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.
6
,
Decem
ber
2020
, p
p.
5687
~
5693
IS
S
N:
20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v
10
i
6
.
pp
5687
-
56
93
5687
Journ
al h
om
e
page
:
http:
//
ij
ece.i
aesc
or
e.c
om/i
nd
ex
.ph
p/IJ
ECE
Motion
artifacts
reduc
tion
in card
iac pu
lse sign
al
ac
qu
ired
from
vid
eo
im
agin
g
Mur
thad
Al
-
Yoonus
1
, Mus
tafa H.
A
lh
ab
i
b
2
, Must
afa
Zuha
er
Na
yef
Al
-
D
aba
gh
3
,
M.
F. L.
Abdu
ll
ah
4
1
Com
pute
r
and
Networks Depa
r
tment,
Pa
it
a
xt T
ec
hni
ca
l
Priva
te
Instit
ute
-
Erb
il, K
RG,
Ira
q
2
Com
m
unic
at
ion
and
Com
puter Enginee
r
ing
Dep
a
rtment,
C
iha
n
Univer
sit
y
-
Erbi
l
,
KRG
,
Ira
q
3
Depa
rtment of
Com
pute
r
Scie
n
ce
,
Coll
ege of
S
ci
en
ce,
Know
le
d
ge
Univer
si
t
y
,
Univer
sit
y
Park
,
Erbi
l
EBL 4460
15,
Kurdistan
Re
gion,
Ir
aq
1,4
Facul
t
y
of Electrical a
nd
Elec
t
ronic
Engi
ne
eri
n
g,
Univer
si
ti T
u
n
Hus
sein
Onn
Malay
s
ia,
Ma
lays
ia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
A
ug
4
, 2
019
Re
vised
A
pr 13
,
2020
Accepte
d
Ma
y
8
, 2
020
Thi
s
stud
y
exa
m
ine
s
the
poss
ibi
l
ity
of
remotely
m
ea
suring
the
c
ard
iac
puls
e
ac
t
ivi
t
y
of
a
pat
i
ent
,
w
hic
h
cou
ld
be
an
al
te
rn
at
iv
e
te
chn
ique
to
t
he
cl
assic
al
m
et
hod
.
Thi
s
t
y
pe
of
m
ea
sur
ement
is
non
-
inva
sive
.
How
e
ver
,
seve
r
al
li
m
it
ations
m
a
y
det
er
iora
t
e
the
a
cc
ura
c
y
of
the
r
esult
s
,
including
cha
nges
in
ambient
i
ll
um
in
at
ion
,
m
oti
on
ar
t
ifa
c
ts
(MA
)
and
othe
r
in
te
rf
ere
nc
es
tha
t
m
a
y
occ
ur
through
vi
deo
re
cor
ding
.
T
he
pape
r
in
hand
pre
sents
a
new
appr
oac
h
as
a
remed
y
for
th
e
afo
r
ementi
on
e
d
proble
m
in
c
a
rdia
c
pulse
sign
al
s
ex
tract
ed
from
fac
ia
l
video
rec
ordings.
Pa
rti
ti
on
ing
provid
es
the
basis
for
the
pre
sente
d
MA
red
uct
ion
m
et
hod;
th
e
a
cq
uire
d
sign
al
s
a
re
par
t
it
ion
ed
in
to
two
sets
for
ea
ch
sec
ond
an
d
eve
r
y
par
ti
t
io
n
is
shift
ed
to
t
he
m
ea
n
le
v
el
and
th
en
al
l
the
par
ti
t
ions
ar
e
re
combine
d
a
gai
n
in
to
one
si
gnal
,
whi
ch
is
f
oll
owed
b
y
low
-
pass
fil
ter
in
g
for
enha
n
ce
m
ent
.
The
proposed
compare
d
wi
th
ordinar
y
pulse
oximetr
y
Phot
o
ple
th
y
s
m
ogra
phic
(PP
G)
m
et
hod.
T
he
resulted
cor
relati
on
coe
ff
ic
i
ent
was found
(0.
957)
when
calc
ul
at
ed
be
twee
n
the
result
s
of
the
proposed
m
et
hod
an
d
the
ordina
r
y
one
.
Exp
erim
ent
s
were
implemente
d
usi
ng
a
comm
on
ca
m
era
b
y
cr
eating
a
dat
a
se
t
f
rom
11
subjec
ts
.
The
e
ase
of
impl
ementa
t
ion
of
th
is
m
et
hod
with
a
sim
ple
tha
t
ca
n
be
used
t
o
m
onit
or
the
c
ard
ia
c
pulse
rates i
n
both
hom
e
and
t
he
c
li
ni
cal
env
ir
onm
ent
s
.
Ke
yw
or
d
s
:
Ca
rd
ia
c
pu
lse
ra
te
FFT
Illum
inati
on
Lo
w
-
pass fil
te
r
Moti
on artifact
s
Pu
lse
oxim
e
te
r
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
:
Murtha
d
Al
-
Y
oonus
,
Com
pu
te
r
an
d Netw
orks De
pa
rtm
ent
,
Pait
axt Tech
ni
cal
Pr
ivate
I
ns
ti
tute
-
Er
bil
,
100 St
reet,
E
r
bi
l,
KRG
,
I
ra
q.
Em
a
il
: eng
.m
ur
tha
d86@gm
ail.co
m
1.
INTROD
U
CTION
The
Me
as
ur
in
g
an
d
m
on
it
ori
ng
of
hum
a
n
car
diac
pul
se
act
ivit
y
are
an
esse
ntial
top
ic
f
or
the
healt
hcar
e
industry
i
n
bo
t
h
cl
inica
l
a
nd
ho
m
e
en
viron
m
ents.
Cu
rr
e
nt
ly
,
P
hoto
pleth
ys
m
og
ra
ph
ic
(
PPG)
base
d
pulse
oxi
m
et
ry
and
el
e
ct
ro
car
diog
raphy
(ECG
)
repr
esent
the
m
os
t
com
m
on
te
ch
niques
bei
ng
use
d
i
n
m
od
ern
ca
r
diac
pulse
si
gn
al
m
on
it
or
in
g
s
yst
e
m
s,
wh
ic
h
are
both
i
ne
xp
e
ns
i
ve
a
nd
reli
able.
How
ever,
these
m
e
tho
ds
us
e
co
ntact
sens
or
s
that
ha
ve
to
be
at
ta
ched
to
the
body
,
wh
ic
h
po
s
e
s
a
disadv
a
nta
ge
since
these
sens
ors
are
wire
d
to
the
m
easur
in
g
equ
ipm
ent,
m
aking
the
proces
s
of
place
m
ent
and
re
m
ov
al
inco
nv
e
nien
t t
o
the p
at
ie
nt.
T
o
overc
om
e
the p
re
-
sta
te
d
prob
le
m
o
f
sk
in
-
at
ta
ched
sen
s
or
s,
so
m
e recen
t
stud
ie
s
il
lustrate
d
a
nu
m
ber
o
f
i
nnova
ti
ve
co
ntact
le
ss so
l
ution
s
that
m
easur
e ca
rd
ia
c pulse rate
s
u
s
ing
a
cam
era [1
-
7].
