
in
con
j
un
ct
ion
with
diff
eren
t
te
chn
iqu
es
fo
r
red
uction
with
diff
eren
t
wind
ow
le
ng
ths
fr
om
sh
or
t
and
lon
g
-
te
rm
reco
rd
ing
s.
The
resu
lt
s
sh
ow
that
the
reco
gn
it
ion
rates
aff
ect
ed
by
the
reco
rd
ing
and
the
wind
ow
le
ng
ths.
The
ob
j
ect
ive
of
this
research
is
to
app
ly
a
ver
ific
at
ion
bio
m
et
ric
syst
em
us
ing
no
n
-
fidu
ci
al
featur
es
with
Eucli
dean
distance
m
at
ching
m
et
ho
do
log
y.
2.
RESEA
R
CH MET
HO
D
The
pro
po
se
d
bio
m
et
ric
au
thentic
at
ion
s
yst
e
m
con
sist
s
of
th
ree
m
ai
n
ste
ps
pre
-
proce
ssin
g,
featur
e
ex
t
racti
on and
re
du
ct
i
on u
si
ng
D
W
T
, and
m
at
ching
proces
s,
as
s
how
n
i
n
F
i
gure
7.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
3
,
J
une
2020
:
30
0
7
-
30
1
3
3010
Figure
7. Bl
oc
k diag
ram
o
f
th
e pro
posed
sys
tem
2.1. D
atase
ts descr
iptio
n
In
this
resea
rc
h,
90
sub
j
ect
s
are
us
e
d
from
t
wo
da
ta
set
s.
The
first
dataset
is
ECG
-
ID
Da
ta
base
wh
il
e
the
seco
nd
is
the
MIT
-
B
IH
Arrh
yt
hm
ia
Database
[22].
The
num
ber
of
te
st
su
bject
s
us
e
d
in
the
propose
d
syst
e
m
is 7
2
s
ubj
ect
s
inclu
des
11 una
utho
rized s
ubj
ect
s.
2.2. Pre
-
proc
essing
The
first
datas
et
has
filt
ered
sign
al
s
s
o
it
use
d
directl
y
w
hi
le
the
Butt
erwor
t
h
filt
er
was
app
li
ed
on
sign
al
s
from
t
he
sec
ond
dataset
.
The
Pa
n
a
nd
Tom
pk
ins
al
gorithm
was
app
li
ed
f
or
R
peak
detect
io
n
[
23]
.
The
cy
cl
e
le
ng
th
was
fi
x
e
d
at
20
0
sam
pl
es
becau
se
fe
at
ur
es
vect
or
s
m
us
t
hav
e
an
eq
ual
le
ng
th
fo
r
al
l
su
bject
s.
A
norm
alizat
ion
process
int
o
the
range
f
ro
m
0
to
1
is
a
pp
li
ed
to
the
am
plitu
de
of
al
l
poi
nts
f
or
each R
-
R
cyc
le
.
2.3.
Fe
at
ure
e
xt
r
act
i
on
Discrete
Wav
e
le
t
Deco
m
po
si
ti
on
was
a
pp
li
ed
as
a
featu
re
extracti
on
m
e
thod
in
e
ach
R
-
R
cy
cl
e
by
us
in
g
Da
ubec
hies
wa
velet
s
(db8)
with
five
-
le
vel
dec
om
po
sit
ion
as
sh
ow
n
in
F
igure
8.
Sele
ct
ing
the
Daubec
hie
s
fam
il
y
in
this
work
de
pend
ing
on
it
s
sh
a
pe
an
d
ene
rg
y
sp
ect
r
um
wh
ic
h
cl
os
e
to
th
e
ECG
sign
al
wh
il
e
s
el
ect
ing
db
8
f
r
om
pr
evio
us
work
w
hich
ge
ts
good
res
ults
than
oth
e
rs.
The
total
co
ef
fici
ents
nu
m
ber
after
deco
m
po
sit
io
n
is
27
2
as
sho
wn
in
Fig
ure
8.
Fig
ur
e
9.
