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.
8
, No
.
6
,
Decem
ber
201
8
, p
p.
4212
~
4220
IS
S
N:
20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v
8
i
6
.
pp
4212
-
42
20
4212
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
Pac
k
ets
W
avelet
s a
nd
St
oc
kw
ell
T
ra
ns
for
m
A
nal
ysis of
F
em
oral
Dopple
r Ultras
ound
S
ignals
M.
L
atf
aoui
1
, F
. Bereksi
Re
gu
ig
2
1
Depa
rt
m
ent
of Electrical E
ng
in
ee
ring
,
Fa
cul
t
y
o
f
Technol
og
y
,
T
ahr
i
Moham
ed
B
ec
har
Univer
si
t
y,
Alger
ia
2
Biom
edi
cal
E
ng
ine
er
ing
L
abor
a
t
or
y
,
Biom
edica
l Engi
ne
eri
ng
D
ep
art
m
ent
,
Facu
lty
of
Technol
og
y
,
Univer
sit
y
of
T
l
emce
n,
Alger
i
a
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Dec
15
, 201
7
Re
vised
Jan
14
, 201
8
Accepte
d
J
ul
26
, 2
01
8
Ultra
sonic
Doppler
signal
s
are
widely
used
in
the
det
ec
t
ion
of
ca
r
diova
scul
ar
pat
hologies
or
the
eva
lu
at
ion
of
t
he
degr
ee
of
ste
nosis
in
the
fem
ora
l
arteri
es.
The
pre
sen
ce
of
stenosis
ca
n
be
i
ndic
a
te
d
b
y
dist
urbing
the
blood
flow
in
the
femoral
arteri
es,
ca
using
spe
ct
r
al
broa
d
eni
ng
of
the
Doppler
signal
.
T
o
ana
l
y
z
e
th
ese
t
ypes
of
signal
s
a
nd
det
ermine
st
enosis
inde
x,
a
num
ber
of
ti
m
e
-
fre
quen
c
y
m
et
hods
have
be
en
develope
d
,
such
as
the
short
-
tim
e
Fourier
tra
nsform
,
the
c
onti
nuous
wave
l
et
s
tra
nsform
,
th
e
wave
let
pac
k
et
tra
nsform
,
and
th
e
S
-
tra
nsfo
rm
Ke
yw
or
d:
C
on
ti
n
uous wa
velet
tran
s
f
or
m
Dop
pler ult
ras
ound
SBI
Stenosi
s
S
-
tra
ns
f
orm
The wa
velet
p
a
cket tra
ns
f
or
m
Copyright
©
201
8
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
:
M. Lat
fa
ou
i,
Dep
a
rtm
ent o
f El
ect
rical
Eng
i
neer
i
ng,
Faculty
of Tec
hnology,
Tahr
i M
oham
e
d
Be
c
har
U
nive
rsity
.
Em
a
il
:
m
.
la
tfaou
i
@g
m
ai
l.com
1.
INTROD
U
CTION
Four
ie
r
a
naly
sis
is
a
basic
to
ol
in
sig
nal
pr
ocessin
g.
It
is
ind
is
pen
sa
ble
in
m
any
areas
of
researc
h;
unf
or
tu
natel
y
it
has
lim
i
ta
ti
on
s
wh
e
n
im
ple
m
ented
beyo
nd
the
stric
t
fr
a
m
ewo
r
k
of
it
s
def
i
niti
on
:
the
area
of
sta
ti
on
ary
finite
ene
rg
y
sig
nals.
I
n
Fou
rier
a
naly
sis,
al
l
the
tem
po
ral
as
pec
ts
bec
om
e
il
le
g
ible
in
t
he
s
pec
trum
.
The
stu
dy
of
non
-
sta
ti
on
a
ry
sign
al
s
the
refor
e
requires
e
it
her
an
e
xten
sion
of
the
F
ourier
T
ran
s
f
orm
(o
r
sta
ti
on
ary m
eth
ods
),
i
ntr
oduc
ing
a
tem
po
ral
aspect,
or the
de
velo
pm
ent o
f speci
fic m
et
hods
.
A
fir
st
so
luti
on,
im
ple
m
ente
d
intuit
ively
in
the
m
id
-
centu
ry,
co
rr
es
pond
s
to
F
ourier
a
na
ly
sis
sli
din
g
wind
ow
or
sho
rt
tim
e
Fo
ur
ie
r
trans
form
(
STFT
),
w
hich
wa
s
intr
oduce
d
in
1945
by
D.