The
cam
era
is
us
e
d
to
rec
ord
fa
ci
al
vid
e
os
w
hile
the
ca
r
diac
pulse
act
i
vity
is
ob
ta
i
ne
d
by
f
ur
t
her
sign
al
proces
sing
a
naly
sis
for
acqu
i
rin
g
the
hear
t
beat
act
ivit
y
[8
]
.
This
kin
d
of
co
nta
ct
le
ss
te
chn
ique
fo
r
card
ia
c pulse
pa
ram
et
ers
m
on
it
or
in
g
ca
n
bec
om
e
an
essenti
al
too
l for h
osp
it
al
healt
h
care
ap
plica
ti
on
s
a
s
well
as
in
the
hom
e
env
ir
onm
ent.
These
novel
te
chn
i
qu
e
s
us
e
a
si
m
il
ar
pr
inciple
to
pulse
oxim
e
try
as
a
con
ta
c
t
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
.
6
,
Dec
e
m
ber
2020
:
5687
-
5693
5688
PPG,
w
her
e
th
e
changes
in
s
kin
col
or
ref
le
ct
s
the
var
ia
ti
ons
in
bl
ood
vol
um
e
accord
in
g
to
the
card
i
ovascular
even
ts
[
9,
10]
.
I
n
oth
e
r
word
s
,
the
hea
rt
pulse
act
ivit
ie
s
af
fect
li
gh
t
transm
issi
on
and
re
flect
anc
e
from
the
hu
m
an
sk
in.
Pu
lse
ox
im
e
try
PPG
is
based
in
the
trans
m
issi
on
m
od
e,
wh
il
e
hear
t
ra
te
m
easur
ing
by
us
in
g
a cam
era is b
as
ed on t
he refle
ct
ance m
od
e [
11
-
14]
.
Howe
ver,
the
l
at
er
m
et
ho
d
s
ti
ll
faces
so
m
e
chall
eng
e
s;
includi
ng
m
otion
a
rtifact
s,
il
lum
inati
on
changes
,
an
d
oth
e
r
inter
fer
e
nces
that
m
ay
occ
ur
durin
g
vid
e
o
rec
ordi
ng.
Ov
e
r
t
he
ye
ars,
var
i
ous
ki
nd
s
of
li
te
ratur
e
pro
pose
d
sig
nal
proces
sin
g
al
go
rithm
s
to
acc
omm
od
at
e
fo
r
a
rtifact
inter
fer
e
nce
an
d
no
ise
cancel
la
ti
on
to
enh
a
nce
the
sign
al
s
ob
ta
in
ed
f
ro
m
the
PPG
s
ources.
S
om
e
of
these
m
et
ho
ds
hav
e
be
e
n
pro
po
se
d
for
reducin
g
the
i
nterf
e
re
nce
ca
us
e
d
by
m
oti
on
a
rtifact
s
[
15
-
17]
.
Ma
ny
al
gorithm
s
relied
on
ind
e
pende
nt
co
m
po
nen
t
a
na
ly
sis
(I
C
A)
for
de
com
po
sin
g
th
e
PPG
sig
nals
in
orde
r
to
obt
ai
n
th
e
desire
d
hear
t
rate
sig
nal
[
18,
19]
.
Anothe
r
stud
y
e
xploit
ed
the
hilb
ert
-
huan
g
t
ran
s
f
orm
(H
HT)
as
a
deco
m
po
sit
io
n
m
e
thod
for
reducin
g
t
he
m
otion
of
arti
fact
e
ff
ect
s
in
Photo
plet
hysm
og
raphic
sig
nals.
Her
e
,
the
intri
ns
ic
m
od
e
functi
on
(I
MF
)
c
om
po
ne
nts
are
ob
ta
ine
d
f
r
om
nonlinear
data
by
ap
plyi
ng
em
pirical
m
od
e
dec
om
po
sit
ion,
wh
ic
h
is
al
s
o
us
e
d
f
or
e
xp
l
ori
ng
the
cha
ng
es
in
am
plit
ude
over
tim
e
[20].
Also
,
a
c
om
par
at
ive
anal
ysi
s
is
pr
ese
nted
bet
w
e
en
the
pe
rfo
r
m
ance
of
a
dapt
ive
filt
erin
g
te
chn
i
qu
e
s
a
nd
wav
el
et
tra
nsf
or
m
.
The
ai
m
was
at
a
c
q
u
i
r
i
n
g
P
P
G
s
i
g
n
a
l
s
t
h
a
t
e
n
j
o
y
m
i
n
i
m
u
m
a
r
t
i
f
a
c
t
s
e
f
f
e
c
t
t
o
b
e
u
s
e
d
t
o
a
p
p
r
o
x
i
m
a
t
e
t
h
e
c
a
r
d
i
a
c
p
u
l
s
e
r
a
t
e
.
I
n
t
h
i
s
c
a
s
e
,
o
n
l
y
l
i
m
i
t
e
d
a
p
p
l
i
c
a
t
i
o
n
c
a
n
b
e
d
r
i
v
e
n
f
r
o
m
m
o
t
i
o
n
r
e
d
u
c
t
i
o
n
-
b
a
s
e
d
w
a
v
e
l
e
t
t
r
a
n
s
f
o
r
m
m
e
t
h
o
d
[
2
1
]
.
Anothe
r
com
par
iso
n
can b
e
s
een
in
[
22
]
w
he
re
dif
fer
e
nt
m
otion
a
rtifact
s
wer
e p
r
ocesse
d
by
dif
fer
e
nt
wav
el
et
tra
nsf
or
m
te
chn
iq
ue
s
f
or
the
possi
bili
ty
of
reduc
ti
on
.
The
im
pl
e
m
entat
ion
of
diff
e
re
nt
wa
ve
le
ts
on
PPG
sig
nals
yi
el
ded
S
pO
2
e
stim
at
es
with
on
ly
sli
ght
va
r
ia
ti
on
s,
wh
il
e
adap
ti
ve
filt
ering
has
bee
n
use
d
to
reduce
the
e
ff
e
ct
s
of
m
otion
a
rtifact
s
on
the
PPG
si
gnal
.
A
ver
it
y
of
dif
fere
nt
ap
proac
hes
wh
e
re
im
ple
m
ente
d
tog
et
he
r
t
o
ac
hieve
the
sam
e
ta
sk
,
s
uch
as
ada
ptive
c
omb
filt
e
r
(
ACF
),
infi
nite
im
pu
lse
res
pons
e
(
IIR
)
a
nd
adap
ti
ve
la
tt
ic
e
no
tc
h
filt
er
(
ALNF)
s
uc
h
as
in
[23].