S
hows
the
sel
ec
te
d
R
-
R
interv
al
and
Figure
10
s
hows
th
at
the
wa
velet
coe
ff
ic
ie
nts
afte
r
100
a
r
e
ap
pro
xim
a
te
l
y
cl
os
ed
to
ze
r
o
s
o
t
ho
se
coe
f
fici
ents
are
neg
le
ct
ed
(d1 an
d d
2) an
d t
he othe
r
c
oeff
ic
ie
nts (
10
4) we
re take
n
i
n our e
xp
e
rim
ent.
Figure
8. The
5
-
le
vel d
isc
rete wa
velet
d
ec
om
po
sit
ion
us
i
ng
Daubec
hies
wav
el
et
s '
db8'
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Perso
nal i
de
ntit
y veri
fi
cation ba
s
ed
ECG
b
i
omet
ric
u
si
ng non
-
fi
duci
al feat
ur
es
(
M
ar
wa
A
. Elsha
hed
)
3011
Figure
9. The
s
el
ect
ed
R
-
R i
nt
erv
al
Figure
10. Wa
velet
co
e
ff
ic
ie
nts
of
t
he
sel
ec
te
d
R
-
R
inter
val
2.4.
A
ut
he
nt
i
cat
i
on
s
tr
at
e
gy
The
Eucli
de
an
distance
al
gor
it
h
m
is
us
ed
f
or
ver
i
ficat
ion
de
ci
sion
s.
E
uclidean
distance
cal
culat
ion
s
are
perform
ed
exactl
y
as foll
ows:
Let
P
(i
)
is t
he
store
d patt
ern f
eat
ur
e m
at
rix
of size
:
×
:
(
)
=
[
1
,
1
⋯
1
,
⋮
⋱
⋮
,
1
⋯
,
]
(1)
If
N
is
the
nu
m
ber
of
sub
j
e
ct
s
so
there
w
her
e
N
sto
red
patte
rn
m
at
ric
es
P
(i
).
P'
(i)
i
s
the
sam
ple
featur
es
m
at
rix
def
i
ne
d as:
′
(
)
=
[
′
1
,
1
⋯
′
1
,
⋮
⋱
⋮
′
,
1
⋯
′
,
]
(2)
The
dista
nce
betwee
n
the
a
tt
ribu
te
s
of
a
sam
ple
and
fe
at
ur
es
m
at
rix
of
a
n
in
div
id
ua
l
i
is
statist
i
call
y
com
pu
te
d usin
g:
(
)
=
[
|
1
,
1
−
′
1
,
1
|
⋯
|
1
,
−
′
1
,
|
⋮
⋱
⋮
|
,
1
−
′
,
1
|
⋯
|
,
−
′
,
|
]
(3)
The
dista
nce
s
cor
e
for
an
in
div
id
ual
is
cal
culat
ed
f
ro
m
t
he
su
m
of
Eu
cl
idean
dista
nc
es
betwee
n
f
eat
ur
e
at
tribu
te
s as:
=
∑
|
,
−
′
,
|
=
1
(4)
The
m
ean
of
di
sta
nce
score
f
or
a
n
in
div
id
ua
l
(i)
shou
l
d
be
cal
culat
ed
due
to
the
var
ia
ti
on
s
in
EC
G
data
set
sign
al
s a
s
f
ollo
ws:
=
1
∑
=
1
(5)
Sm
a
ll
distance
sco
re
value
ind
ic
at
es
go
od
m
at
ching
wh
il
e
la
r
ge
di
st
ance
sco
re
value
i
nd
ic
at
e
s
poor
m
at
ching
[
14]
.
To
get
an
acc
ur
at
e
decisi
on
a
thres
ho
l
d
is
cal
culat
ed
f
rom
the
distance
s
of
t
he
databa
se
an
d
the sam
ple f
e
at
ur
e
m
at
rix
fo
r
each test
ed
in
di
vid
ual.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
3
,
J
une
2020
:
30
0
7
-
30
1
3
3012
3.
E
X
PERI
MEN
T AND
RES
U
LT
S
90
sub
j
ect
s
ar
e
us
e
d
in
the
enrolm
ent
pr
oc
ess
to
c
reate
the
syst
em
dat
abase.