G
abor
with
th
e
idea
of
a
tim
e
-
fr
equ
e
nc
y
plan
w
her
e
tim
e
beco
m
es
an
ad
diti
on
al
pa
ram
et
er
of
fr
e
qu
e
ncy
[
1].
Th
is
m
et
ho
d
sho
w
s
that
a
j
oi
nt
exact
l
ocati
on
in
ti
m
e
and
freq
ue
nc
y
is
i
m
po
ssible,
an
d
i
ntrodu
ce
s
the
idea
of
a
discrete
basis,
m
ini
m
u
m
, r
esu
lt
ing
in
a
f
e
w
c
oeffici
ents
of
t
he
si
gn
al
e
nerg
y dist
rib
ution i
n
ti
m
e
-
fr
eq
uency
p
la
n.
Othe
r
m
et
ho
ds
are
use
d
i
n
thi
s
w
ork
,
nam
ely
the
co
ntinuo
us
wav
el
et
tra
nsfo
rm
and
wa
ve
le
t
pack
et
s.
t
hese
tw
o
va
riances
of
the
w
avelet
trans
for
m
hav
e
exist
ed
in
a
la
te
nt
st
at
e
in
bo
t
h
m
a
them
a
ti
cs
and
sign
al
processi
ng,
but
the
real
e
xpan
sion
be
gan
in
t
he
early
1980s
.
The
la
st
m
e
tho
d
us
e
d
i
n
this
w
ork
(
base
d
on
the
wav
el
et
tra
ns
f
or
m
)
is
the
S
-
t
rasfor
m
pr
op
ose
d
by
Sto
kw
el
le
t
al
.;
it
is
si
m
il
ar
to
ST
FT
with
an
e
xce
pt
ion
that
the am
plit
ud
e
and w
i
dth
of th
e analy
sis wi
ndow a
re a
fun
c
ti
on
of freq
ue
nc
y, as in
the
w
avelet
transf
orm
[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
Packets W
avel
et
s and S
t
ockwel
l
Transfor
m An
alysis of F
e
mo
r
al
Dop
pler
U
lt
ra
s
ound
Si
gnals (M.
L
atfaoui)
4213
2.
METHO
DS
O
F ANAL
YS
I
S
2.1.
Sho
r
t
-
Ti
me
F
ou
ri
er Tr
an
s
f
orm
This
m
et
ho
d
is
base
d
on
t
he
deco
m
po
sit
io
n
of
t
he
si
gn
al
i
nto
sm
al
l
segm
ents
in
w
hic
h
the
Fou
rier
trans
form
is
a
pp
li
ed;
the
re
by
gen
erati
ng
a
local
iz
ed
sp
ec
trum
analy
t
ic
ally
ST
FT
is
give
n
by
the
f
ollo
wing
relat
ion
s
hip
:
2
/
2
/
2
)
(
)
(
)
(
T
T
ft
j
dt
e
t
w
t
X
f
X
(1)
Wh
e
re
w(
t
-
τ)
i
s
a
sel
ect
ed
window
f
un
ct
io
n.
The
act
ion
of
this
window
is
to
locat
e
in
tim
e,
the
res
ulti
ng
local
sp
ect
r
um
.
This
local
iz
at
ion
w
indow
is
the
n
sh
ifte
d
i
n
tim
e
to
pro
du
ce
t
he
local
sp
ect
r
um
fo
r
the
dura
ti
on
of
the ex
ist
e
nce
of
x(
t)
. Th
e res
ul
ti
ng
s
pec
tral
powe
r
is cal
le
d
sp
ect
r
ogram
[
3]
-
[5
]
.
2.2.
Continu
ous
Wavel
et Tr
ansfo
r
m
The
c
onti
nuou
s w
a
velet
trans
form
(
CWT
)
is
def
i
ned b
y:
dt
t
Ψ
t
x
b
a
C
W
T
b
a
)
(
)
(
)
,
(
*
,
(2)
Wh
e
re
x(
t
)
rep
rese
nts
the
analy
zed
sign
al
,
a
an
d
b
represe
nt
resp
ect
ively
the
scal
ing
factor
(d
il
at
at
ion
/c
om
pr
essi
on
coe
ffi
ci
ent)
an
d
th
e
tim
e
(sh
ifti
ng
coe
ff
ic
ie
nt
),
a
nd
the
s
upersc
ript
ast
e
r
isk
(
*)
denotes t
he
c
om
plex
conj
ug
a
ti
on
.