Alte
rn
at
ively
,
the
work
i
n
[24]
s
uggeste
d
a
two
-
sta
ge
m
et
ho
d
for
m
otion
arti
facts
reducti
on:
in
the
fi
rst
sta
ge
,
the
i
nh
e
ren
t
tim
e
and
f
re
qu
e
ncy
do
m
ain
(
FD)
char
act
e
risti
cs
of
the
PP
G
sign
al
are
us
e
d
to
detect
m
oti
on
a
rtifact
s
w
hile
the
seco
nd
sta
ge
a
pp
li
e
d
sig
nal
processi
ng enh
ancem
ent f
eat
uri
ng a
ne
w
m
a
gn
it
ude al
gorithm
b
ased
on a
n
F
D
-
ICA
app
r
oach [
24]
.
The
st
udy
pre
sented
in
[
25
]
pro
posed
a
m
ul
ti
-
wav
el
en
gth
m
otion
ar
ti
facts
reducti
on
al
gorithm
fr
am
ewo
r
k
for
wr
ist
-
w
orn
P
PG
se
nsor
s,
ut
il
iz
ing
the
green
c
om
po
ne
nt
of
the
PP
G
sign
al
for
hea
rt
rate
m
on
it
or
ing
in
add
it
io
n
to
al
locat
ing
an
inf
r
ared
PP
G
sig
na
l
as
the
m
oti
on
re
fer
e
nce.
The
a
uthors
pr
esente
d
four
m
ai
n
ste
ps
f
or
syst
em
reali
zat
ion
:
Moti
on
detect
ion
,
m
otion
r
e
m
ov
al
us
i
ng
con
ti
nu
ous
wav
el
et
trans
form
, ap
pro
xim
at
e H
R est
i
m
ation
a
nd s
ign
al
reconstr
uc
ti
on
.
The
al
go
rithm
pr
ese
nte
d
in
t
his
pa
pe
r
dem
on
st
rates
a
ne
w
m
et
ho
d
f
or
m
ot
ion
arti
fact
s
reducti
on
us
in
g
a
rem
ote
ly
-
ob
ta
ine
d
P
ho
t
op
le
thysm
og
ra
ph
(rPP
G)
sign
al
f
ro
m
facial
vid
e
o
r
ecordin
gs
of
el
eve
n
par
ti
ci
pa
nts
us
i
ng
face
detect
ion
al
gorithm
[
26
]
.
T
his
m
et
h
od
ca
n
m
easur
e
the
card
ia
c
pulse
rate
of
a
pe
rs
on
from
a
sh
or
t
di
sta
nce
up
to
t
wo
m
et
ers
by
us
in
g
a
cam
era
connecte
d
to
a
PC.
This
is
fo
ll
owe
d
by
us
in
g
the
MATLAB
(Mat
hWorks
TM
)
so
ftwa
re
for
processi
ng
the
sign
al
da
ta
(sign
al
ex
tract
ion
an
d
f
il
te
r
enh
a
ncem
ent
f
or
m
otion
arti
f
act
s
reducti
on)
.
Finall
y,
the
e
nh
a
nce
d
si
gn
al
with
t
he
fr
e
q
ue
ncy
of
t
he
ca
rd
ia
c
pu
lse
is d
is
pla
ye
d
as
res
ults.
Ele
ven
s
ubj
ec
ts
had
par
ti
ci
pated
in
t
he
exp
e
rim
ents
fo
r
recordi
ng
30
-
sec
ond
vide
os
f
or
ea
c
h
ind
ivi
du
al
.
T
he
contrib
utio
n o
f
this
resea
rch
can
be
s
umm
ar
iz
ed
as
fo
ll
ows
:
The
extracti
on
of
the
car
diac
pu
lse
sig
nal
from
the
gr
ee
n
channel
in
facia
l
vid
eos
,
(usin
g
the
colo
r
RG
B
separ
at
io
n
m
et
hods
sta
te
d
in
the m
et
ho
dolo
gy
secti
on
)
.
Partit
ion
in
g
t
he
green
co
m
ponen
t i
nto
60
pa
rtit
ion
s
(tw
o pa
rtit
ion
s
per a se
cond
).
Partit
ion
s
En
ha
ncem
ent.
Re
com
bin
ing
t
he
e
nh
a
nce
d
pa
rtit
ion
s int
o o
ne
si
gn
al
.
To fin
d
the
car
diac pulse
rate
fr
e
qu
e
ncy,
fast
fourie
r
tra
nsfo
rm
(F
FT)
is
appli
ed
to
g
et
the
d
esi
re
d si
gnal
.
The
arr
a
ngem
e
nt
of
this
rese
arch
will
be
as
fo
ll
ows:
Sect
ion
2
descr
i
bes
the
pr
op
os
e
d
m
et
ho
d
f
or
card
ia
c
pu
lse
s
ign
al
de
te
ct
ion
and
m
otion
a
r
ti
facts
reducti
on
usi
ng
the
green
com
pone
nt
of
the
PP
G
s
ign
al
.
Sect
ion
3
pres
ents
the
ex
per
i
m
ental
resu
lt
s
and
ex
plains
th
e
eff
ect
ive
ness
of
the
pro
pose
d
m
et
ho
d,
re
ga
rd
i
ng
no
ise
reducti
on
in
the
ca
rd
i
ac
pu
lse
sig
na
ls
ob
ta
ine
d
f
rom
vi
deo
rec
or
dings.
Fi
nally
,
Sect
ion
4
dis
cusses
the r
es
ults,
pr
e
sents the
concl
us
io
n
a
nd
offers f
uture rec
omm
end
at
ions.
2.
METHO
D
2.1.
Ex
peri
ment
al setup
I
n
this
researc
h,
el
eve
n
ind
i
viduals
rangin
g
betwee
n
18
-
35
ye
ars
old
wer
e
par
ti
ci
pat
ed
f
or
acq
uiri
ng
the
PP
G
sig
na
ls.
An
ordi
nary
ca
m
era
is
us
ed
to
re
co
rd
t
he
facial
vi
deos
in
true
c
olor
sta
nd
a
rd
(24
-
bit
per
pix
el
RGB
for
m
at
with
3
chan
nels
×
8
bits/
channel)
at
30
fr
am
es
per
second
(
f
ps),
with
a
pix
el
reso
lut
ion
of
720×
404.
The
vid
e
os
wer
e
sa
ved
in
AVI
fo
r
m
at
fo
r
furthe
r
analy
sis
us
i
ng
MATL
AB
s
oft
war
e.
The
rec
ordin
g
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
Motio
n arti
facts red
uctio
n
in
car
dia
c
pulse
s
ignal ac
quire
d
from
vi
de
o
im
agin
g
(
Mu
rth
ad
Al
-
Yoo
nus
)
5689
sessions
to
ok
place
at
di
ff
e
re
nt
tim
es
of
t
he
day.
A
var
yi
ng
am
ou
nt
of
a
m
bient
su
nlig
ht
was
the
only
so
urce
of
il
lum
inati
on
in
this
exp
e
rim
ent.
The
pa
rtic
ipants,
ha
ving
dif
fer
e
nt
sk
i
n
colo
rs,
w
ere
seat
ed
on
a
c
hair
i
n
fron
t
of
the
ca
m
era
at
a
dista
nce
of
ap
pro
xim
at
ely
on
e
m
e
te
r.