Da
ubec
hies
wa
vele
t
'
db
8'
as
a
fea
ture
e
xtracti
on
m
et
ho
d
is
use
d.
T
he
E
uclidean
distance
al
gorithm
is
us
e
d
as
a
m
atch
in
g
al
gorithm
.
The
perform
ance
of
the
pro
posed
ECG
ve
rificat
ion
syst
e
m
is
est
i
m
at
e
d
by
cal
culat
ing
s
om
e
m
et
rics as the
accuracy
or ve
rificat
ion rate
.
v
(
%
)
=
(6)
Be
sides,
s
om
e
m
e
tric
s
are
use
d
s
uc
h
as
r
ecal
l,
pr
eci
si
on,
a
nd
F
-
sc
or
e
to
e
valuate
t
he
pro
po
s
e
d
syst
e
m
p
erform
ance (6)
.
Sen
sitivi
t
y
(
%
)
=
TP
TP
+
FN
×
100
(7)
Specificit
y
(
%
)
=
TN
FP
+
TN
×
100
(8)
(
%
)
=
+
∗
100
(9)
−
=
2
∗
∗
(
+
)
(10)
W
he
re,
T
P
de
no
te
s
the
nu
m
ber
of
tr
ue
pos
it
ive
sa
m
ples,
TN
de
note
s
th
e
nu
m
ber
of
tr
ue
ne
gative
sa
m
ples
,
FP
denotes
t
he
nu
m
ber
of
false
-
po
sit
iv
e
sam
ples
an
d
F
N
de
note
s
the
num
ber
of
false
-
negat
ive
sam
ples
[2
4,
25
]
.
The
num
ber
of
te
st
subj
ect
s
us
e
d
i
n
the
propose
d
syst
e
m
is
72
subj
ect
s
incl
ud
e
s
11 una
uth
or
iz
e
d
s
ubj
ect
s.
T
he
foll
ow
i
ng T
ab
le
1
s
umm
arize
s the
perf
or
m
a
nce
of the syst
e
m
.
Table
1.
T
he
P
erfor
m
ance of
the
pr
opos
e
d
s
yst
e
m
Para
m
eters
Percent
(%)
Verificatio
n
Rate
9
4
.44
Sen
sitiv
ity
9
5
.08
Sp
ecif
icity
9
0
.9
Precisio
n
9
8
.3
F
-
Sco
re
9
6
.66
4.
CONCL
US
I
O
N
So
m
e
bio
m
e
tric
trai
ts
cou
ld
be
im
it
at
ed
but
the
ECG
sig
nal
is
consi
dered
as
a
real
-
ti
m
e
trai
t
that
ind
ic
at
es
that
the
perso
n
is
al
i
ve
an
d
pr
e
sent
by
him
se
lf.
A
ver
ific
at
io
n
bi
om
et
ric
s
yst
e
m
based
ECG
si
gn
al
is
pro
po
se
d
in
thi
s
researc
h.
Da
ub
ec
hies
wa
ve
le
t
'
db
8'
as
a
featur
e
ext
racti
on
m
et
ho
d
is
use
d
an
d
the
E
uc
li
dean
distance
al
gori
thm
fo
r
ver
i
ficat
ion
decisi
on.
90
s
ubj
e
ct
s
are
us
ed
t
o
buil
d
t
he
syst
em
data
base
,
72
sub
j
ec
ts
fo
r
te
sti
ng
our
ve
r
ific
at
ion
bi
om
e
tric
syst
e
m
.
The
syst
e
m
per
form
ance
is
evaluated
by
cal
cu
la
ti
ng
so
m
e
m
et
rics.
The
ob
ta
ine
d
f
-
Sc
or
e
is
96.66
%.
The
pro
pose
d
syst
em
is
m
or
e
su
it
able
f
or
a
la
rg
e
num
ber
of
sub
j
ect
s
wh
ic
h
giv
es
it
s
decisi
on
i
n
a
ver
y
s
hort
tim
e
with
good
acc
urac
y
du
e
to
usi
ng
the
ver
ific
at
io
n
ap
proac
h
ba
sed
on
non
-
fid
ucial
f
e
at
ur
es.