Ψ
a,b
(
t)
is
ob
ta
ine
d by sc
al
ing
the
w
a
vel
et
at tim
e
b
an
d
scal
e
a
:
a
b
t
Ψ
a
t
Ψ
b
a
1
)
(
,
(3)
Wh
e
re
ψ(t)
represents
the
wa
velet
tim
e fu
nc
ti
on
[6
]
.
2.3.
The W
av
el
et
Packe
t
Tr
an
s
f
orm
Wav
el
et
pac
ke
ts
us
e
d
t
o
dec
om
po
se
the
si
gnal
to
a
la
rg
e
num
ber
of
ba
ses
and
sel
ect
ed
w
it
h
a
ce
rtai
n
crit
erion
;
the
one
that
best
re
pr
ese
nts
t
he
si
gn
al
.
D
uri
ng
de
com
po
sit
ion
,
the
low
-
pass
a
nd
hi
gh
-
pass
filt
ered
ver
si
ons
of
th
e
sig
nal
are
de
com
po
sed
.
T
he
ap
pro
xim
a
tio
n
of
t
he
deta
il
s
and
the
det
ai
ls
of
the
det
ai
ls
are
therefo
re
a
dded
to
the
ap
pro
xim
a
ti
on
of
the
sig
nal.
The
coe
ff
ic
i
ents
f
r
om
this
dec
om
po
sit
ion
ar
e
char
act
e
rized by
thr
ee par
am
et
ers: the level of
deco
m
po
sit
ion, f
reque
ncy ind
e
x,
a
nd tim
e
ind
e
x.
Eac
h
w
avelet
pack
et
is
ca
rr
y
ing
tri
ple
in
for
m
at
ion
{
f,
s,
p
},
f
reque
ncy,
s
cal
e,
an
d
po
sit
i
on
w
her
e
the
wav
el
et
has
only
two
par
am
et
ers: scal
e and posit
io
n [7
]
-
[8
]
.
In
wa
velet
pack
e
t
a
naly
s
is,
the
si
gn
a
l
is
deco
m
po
se
int
o
a
pproxim
ation
s
a
nd
detai
ls
.
The
a
pproxim
a
ti
on
is the
n
it
sel
f
dec
om
po
sed i
nto
app
roxim
a
ti
on
and
detai
l i
n
seco
nd level
, and
the
proce
ss is
rep
eat
e
d.
For
a
dec
om
po
sit
ion
of
n
le
vel,
there
a
re
(n+
1)
po
ssibil
it
ie
s
to
d
ecom
po
se
or
to
c
ode
the
sig
nal.
The
Fig
ur
e
1
s
hows
t
he dec
om
po
sit
ion
of a
dig
it
al
sig
nal
in
wav
el
et
tra
nsfo
rm
at three levels.
Figure
1
.
W
a
ve
le
t t
ran
sf
or
m
d
ecom
po
sit
io
n schem
e
S
D
1
A
1
D
3
D
2
A
2
A
3
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.
8
, N
o.
6
,
Dece
m
ber
201
8
:
4212
-
4220
4214
Wh
e
re
S=
A
1
+D
1
,
S=
A
2
+
D
2
+D
1
,
S=
A
3
+D
3
+D
2
+D
1
.
I
n
wav
el
et
pack
et
s
analy
sis,
dec
om
po
sit
ion
into
ap
pro
xim
at
ion
an
d
deta
il
is
m
ade
on
ly
on
ap
pro
xim
at
ion
s
bu
t
al
so
on
detai
ls.
In
oth
er
wor
ds,
whe
n
analy
zi
ng
in
wav
el
et
pac
ke
ts,
it
is
no
lo
ng
e
r
only
the
filt
ered
low
-
pass
ve
rsions
of
the
sig
n
al
that
are
deco
m
po
se
d,
but
al
so
the
filt
ered
high
-
pass
ver
sio
ns.
In
a
no
t
her
way,
th
e
hig
h
fr
e
qu
e
nc
ie
s
are
al
so
cut
into
su
b
-
bands
a
nd
the
dec
om
po
sit
ion
tree
de
viate
sy
m
m
et
rical
l
y.
The
wa
velet
pack
et
s
decom
po
sit
ion
le
ad
s
to
a
deco
m
po
sit
io
n i
nto
fr
e
qu
e
ncy
s
ub
-
bands o
f
t
he
si
gn
al
[9
]
-
[
11
]
.
2.4.