D
ur
i
ng
the
recordi
ng,
pa
r
ti
ci
pan
ts
wer
e
aske
d
to sit st
il
l and s
ta
re at the cam
era lens
.
2.2.
C
ardi
ac pul
se sig
na
l e
xt
r
act
i
on
The
si
gn
al
of
interest
f
or
thi
s
w
ork
is
t
he
card
i
ov
asc
ular
pulse
wav
e
that
pr
op
a
gates
thr
ough
out
the
body.
The
Viola
-
J
on
es
a
lgorit
hm
pr
ov
i
des
the
basis
f
or
the
re
gion
of
interest
(R
OI)
sel
ect
ion
f
or
fac
e
detect
ion.
T
he
cam
era
senses
the
c
hanges
in
the
am
ount
of
re
flect
ed
li
gh
t,
w
hich
i
nd
ic
at
es
t
he
tim
i
ng
of
card
i
o
pulse
e
ven
ts
.
Thes
e
even
ts
relat
e
to
vo
l
um
et
ric
chan
ge
s
in
the
R
OI
blood
vess
el
s
durin
g
the
card
ia
c
cy
cl
e.
The
RG
B
channels
f
or
each
f
ram
e
c
on
sist
of
th
ree
com
po
ne
nts;
these
are
r(
t)
,
g(t
)
an
d
b(
t)
for
red,
gr
ee
n
a
nd
blue
res
pecti
vely
.
The
e
xtracted
sig
nal
f
r
om
t
he
gr
ee
n
col
or
sens
or
is
t
he
best
si
gn
al
t
o
sho
w
the car
diac act
ivit
y, bein
g at
te
nu
at
e
d by m
oti
on artifact
noise
.
The
pr
es
um
pti
on
is
t
hat
the
vo
l
um
et
ric
blo
od
va
riat
ion
s
r
esult
in
facial
colo
r
cha
nges
and
he
nce
facial
li
gh
t
int
e
ns
it
y
var
ia
ti
ons.
Fig
ur
e
1
sh
ows
the
PP
G
sig
nals
bei
ng
acq
uire
d
from
the
facial
vid
e
o
recordi
ng
m
easur
em
ents
of
two
par
ti
ci
pa
nts
by
us
in
g
a
ca
m
era
conne
ct
e
d
to
a
PC
at
diff
ere
nt
ti
m
es
of
the
day.
T
he
ob
s
er
ved
sm
all
rip
ple
in
the
green
c
om
po
nen
t
is
acc
ord
in
g
to
t
he
ca
r
diovasc
ular
e
ve
nts;
howe
ver,
the
r
ed
li
ne
ind
ic
at
es
the
sign
ific
a
nt
a
m
plit
ud
e
changes
res
ulte
d
from
un
desir
able
m
otion
ar
ti
facts.
The
si
gn
al
s
ob
ta
ine
d
from
these
gr
ee
n
com
po
ne
nts
show
evi
de
nt
card
ia
c
pulse
cha
racteri
st
ic
s
an
d
accom
m
od
at
es
the
m
otion
of
arti
facts,
s
uch
as
sm
a
ll
facial
m
ot
ion
a
nd
il
lum
inati
on
cha
ng
e
s.
T
he
ai
m
her
e
i
s
at
r
ed
ucin
g
m
otion
a
rtifact
s to
enhance
this si
gn
al
a
nd
on
ly
yi
el
d
the cle
ar
card
ia
c
pulse
s
ign
al
.
Figure
1. Tw
o PPG
car
diac
pulse
sig
nals
wi
th m
otion
artif
a
ct
s noiseo
f
tw
o partic
ipa
nts,
t
he
sig
nals
were t
aken at
a
dis
ta
nce 1 m
et
er
2.3.
M
oti
on a
rtifa
c
ts
re
duc
t
ion
Sign
al
en
ha
nc
e
m
ent
and
m
ot
ion
arti
facts
el
im
inati
on
requi
res
the
i
m
ple
m
entat
ion
of
a
ne
w
m
e
tho
d
wh
ic
h
is
based
on
pa
rtit
ion
in
g
the
sign
al
,
s
hifti
ng
par
ti
ti
ons
to
the
m
ean
le
vel,
reco
m
bin
ing
them
and
then
app
ly
in
g
lo
w
-
pass
filt
er
(LPF)
to
obta
in
t
he
fi
nal
sig
nal
.
Fig
ur
e
2
sho
ws
the
acq
uir
ed
sig
nal
as
well
as
the enha
ncem
e
nt pr
ocesses.
Figure
2. (a
)
T
he (RO
I)
is
au
t
om
atical
ly
cro
pp
e
d usi
ng
fac
e d
et
ect
io
n,
(b)
t
he
R
OI
dec
om
po
ses into
t
he
RGB
channe
ls a
nd t
he raw
RGB tr
aces, (c
)
t
he gr
een c
om
po
ne
nt
is en
ha
nced,
(d)
F
FT is a
ppli
ed
to
rec
ov
e
r
t
he fre
qu
e
ncy
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
.
6
,
Dec
e
m
ber
2020
:
5687
-
5693
5690
The
sta
nda
rd
c
ard
ia
c
pu
lse
ra
te
of
a
n
ad
ult
ranges
from
60
to
100
beats
per
m
inu
te
;
m
eanin
g
th
at
ever
y
sec
ond
con
ta
in
s
one
or
tw
o
pulse
s
a
ccordin
g
to
th
e
par
ti
ci
pa
nt’s
healt
h
c
onditi
on.
T
he
par
ti
ti
one
d
sign
al
is
obta
in
ed
from
the
green
com
ponen
t
ov
e
r
30
seco
nds
an
d
is
div
i
de
d
into
60
par
t
it
ion
s
(tw
o
pa
r
ti
ti
on
s
per
seco
nd)
to
sepa
rate
e
very
poss
ible
pulse
in
one
par
t
it
ion
a
nd
the
n
sh
ift
pa
rtit
ions
to
t
he
m
ean
le
vel.
The
n
al
l
par
ti
ti
on
s
are
reco
m
bin
e
d
again
t
o
fo
rm
the
signa
l.
Last
ly
,
us
ing
a
4H
z
lo
w
-
pa
ss
filt
er
(LPF)
will
resu
lt
in
the
fi
nal
en
han
ce
d
s
ign
al
.
T
he
f
ollow
i
ng
ste
ps
ar
e
app
li
ed
for
r
edu
ci
ng
t
he
effe
ct
of
m
otion
arti
facts
in car
diac
pu
ls
e sig
nals obtai
ned f
ro
m
a v
id
eo reco
rd
i
ng of
the
hu
m
an
fac
e:
Partit
ion
in
g
t
he
green
co
m
ponen
t i
nto
60
pa
rtit
ion
s as
in Fi
gure
3(
a
).