REFERE
NCE
S
[1]
Ze
esha
n
Hass
an
,
S
y
ed
Om
er
Gila
ni
and
Mohs
in
Jam
il
,
“
Revi
e
w
of
Fiduci
al
a
nd
Non
-
Fiduci
al
Te
chni
qu
es
of
Feat
ure
Ex
traction
i
n
ECG
base
d
Biom
et
ric
S
y
s
te
m
s,”
Indian
Jo
urnal
of
Sci
ence
and
Technol
ogy
,
vol
.
9,
no.
21
,
Jun.
2016
.
DO
I: 10.17485/
i
jst/
20
16/v9i
21/94841
[2]
Manjuna
th
sw
amy
B.
E.,
Appaj
i
M
Abhishek,
Thri
ven
i
J,
Ven
ugopal
K
R,
a
nd
L.
M.
Patn
a
ik,
“
Multi
m
odal
Biom
et
ric
Auth
ent
i
ca
t
ion
using
E
CG
and
F
inge
rprin
t,
”
In
t
ernati
onal
Jour
nal
of
Computer
Appl
ic
a
ti
ons
(
0975
-
8887
)
,
vol
.
111
,
no
.
13
,
Fe
b.
2015
[3]
Ian
McAte
er
,
Ahm
ed
Ibra
him,
Guanglou
Zhe
ng
,
W
enc
heng
Yan
g
and
Crai
g
Val
li
,
“
Inte
gr
at
ion
o
f
Biom
et
ric
s
and
Stega
no
gra
ph
y
:
A Com
pre
hensive
Rev
ie
w,
”
Te
ch
nologi
es
,
vol
.
7
,
no.
34
,
2019
.
D
OI:10.
3390/tech
nologi
es702003
4
[4]
VP
S
G4,
“
Princ
ipl
es
of
ECG
,”
Vascul
ar
Posit
i
oning
Syste
m
G4,
Vascul
ar
Posit
ioni
ng
Te
chnology
Educ
at
ion
,
[Online
]
.
Avai
lable:
htt
p
:/
/www
.
arr
owvascul
ar
.
c
om
/vpseduc
at
io
n/pre
-
tr
ia
l
/pri
nc
i
ple
s
-
of
-
ec
g
.
html
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Perso
nal i
de
ntit
y veri
fi
cation ba
s
ed
ECG
b
i
omet
ric
u
si
ng non
-
fi
duci
al feat
ur
es
(
M
ar
wa
A
. Elsha
hed
)
3013
[5]
Ahm
ed
Younes
Shdefa
t,
Moon
-
I
l
Joo,
Sung
-
Hoon
Choi,
Hee
-
Ch
eol
Kim
,
“
Util
izing
ECG
W
ave
form
F
ea
ture
s
as
New
B
iometri
c
Authent
i
ca
t
ion
Method,
”
Int
ernati
onal
Journal
of
El
e
ct
rica
l
and
Computer
Engi
nee
ring
(
IJE
CE
)
,
vol.
8
,
no
.
2
,
pp
.
658~665
,
Apr.
2
018.
DOI
:
10
.
11
591/i
jece
.
v8i2
.
p
p658
-
665
.
[6]
Vane
ss
a
Novae
s
Barr
os,
“
The
h
e
art
c
y
cle:
rev
i
ew
,
”
MO
J
Women’
s
Healt
h
,
vol
.
8
,
no.
1
,
2019
.
[7]
W
oo
-
Hy
uk
Jung
1
and
Sang
-
Go
og
Le
e,
“
ECG
Ide
nti
f
ic
a
ti
on
B
ase
d
on
Non
-
Fiduci
a
l
Feat
ure
Ext
ra
ct
i
on
Us
in
g
W
indow
Removal
Me
thod,”
MD
PI
journal
,
vo
l.
7
,
no
.
11
,
Nov
.
2
017
.
DO
I:
10
.
33
90/a
pp7111205
.
[8]
Moham
ad
O.
Diab,
Ala
a
Seif
,
Moh
amad
El
-
A
bed
and
Mahe
r
Sabbah,
“
Indiv
idua
l
Id
entificat
ion
Us
ing
ECG
Signal
s,”
Journa
l
of
Computer
an
d
Comm
unic
ati
o
ns
,
vol. 8, pp. 74
-
80,
2018
.