The S
-
Tr
an
s
f
orm
The
S
-
tra
nsfo
r
m
pr
ovides
a
ti
m
e
-
fr
e
qu
e
ncy
represe
ntati
on
of
a
si
gn
al
.
It
on
ly
com
bin
es
a
dep
e
ndent
fr
e
qu
e
ncy
re
sol
ution
with
si
m
ul
ta
neo
us
lo
cat
ion
of
the
real
an
d
im
aginar
y
pa
rt
of
t
he
s
pectr
um
.
It
was
pub
li
sh
e
d
for
t
he
fi
rst
tim
e
in
1996
by
Sto
kwel
l
.
Th
e
ba
sic
idea
of
t
his
ti
m
e
-
fr
e
qu
e
ncy
distrib
ution
is
si
m
il
ar
to
the
F
ourier
trans
form
sli
di
ng
wind
ow,
e
xcep
t
that
t
he
a
m
plit
ud
e
an
d
width
of
the
analy
sis
window
are
var
ia
ble
depen
ding
on
the
f
r
equ
e
ncy
as
is
the
case
in
wav
el
et
analy
s
is
[1
1]
-
[
12
]
.
T
he
S
-
t
ra
ns
f
or
m
of
a
functi
on
x(
t
)
can
be
de
fine
d
a
s
a
tran
sf
or
m
i
nto
wa
velet
s
w
it
h
a
quit
e
sp
ec
ific
m
oth
er
wa
velet
m
ult
ipli
e
d
by
a
ph
a
se f
act
or
:
)
,
(
)
,
(
2
d
W
f
S
f
i
e
(4)
Wh
e
re
W(
τ
,d
)
is t
he
c
onti
nuou
s w
a
velet
trans
form
o
f
the
sig
nal
x(
t
)
def
ine
d by:
dt
d
t
w
t
x
d
W
)
,
(
)
(
)
,
(
(5)
Wh
e
re t
he
m
oth
er
w
a
velet
is
def
i
ned b
y:
ft
i
e
f
t
f
e
f
t
w
2
2
2
2
2
,
(6)
Let
u
s
note
tha
t t
he
dilat
at
ion
factor
d
is t
he
re
ver
se
of t
he
f
r
equ
e
ncy
f
[13]
-
[14].
3.
RESU
LT
S
A
ND AN
ALYSIS
The
D
oppler
si
gn
al
s
stu
died
in
t
his
w
ork
ar
e
f
ro
m
the
S
t
Ma
rie
hos
pital
in
Lei
cest
er
(
En
glan
d)
;
t
he
sign
al
file
s
are
in
.w
a
v
form
with
a
sam
pling
fr
e
qu
e
ncy
of
44
KHz
a
nd
a
durati
on
of
4.34
s
,
c
orres
pondin
g
t
o
191
390
sam
ples.
The
F
ig
ur
e
2
sh
ows
a
te
m
po
ral
rep
rese
ntati
on
an
d
spe
ct
ral
analy
sis
of
a
sign
al
da
ta
base
.
Fr
om
the
resu
l
ts
of
Keet
on
a
nd
Sa
dik
giv
e
n
in
[15]
-
[
16]
,
a
Dopp
le
r
si
gnal
can
be
regarde
d
as
a
Ga
us
sia
n
sign
al
in
a
se
gm
ent
of
10
m
s
to
12
m
s
.
Althou
gh
it
has
not
al
ways
gu
a
r
anteed
t
hat
the
Dop
pler
si
gn
a
ls
are
Gau
s
sia
n
at
10
m
s
, th
is rem
a
ins tru
e
f
or se
gme
nts
belo
w
10
m
s
.
The
pri
nci
pal
obj
ect
ive
of
th
is
researc
h
is
to
com
par
e
ST
FT,
C
W
T
,
P
W
T
and
S
-
tra
nsf
or
m
m
et
ho
ds
in
the
case
of
the
res
olu
ti
on
of
ti
m
e
-
fr
eq
ue
ncy
of
ultras
on
ic
D
oppler
sign
al
s.
T
he
pur
pose
of
the
tim
e
-
fr
e
qu
e
ncy
a
naly
sis
is
to
pro
vi
de
a
m
or
e
in
form
ative
desc
ription
of
the
sig
nal
re
veali
ng
t
he
te
m
po
ral
va
riat
io
n
of
it
s
fr
e
quenc
y
con
te
n
t.