Get the m
ean l
evel f
or eve
ry
par
ti
ti
on as s
ho
wn in t
he fo
rm
ula b
el
ow
:
̅
=
∑
=
1
(
)
(1)
w
he
re
n rep
res
ents the
No.
p
a
rtit
ion
s
Sh
ift t
he
sam
ples of e
ver
y
part
it
ion
to
t
he
m
e
an
le
vel
as in
F
igure
3(b).
, usi
ng
:
(
)
=
(
)
−
̅
(2)
Re
peat
ste
p
3
f
or
al
l
par
ti
ti
ons
,
then
rec
om
bin
e
al
l
the
par
ti
ti
on
s
to
yi
el
d
th
e
cl
ear
card
io
pu
lse
sig
nal
wi
th
the r
e
duct
i
on of ar
ti
facts
as in
Fig
ur
e
3(c
).
Use
l
ow
-
pa
ss fi
lt
er ite
rati
on
(4 tim
es)
of
ord
er
4H
z
for t
he final
ste
p o
f
e
nhancem
ent, F
i
gure
3(d
).
The
proces
sin
g
of
data
requir
es
the
transform
at
ion
into
the
fr
eq
ue
ncy
dom
ai
n
(F
D)
by
Fast
Fo
uri
er
.
Transf
or
m
,
wh
ic
h
will
est
im
at
e
the
fr
eq
uen
cy
of
the
card
ia
c
pu
ls
e
to
facil
it
at
e
sign
al
enh
a
nc
e
m
ent.
At
the
sam
e
tim
e,
a
pu
lse
oxim
et
er
m
easur
es
the
re
fer
e
nce
of
t
he
car
diac
pu
lse
rate.
T
hi
s
pulse
oxim
eter
data
act
s
as
gro
und
tr
uth
that
e
nab
le
s
acc
ur
a
c
y
com
par
ison
and
as
sessm
ent
of
the
ca
m
era
al
go
rit
hm
-
based
m
easur
em
ent m
et
ho
d.
Figure
3. (a
)
P
arti
ti
on
in
g,
(
b)
sh
ifti
ng
par
ti
ti
on
s
to
the m
ean,
(c)
r
ec
om
bin
ing
the
Pa
rtit
ion
s
, (d) LPF
E
nh
a
ncem
ent
3.
E
X
PERI
MEN
TAL RES
UL
TS
In
t
he
e
xp
e
rim
ents,
the
cam
e
ra
wa
s
ad
juste
d
to
be
a
bout
1m
et
er
away
fr
om
the
subj
ec
t
to
colle
ct
RGB
data
wei
gh
t
var
ia
ti
ons
durin
g
vid
e
o
r
ecordin
g.
T
he
exp
e
rim
ents
wer
e
pe
rfor
m
ed
indoor
a
nd
outdoo
r
at
diff
e
re
nt
tim
es
of
the
day.
Th
e
faces
of
the
s
ubj
ect
s
are
co
nsi
der
e
d
to
be
the
reg
i
on
of
in
te
rest,
bein
g
cr
oppe
d
for
analy
sis
after
face
detect
ion.
From
the
e
xp
e
rim
ents,
the
gr
ee
n
sens
or
colle
ct
ed
the
be
st
data
wh
e
re
weig
ht
var
ia
ti
ons
co
rresp
ond
to
ca
rd
i
ov
a
scular
act
iv
it
ie
s.
Also
,
th
e p
rese
nce o
f
othe
r
fluct
uation
s
ources
in
li
ght wit
h
the
obta
ine
d
da
ta
will
introd
uce
no
ise
t
o
th
e
sign
al
.
T
he
m
ot
ion
arti
fact
s
re
du
ct
io
n
m
et
hod
was
a
pp
li
ed
on
the obtai
ne
d
si
gn
al
us
i
ng MA
TLAB s
oft
ware.
Figure
4
s
how
s
the
ty
pical
r
esults
of
the
c
ard
ia
c
p
ulse
re
m
ote
m
easur
em
ent
and
m
oti
on
a
rtifact
s
reducti
on
obta
ined
by
this
m
et
ho
d
f
or
thr
ee
su
bject
s
Fi
gure
4(a)
s
ho
ws
the
sig
nals
in
tim
e
do
m
a
in
with
diff
e
re
nt
am
pli
tud
es
cau
sed
by
li
gh
t
inte
ns
it
y
var
ia
ti
ons
a
c
cordin
g
t
o
t
he
card
i
ov
asc
ular
eve
nts
a
nd
m
oti
on
arti
facts.
Fi
gur
e
4(b
)
s
hows
t
he
re
su
lt
s
of
t
he
arti
facts
re
duct
ion
m
et
ho
d
by
pa
rtit
ion
in
g,
s
hiftin
g
t
o
a
m
ean
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
Motio
n arti
facts red
uctio
n
in
car
dia
c
pulse
s
ignal ac
quire
d
from
vi
de
o
im
agin
g
(
Mu
rth
ad
Al
-
Yoo
nus
)
5691
le
vel,
rec
om
bin
in
g
a
nd
the
n
app
ly
in
g
a
l
ow
-
pass
filt
er
it
er
at
ively
fo
r
eac
h
case.
Fig
ure
4(
c
)
s
hows
t
he
FFT
of
the
en
hanced
sign
al
s
in
freq
uen
cy
do
m
ai
n,
the
FFT
s
pec
tra
cl
early
sh
ow
a
hi
gh
e
r
pe
ak;
this
pea
k
is
at
the freq
ue
ncy
of the ca
rd
ia
c
pu
lse
.
In
orde
r
t
o
te
s
t
the
he
art
rate
m
easur
e
m
ent
accuracy
of
th
is
m
e
tho
d
f
or
m
ot
ion
a
rtifact
re
du
ct
io
n,
11
sub
j
ect
s
pa
rtic
ipate
d
in
th
e
ex
per
im
ents.
The
re
s
ults
of
this
ne
w
hea
r
t
rate
m
easur
ing
m
et
ho
d
c
om
par
e
favor
a
bly
with
the
re
fer
e
nce
res
ults
pr
e
vi
ou
sly
ob
ta
ine
d
from
a
pu
lse
oxim
e
te
r
as
sh
ow
n
in
Tab
le
1.
This
m
e
tho
d
pro
duces
a
sta
bl
e
cor
relat
io
n
coeffic
ie
nt
of
(0.95
7)
cal
cula
te
d
betwee
n
th
e
card
ia
c
puls
e
rate
m
easur
em
ents acqu
i
red he
re a
nd the classi
cal
pulse
ox
im
et
e
r
m
et
ho
d, as s
how
n
i
n
Fi
gure
5.
Figure
4.
(a
)
Si
gn
al
s
in
ti
m
e d
om
ai
n
with d
i
f
fer
e
nt am
pli
tud
es, (b
)
arti
facts
r
e
du
ct
io
n,
(c)
high
pec
ks
at
card
ia
c
puls
e rate in
freq
ue
ncy dom
ai
n
Table
1.