DO
I: 10.4236/
j
cc.201
8.
61008.
[9]
Di
W
ang,
Yujuan
Si,
W
eiy
i
Yan
g,
Gong
Zha
ng
and
Jia
Li,
“
A
Novel
El
e
ct
ro
cardiogra
m
Bi
om
et
ric
Ide
n
ti
fi
catio
n
Method
Based
o
n
T
emporal
-
Fr
eq
uency
Auto
enc
o
ding,
”
Elec
troni
cs
,
2019
.
DO
I:
1
0.
3390/electroni
cs8060667
.
[10]
Ţa
ră
lungă
Drag
oș
Danie
l
and
Mihae
l
a
Nea
gu
,
“
Canc
el
l
ing
Ha
rm
onic
Pow
er
Li
ne
Int
erf
er
ence
in
Biopot
en
tials,
”
book
cha
p
te
r
,
In
t
ec
h
Ope
n
,
2018
.
DO
I:
10.
5772/i
n
te
chop
en.
74579
.
[11]
Pankaj
Sare
en
,
“
Biom
et
ric
s
-
Intr
oduct
ion
,
Charact
er
isti
cs,
B
asic
te
chni
qu
e,
i
ts
T
y
p
es
an
d
Vari
ous
Perform
anc
e
Mea
sures,”
Inte
r
nati
onal Journal
of Emerging Re
search
in
M
anag
eme
nt
&
Technol
ogy
,
vol
.
3
,
no
.
4,
2014
.
[12]
Sus
hm
a
Jaiswal,
Sarita
Singh
B
hada
u
ri
a
a
nd
Ra
kesh
Singh
Jado
n,
“
Biom
et
ri
c:
Case
Stud
y
,
”
Jo
urnal
of
Globa
l
Re
search in Co
mputer
Scienc
e
,
vol.
2
,
no
.
10
,
O
ct
.
2011
.
[13]
Ahm
ed
M.
Hama
d,
R
asha
Sal
ah
El
ha
d
ar
y
,
Ahm
ed
Om
ar
El
kh
atee
b,
“
Multi
m
odal
Biom
et
r
ic
Id
e
nti
ficat
ion
Us
ing
Fingerpr
int,
Fa
c
e
and
Iris
R
ec
og
nit
ion,”
In
te
rnat
i
onal
Journal
of
I
nformation
Sc
ience
and
Int
el
l
igent
Syst
em
,
2014
.
[14]
Yogendra
Nara
i
n
Singh,
S.
K.
S
ingh,
“
Eva
luatio
n
of
El
ec
tro
ca
rd
iogra
m
for
Biom
et
ric
Authentic
at
ion
,
”
Journal
of
Informat
ion
Se
c
urity
,
vol
.
3,
no
.
pp.
39
-
48
,
2012
.
[15]
Ahm
ed
F.
Hu
sse
in,
Abbas
K.
AlZuba
idi,
Ali
Al
-
Ba
y
aty
and
Qai
s
A.
Haba
sh,
“
An
IoT
Rea
l
-
Ti
m
e
Biom
et
ri
c
Authent
i
ca
t
ion
S
y
stem
Based
on
ECG
Fiduci
a
l
Ext
r
ac
t
ed
Feat
ure
s
U
sing
Discre
te
Co
sine
Tra
nsform
,
”
arXiv
:
1708
.
081
89
,
2017
.
[16]
Jam
ee
l
K
adhi
m
Abed
and
Nahra
in
N.
Abd,
“
EC
G
Biom
e
tri
c
Veri
ficati
on
b
y
usin
g
PC
A,”
Inte
rnational
Journal
o
f
Engi
ne
ering
Tr
e
nds and
Technology
(
IJE
TT
)
,
v
ol.
65
,
no
.
1
,
Nov
.
2018.