Ca
rd
i
ac p
ulse r
at
e
r
e
su
lt
s co
m
par
e
d wit
h refe
re
nce
r
es
ults
Su
b
jects
Prop
o
sed
M
eth
o
d
Pu
lse Ox
i
m
ete
r
1
55
58
2
72
76
3
63
63
4
88
86
5
83
81
6
60
57
7
60
64
8
95
83
9
86
79
10
56
58
11
69
69
Figure
5. The
c
ard
ia
c
pulse
r
at
e co
rr
el
at
io
n
c
oeffici
ent
betw
een th
e
cam
era m
e
tho
d an
d
t
he
pulse
ox
im
eter
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
.
6
,
Dec
e
m
ber
2020
:
5687
-
5693
5692
4.
DISCU
SSI
ON A
ND CON
C
LUSIO
N
This
pa
pe
r
pre
sents
a
te
ch
nique
f
or
m
otion
arti
facts
reduct
ion
in
ca
rd
ia
c
pu
lse
si
gn
al
s
obta
ined
by
rem
ote
vid
eo
recordi
ng.
T
his
m
et
ho
d
c
on
sist
s
of
t
w
o
sta
ges:
first
ly
,
par
ti
ti
on
in
g
the
si
gn
al
,
sh
ifti
ng
the
par
ti
ti
ons
to
the
m
ean
l
evel
an
d
rec
om
bin
ing
them
back
int
o
on
e
sign
al
;
the
s
econd
sta
ge
is
sign
al
enh
a
ncem
ent
by
low
-
pa
ss
fi
lt
ering
.
In
ord
er
to
te
st
the
perform
ance
of
this
arti
facts
reducti
on
m
et
hod,
the appli
ed
FF
T enha
nces the
car
diac pulse
r
at
e b
y i
de
ntifyi
ng h
i
gh p
ea
ks i
n
the
freq
ue
nc
y do
m
ai
n.
The
rem
ote
pulse
m
easur
em
ent
res
ults
co
m
par
e
fav
ora
bl
y
to
the
resu
l
ts
ob
ta
ine
d
f
r
om
a
reg
ular
pu
lse
ox
im
et
er.
The
correla
ti
on
coe
ff
ic
ie
nt
between
the
resu
lt
s
of
the
two
m
et
ho
ds
is
found
to
be
(
0.957
).
The
te
ch
nique
can
re
duce
t
he
ef
fect
of
m
ot
ion
arti
fac
ts
in
car
diac
pu
lse
PP
G
f
or
sign
al
e
nh
a
nc
e
m
ent.
Fu
rt
her
m
or
e,
this
stu
dy
dem
on
st
rates
that
t
his
m
et
ho
d
ca
n
work
ef
fici
ently
with
vid
e
os
recor
ded
at
diff
e
re
nt
tim
es
of
t
he
da
y.
This
te
ch
ni
qu
e
is
cu
rr
e
nt
ly
under
de
vel
op
m
ent
but
it
can
be
util
iz
ed
la
te
r
f
or
bui
lding
a
r
obus
t,
rem
ote,
accu
rate
ca
rd
ia
c
pulse
rat
e
m
easur
ing
s
yst
e
m
s.
Top
ic
s
f
or
f
uture
re
search
m
ay
include:
creati
ng
a
real
-
tim
e
re
m
ote
m
easur
i
ng
syst
em
based
on
t
hi
s
m
e
tho
d
f
or
m
ulti
-
par
am
et
ers
su
c
h
as
car
dio
pu
ls
e
rate va
riabil
it
y, res
pirato
ry an
d
a
rterial
b
l
ood o
xyge
n
sat
ur
at
ion
.
REFERE
NCE
S
[1]
Poh,
Ming
-
Zhe
r,
Danie
l
J.
McDuff,
and
Rosali
nd
W
.
Pica
rd
,
"N
on
-
cont
a
ct
,
aut
om
a
te
d
ca
rd
iac
pulse
m
ea
surem
ent
s
using vi
deo
ima
ging
and
bli
nd
s
ourc
e
sep
arati
on
,
"
Opti
cs
e
xpre
ss
,
vol
.
18
,
no
.
10
,
pp.
10762
-
1077
4,
2010
.
[2]
Garbe
y
,
et
al.
,
"Contact
-
fre
e
m
ea
surem
ent
of
ca
rd
ia
c
pulse
base
d
on
the
ana
l
y
sis
of
th
ermal
imager
y
,
"
IEE
E
Tr
ansacti
o
ns
Bi
omedi
cal E
ngine
ering
,
vol
.
54,
no
.
8
,
pp
.
14
18
-
1426,
2007
.
[3]
Cenni
ni
,
et
al.
,
"H
ea
rt
ra
te
m
onit
oring
vi
a
r
emote
photoplet
h
y
sm
ogra
ph
y
w
it
h
m
oti
on
artif
a
ct
s
red
u
ct
ion
,
"
Optic
s e
xpre
ss
,
vol.
18
,
no
.
5
,
pp
.
4867
-
4875
,
20
10.
[4]
Le
wandows
ka,
M.,
Rum
iński,
J.,
Koce
jko,
T
.
,
a
nd
Now
ak,
J
.
,
"
Mea
suring
pulse
rat
e
with
a
webc
am
-
a
non
-
cont
a
c
t
m
et
hod
for
evalua
ti
ng
cardiac
ac
t
ivi
t
y
,
"
Fe
d
erate
d
Conf
ere
nc
e
Computer
Sc
i
enc
e
and
Infor
mation
Syste
ms
(
Fe
dCSIS)
,
pp.
4
05
-
410,
2011
.
[5]
Ta
kano
,
and
Yuji
Ohta
,
“
Hea
r
t
r
at
e
m
ea
surem
ent
base
d
on
a
ti
m
e
-
la
pse
image
,
"
Me
dic
a
l
engi
n
eer
ing
and
physic
s,
vol.
29
,
no
.
8
,
pp
.
853
-
857
,
2007
.
[6]
Li
,
X
.
,
Ch
en,
J.,
Zha
o
,
G.
,
and
Piet
ikainen
,
M
,
"Rem
ot
e
hea
r
t
r
at
e
m
e
asure
m
ent
from
face
vid
e
os
under
re
al
ist
i
c
situa
ti
ons,"
Proc
ee
dings o
f
th
e
I
E
EE
conf
ere
n
ce o
n
compute
r
vi
sio
n
and
pat
te
rn re
cogni
ti
on
,
pp
.
42
64
-
4271,
2014
.
[7]
Rouast,
P.
V.,
Adam
,
M.
T.
,
C
hiong,
R.
,
Cornf
orth,
D.
,
Lux,
E
.
,
"Rem
ote
heart
rat
e
m
ea
surem
e
nt
using
low
-
cost
RGB fa
ce vi
d
eo: a
technical
liter
at
ure
rev
ie
w,"
F
rontie
rs
of
Com
pute
rSci
enc
e
,
vo
l.
12
,
no
.
5
,
pp
.
8
58
-
872,
2018
.
[8]
Ibra
him,
W
.,
To
m
ari
,
R.
,
Za
k
aria,
W
.
N.
W
.
,
an
d
Othm
an,
N
,
"N
on
-
cont
act
heart
rate
m
onit
orin
g
ana
l
y
sis
from
var
ious
dist
anc
e
s
with
diff
ere
nt
face
r
egi
ons,"
Inte
rnational
Jo
urnal
of
El
e
ct
ri
cal
and
Computer
Eng
ine
eri
n
g
(
IJE
CE)
,
vol. 7,
no.
6
,
pp
.
3030
-
3036
,
2017
.
[9]
Hass
an,
et
al.,
"
Hea
rt
r
at
e
esti
m
at
ion
using
faci
al
vid
eo:
A
r
eview,"
Bi
omed
ic
a
l
Signal
Pro
ce
ss
i
ng
and
Control,
vol.
38
,
pp
.
346
-
360,
2017
.
[10]
N
.
Ibra
him,
et
a
l
.
,
"A
na
l
y
s
is
of
m
ini
m
um
fac
e
v
ide
o
dur
at
ion
an
d
the
eff
ec
t
of
v
ide
o
compress
ion
to
imag
e
-
base
d
non
-
c
o
n
t
a
c
t
h
e
a
r
t
r
a
t
e
m
o
n
i
t
o
r
i
n
g
s
y
s
t
e
m
,
"
B
u
l
l
e
t
i
n
o
f
E
l
e
c
t
r
i
c
a
l
E
n
g
i
n
e
e
r
i
n
g
a
n
d
I
n
f
o
r
m
a
t
i
c
s
,
v
o
l
.
9
,
n
o
.
1
,
p
p
.
4
0
3
-
4
1
0
,
2
0
2
0
.
[11]
Aart
s,
e
t
al
.
,
"N
on
-
cont
a
ct
h
ea
r
t
r
at
e
m
onit
oring
uti
lizing
ca
m
er
a
photopleth
y
sm
o
gra
ph
y
in
t
he
ne
onat
a
l
in
te
nsiv
e
ca
re
unit
-
A pi
lo
t
stud
y
,
"
Early hu
man de
velopmen
t,
vo
l. 89,
no.
12,
pp.
943
-
948,
20
13.
[12]
Al
-
Naji
,
a
nd
Java
an
Chahl
,
"N
oncont
a
ct
he
art
ac
t
ivi
t
y
m
ea
sure
m
ent
sy
st
em
base
d
on
vide
o
imaging
ana
l
y
sis
,
"
Inte
rnational
Jo
urnal
of Pat
t
ern
Re
cogn
it
ion
and
Arti
f
icial
In
telligence,
vo
l
.
31,
n
o.
2
,
pp
.
175700
1,
2017
.
[13]
Alghoul,
e
t
al.,
"H
ea
rt
ra
te
var
i
abi
lit
y
ex
tra
c
ti
o
n
from
vid
eos
signal
s:
ICA
vs.
EVM
compari
son,"
IEEE
Acce
ss
,
vol.
5
,
pp
.
4711
-
4719,
2017
.
[14]
MasnaniBt
Moham
ed,
et
al
.
,
"
Non
Conta
ct
m
onit
oring
of
he
art
rate
respons
es
to
ta
ste
stim
uli
using
a
vid
eo
ca
m
era,"
Indone
sian J
ournal
o
f El
e
ct
rica
l
Eng
in
ee
ring a
nd
Computer
Sc
ie
nc
e,
v
ol.
18
,
no
.
1
,
pp
.
293
-
300,
2020
.
[15]
Ta
nwee
r
,
e
t
a
l
.
,
"M
oti
on
artifa
ct
red
u
ct
ion
fro
m
PPG
signal
s
during
intens
e
exe
rc
ise
using
fil
t
ere
d
X
-
LM
S,
"
IEE
E
Inte
rnat
io
nal
Symposium
on
Circui
ts
and
Syste
ms
(
ISCAS)
,
pp
.
1
-
4
,
2017
.
[16]
Dao,
et
a
l.,
"A
robust
m
oti
on
art
ifact
de
t
ec
t
ion
al
gori
th
m
for
ac
cur
at
e
det
ection
of
hea
rt
ra
te
s
fro
m
photopl
et
h
y
sm
ogra
phic
sign
al
s
using
ti
m
e
–
fre
q
uency
spe
ct
ra
l
f
ea
tur
es,
"
IE
EE
j
ournal
of
biome
dic
al
and
health
inf
orm
ati
cs,
vol
.
21,
no.
5,
pp.
12
42
-
1253,
2016
.
[17]
Nordin,
W
.
David
Hair
ston,
a
nd
Danie
l
P.
F
err
is
,
"D
ual
-
e
lectr
ode
m
oti
on
art
if
ac
t
c
ancel
l
a
ti
on
for
m
obil
e
el
e
ct
roe
n
ce
ph
al
o
gra
ph
y
,
"
Journal
of
n
eural engi
ne
ering
,
vo
l.
15,
n
o.
5
,
2018
.
[18]
Y.
Jianc
hu
,
and
S.
W
arr
en,
"A
Short
Stud
y
to
As
sess
the
Potentia
l
of
Ind
epe
nden
t
Com
ponent
Anal
y
s
is
for
Motio
n
Artifa
c
t
Separ
at
i
on
in
W
ea
r
abl
e
Puls
e
Oxim
et
er
Signal
s,"
IE
EE
Confe
renc
e
of
t
he
Engi
n
ee
ring
in
Me
di
ci
n
e
and
Bi
ology Society
,
pp.
3585
-
3588
,
2005
.
[19]
B.
S.
Kim
,
and
Sun
K.
Yoo
,
"M
oti
on
artifa
ct
red
uction
in
pho
topl
et
h
y
sm
ogra
p
h
y
using
inde
p
e
ndent
componen
t
ana
l
y
sis,"
IEEE T
rans
.
Bi
omed.
Eng.
,
vol
.
53
,
no
.
3
,
pp
.
566
-
568
,
2006.
[20]
Raghura
m
,
e
t
al
.,
"H
HT
base
d
signal
de
compos
it
ion
for
red
u
ction
of
m
oti
on
a
rti
fa
ct
s
inphot
op
le
th
y
sm
ogra
ph
i
c
signal
s,"
Instrum
ent
ati
on
and
M
e
asur
eme
nt
Tech
nology
Con
fe
ren
ce
(
I2MTC)
,
pp.
1730
-
1734,
201
2.
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
Motio
n arti
facts red
uctio
n
in
car
dia
c
pulse
s
ignal ac
quire
d
from
vi
de
o
im
agin
g
(
Mu
rth
ad
Al
-
Yoo
nus
)
5693
[21]
Foo
and
Jong
Y
ong
A
,
"Com
par
ison
of
wave
let
tra
nsform
at
ion
a
nd
ada
p
ti
ve
fi
lt
e
ring
in
restor
ing
artifa
c
t
-
indu
ce
d
ti
m
e
-
relate
d
m
eas
ure
m
ent
,
"
B
iomed.
S
ignal P
roce
ss
.
Control,
vol
.