[17]
S
ugondo
Hadi
y
oso,
Suci
Aulia
and
Achm
ad
Riz
a
l,
“
One
-
Lead
El
ectroc
ard
iogr
am
for
Biom
et
ric
Authent
i
ca
t
ion
using
Ti
m
e
Ser
i
es
Anal
y
sis
and
Support
Vec
tor
Mac
hine,”
(
IJACSA)
Inte
rnatio
nal
Journal
of
A
dvanc
ed
Comput
er
Sci
en
ce and
App
li
cations
,
vo
l. 10
,
no
.
2
,
2019
.
[18]
Nikit
a
Sam
ari
n
,
“
A
Ke
y
to
Y
our
Hea
r
t
:
Bio
m
et
ric
Auth
e
ntication
Based
o
n
ECG
Signal
s,
”
Project
Repo
r
t
Com
pute
r
Scie
n
ce
Schoo
l
of
Inf
orm
at
ic
s Univer
sit
y
of Edinburg
h
,
2018
.
[19]
Robin
Ta
n
and
Mare
k
Perkows
ki,
“
Towa
rd
Im
proving
Elec
tro
ca
rdiogr
am
(EC
G)
Biom
et
ric
V
eri
fi
ca
t
ion
using
Mobile
Sensors
: A
Two
-
Stage Cl
assif
ie
r
Appro
ach,”
S
ensors
,
vol
.
17,
no.
2
2017
.
DO
I:
10.
3390/s
17020410
.
[20]
M
Heja
z
i,
SA
R
Al
-
Hadda
d,
YP
Singh,
SJ
Hashim
and
AF
A
Aziz
,
“
ECG
biomet
ric
aut
hen
ti
c
at
io
n
base
d
on
non
-
fiduc
i
al
appr
oa
c
h
using ke
rn
el m
et
hods,”
Digit
a
l Signal Proce
ss
in
g
,
vol
.
52
,
pp
.
72
-
86,
Ma
y
2016.
[21]
M
Heja
z
i,
SA
R
Al
-
Hadda
d,
SJ
Hashim
,
AF
A
A
zi
z
and
YP
Sing
h,
“
Feat
ur
e
l
evel
fusion
for
b
iometric
ver
ifica
t
ion
with
two
-
lead
E
CG
signal
s,”
I
E
EE
12th
In
te
rnat
ional
Coll
oqu
iu
m
on
Signal
Pro
ce
ss
ing
&
Its
Appl
ic
at
ions(
CSPA
)
,
pp.
54
-
59
,
2016
.
[22]
“
Phy
sioBank
A
TM
,”
Phy
sioN
et
,
[Online
]
Avai
lable:
htt
ps:
//
ar
chive.
ph
y
sione
t.
org
/c
gi
-
bin
/atm
/ATM
[23]
Jiapu
Pan
and
W
il
li
s
j.
Tompkins,
“
A
Rea
l
-
Tim
e
Qrs
Dete
ct
i
on
Algorit
hm
,
”
IEE
E
Tr
ansactions
on
Bi
omedic
al
Engi
ne
ering
,
vol
.
32
,
no
.
3
,
Mar
.
1985.
[24]
Renuka
Jos
hi
,
“
Acc
ura
c
y
,
Pre
ci
s
ion,
R
ecal
l
&
F1
Score
:
Int
erp
retati
on
of
Perform
anc
e
Mea
sures
,
”
2016.
[Online
]
.
Avail
ab
le
:
ht
tp
s://
blog.exsil
io
.
c
om
/a
ll
/acc
ur
acy
-
pre
ci
sion
-
r
ecal
l
-
f1
-
score
-
interpre
ta
ti
on
-
of
-
per
for
m
anc
e
-
m
ea
sures
/
[
Acc
essed
5
/8/
2
019]
.
[25]
Afza
l
Godil
,
R
oger
Bostel
m
an
,
W
il
l
Shack
le
fo
rd,
Tsa
i
Hong
and
Michael
Sh
nei
er
,
“
Perform
anc
e
M
et
ri
cs
fo
r
Eva
lu
at
ing
Obj
e
ct
and
Hum
an
Detect
ion
and
Tr
a
cki
ng
S
y
st
ems
,
”
Nati
onal
Insti
tu
te
of
Standards
and
Technol
ogy
,
Jul.
2014.
Evaluation Warning : The document was created with Spire.PDF for Python.