1,
pp
.
93
-
98
,
20
06.
[22]
Raghura
m
,
et
al
.
,
"Eva
luation
of
wave
lets
for
red
uct
ion
of
m
oti
on
art
ifacts
inphotoplet
h
y
sm
ogra
p
hic
signal
s,"
10t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
I
n
f
o
r
m
a
t
i
o
n
S
c
i
e
n
c
e
s
S
i
g
n
a
l
P
r
o
c
e
s
s
i
n
g
a
n
d
t
h
e
i
r
A
p
p
l
i
c
a
t
i
o
n
s
(
I
S
S
P
A
)
,
p
p
.
4
6
0
-
46
3
,
2
0
1
0
.
[23]
Le
e
,
e
t
al
.
,
"A
dapt
ive
comb
fil
t
er
ing
for
m
oti
on
a
rti
fa
ct
r
educ
t
ion
from
PPG
with
a
struct
ur
e
of
ad
apt
iv
e
lattice
IIR
notc
h
f
il
t
er,
"
Annual
In
te
rn
at
ion
a
l
Confer
ence
En
gine
er
ing
in
Me
di
ci
n
e
and
Bio
lo
g
y
Societ
y
,
pp.
7
937
-
7940,
2011
.
[24]
Krishnan,
et
al
.
,
"Two
-
stage
appr
oac
h
for
det
e
c
ti
on
and
red
uc
tion
of
m
oti
on
ar
ti
fa
ct
s
in
photopl
et
h
y
sm
ogra
ph
i
c
dat
a
,
"
IE
EE Tr
ansacti
ons
B
iome
dic
al
Engi
n
ee
rin
g,
vol
.
57
,
no
.
8
,
pp.
1867
-
1876
,
2010.
[25]
Zha
ng,
et
al.,
"M
oti
on
Artifact
Reduc
t
ion
for
W
rist
-
W
orn
Ph
otopl
et
h
y
sm
ogra
p
h
Sensors
Based
on
Diffe
ren
t
W
ave
le
ngths,"
S
ensors
,
vol. 19,
no.
3
,
1
-
18
,
201
9.
[26]
Viola
,
and
Mich
ae
l
Jones
,
"Rapi
d
Objec
t
Detect
i
on
using
a
Boost
ed
Casca
de
of
Sim
ple
Feat
ure
s,"
IEE
E
Compute
r
Soci
e
ty
Con
fe
re
nce
on
Compute
r V
ision
and
Pa
t
te
rn R
ec
ogni
ti
on
,
vol
.
1
,
pp
.
511
-
518,
2001
.
BIOGR
AP
HI
ES OF
A
UTH
ORS
Murt
had
Al
-
Yoon
us
rec
ei
ved
a
M.
Eng
in
El
e
ct
roni
c
and
Com
pute
r
En
gine
er
ing
from
Univer
sit
y
T
ec
h
nic
a
l
Mel
aka
M
al
a
y
si
a,
h
e
wor
ked
in
Al
Kuno
oze
Univ
ersity
Coll
ege
2016
-
2019,
te
a
chi
ng
and
supervision
of
gr
adua
t
ed
Projects.
Curre
nt
l
y
le
ct
ur
ing
at
Pai
ta
xt
T
ec
hni
cal
Privat
e
Insti
tut
e
,
Erbi
l,
KRG
of
Ira
q.
His
rese
ar
ch
int
er
ests
inc
l
ude
Im
age
,
sign
al
proc
essing,
dee
p
m
ac
hin
e
learni
ng
and
Mobi
le
Roboti
c
S
y
ste
m
.
Published
four
pape
rs
in
the
fie
lds
of
Im
age
Proce
ss
ing
and
p
at
t
ern
r
ec
ogni
ti
o
n.
Mustafa
H.
Mo
hammed
Alhabi
b
completed
B
Sc.
degr
ee
from
the
D
epa
r
tment
of
Com
pute
r
Engi
ne
eri
ng,
Co
ll
eg
e
of
Engi
ne
eri
ng,
U
nive
rsit
y
of
Mos
ul
in
2008,
obt
ai
ned
MS
c
degr
ee
in
Te
chn
ic
a
l
Com
pute
r
Engi
ne
ering
from
the
T
ec
hni
ca
l
Engi
n
e
eri
ng
Col
le
g
e
Mos
ul
in
2013.
Speci
a
li
z
ed
in
Im
age
Proce
ss
ing
and
Artifici
al
Int
el
l
ige
nc
e
fie
lds.
Curr
ent
l
y
l
ec
t
uring
a
t
the
dep
art
m
ent
of
Com
m
unic
at
ions
and
Com
pute
r
Engi
n
ee
ring
a
t
ci
han
Univ
ersity
,
Erbil,
KRG
of
Ira
q
.
Publishe
d
5
pap
ers
i
n
th
e
fields of
Med
ica
l
Im
age Proce
ss
i
ng
and
p
atter
n
re
cogni
ti
on
.
Mustafa
Z
.
Na
y
ef
Al
-
Dab
agh
rec
e
ive
d
th
e
B.
Sc.
and
M.Sc
d
egr
ee
in
Com
pute
r
Eng
ine
er
ing
Te
chno
log
y
fro
m
Northe
rn
Techni
c
al
Univ
ersi
t
y
,
Ir
aq
in
200
8
and
2014.
He
is
working
as
a
lectur
er
on
K
nowledge
univ
e
rsit
y
.
His
cur
re
nt
rese
ar
ch
inter
ests
inc
lude
Im
age
Proce
ss
ing,
Patt
ern
Rec
ogn
ition, and
Ma
chi
n
e
L
ea
rn
ing, Bi
o
m
edi
ca
l
Engi
n
eering,
and
Robo
tics S
y
st
ems
.
M.
F
.
L
.
Ab
du
l
lah
He
recei
ved
Phd
degr
ee
El
e
ct
ri
ca
l
Engi
n
ee
ri
ng
from
(Unive
rsit
y
W
arwi
ck
)
,
Master
E
le
c
trica
l
Engi
n
ee
r
ing
(
Univer
sit
y
T
ec
h
nolog
y
Malay
si
a),
BS
c
Elec
t
rical
Eng
ineeri
n
g
(Unive
rsit
y
T
echnolog
y
Ma
lays
ia
),
Dip
Edu
(UTM),
Certi
fic
a
te
E
lectr
oni
c
Engi
n
ee
rin
g
(Pol
y
t
ec
hni
c
U
ngku
Om
ar)
.
Curre
ntly
lectu
ring
a
t
th
e
D
epa
rtmen
t
of
Com
m
unic
at
ion
Engi
ne
eri
ng,
Univer
siti
Tun
H
uss
ei
n
Onn
Malay
s
ia.
Publishe
d
153
pape
rs
i
n
the
fi
el
d
of
El
e
ct
ri
ca
l
and
C
om
m
unic
at
ion, signa
l
proc
essing
a
nd
fib
er
op
ti
cs
.
Evaluation Warning : The document was created with Spire.PDF for Python.