A
s
olu
ti
on,
w
hich
is
consi
der
e
d
a
s
the
m
or
e
intu
it
ive,
co
ns
ist
s
in
ass
ociat
ing
a
non
-
sta
ti
on
ary
si
gnal
a
seq
ue
nce
of
Fou
rier
tra
nsfo
rm
s
sh
ort
-
te
rm
to
adap
t
t
he
su
ccessi
ve
observ
at
io
n
wind
ow
s
to
structu
ral
var
ia
ti
on
s
of
the
sign
al
in
suc
h
a
way
that
the
sta
ti
on
arit
y
assum
ption
s
are
local
ly
sat
isfie
d
[1
6]
.
The
disa
dvanta
ge
of
the
F
ouri
er
trans
f
or
m
is
the
sta
ti
on
ary
of
si
gn
al
s,
a
nd
therefo
re
does
no
t
al
lo
w
obta
inin
g
tim
e
info
rm
at
i
on.
S
TFT
im
p
li
ci
tly
is
con
si
der
e
d
f
or
a
non
-
sta
ti
onary
sign
al
as
a
serie
s
of
q
uasi
-
sta
ti
on
a
r
y
sit
uations
a
cr
oss t
he
a
naly
sis
window.
The
te
m
po
ral
res
olu
ti
on
of
su
c
h
a
n
a
naly
sis
is
determ
ined
by
the
width
of
the
window,
the
fr
e
qu
e
ncy res
ol
ution
is
deter
m
ined
by the
width
of it
s Four
ie
r
t
ran
s
f
or
m
. F
or a
highly
n
on
-
sta
ti
on
a
ry sign
al
a
s
an
ultraso
nic
Dop
pler
si
gn
al
,
go
od
te
m
po
r
al
reso
l
ution
is
re
qu
ire
d,
w
hi
ch
re
qu
i
res
w
orkin
g
with
a
short
window.
The
m
ajo
r
disad
va
ntage
of
t
his
tr
ansfo
rm
is
the
lim
it
at
ion
of
th
e
fr
e
qu
e
ncy
re
so
luti
on.
T
he
pro
blem
of
the
ST
FT
is
that
it
us
es
a
fi
xed
siz
e
o
f
window
co
ve
rin
g
the
tim
e
-
fr
eq
ue
ncy
do
m
ai
n.
A
no
t
her
d
isa
dva
ntage
of
t
his
tran
sf
orm
is
the
fixed
te
m
po
ral
an
d
f
reque
ncy
re
so
luti
on
[
17]
.
Pr
oc
essin
g
in
gen
e
ral,
or
by
us
in
g
wav
el
et
S
-
tra
nsfo
rm
,
of
fe
r
th
e
possibil
it
y
to
have
a
wi
ndow
that
a
dap
ts
accor
ding
to
th
e
irre
gu
la
riti
es
of
t
he
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
Packets W
avel
et
s and S
t
ockwel
l
Transfor
m An
alysis of F
e
mo
r
al
Dop
pler
U
lt
ra
s
ound
Si
gnals (M.
L
atfaoui)
4215
sign
al
.
W
a
vele
ts
are
a
fam
il
y
of
functi
ons
l
ocali
zed
i
n
ti
m
e
and
f
reque
ncy
a
nd
f
or
m
an
or
t
honorm
al
basis.
They
are
gen
e
r
at
ed
one
f
ro
m
the
oth
e
r
by
tr
anslat
ion
a
nd
dilat
at
ion
.
Eac
h
wa
velet
is
use
d
to
de
com
pose
the
sign
al
an
d
is
us
ed
as
each
ex
pone
ntial
fu
nct
ion
in
the
Fou
rier
trans
form
.
The
diff
e
re
nc
e
is
that
the
wav
el
et
functi
ons a
re
well
locali
zed i
n
ti
m
e u
nlike
ex
pone
ntial
s [18
]
.
(a)
(b)
Figure
2
.
(
a
)
T
e
m
po
ral
repres
entat
ion
.
(
b) S
pectr
um
o
f
D
opple
r Ult
rasou
nd Sig
nal
The
wa
velet
pa
ckets
trans
for
m
is
a
gen
eral
iz
at
ion
of
the
D
WT
;
it
allows
deco
m
po
sin
g
the
detai
l
s
appr
ox
im
at
ion
.
This
trans
f
orm
al
lows
decom
po
sing
the Doppler
si
gn
al
i
nto
s
ub
-
f
re
qu
e
ncy
ba
nd
s b
y
m
eans
of
a filt
er b
a
nk. T
he
c
oeffici
ents
of
t
his d
e
com
po
sit
ion
give
a t
i
m
e
-
fr
eq
ue
ncy
represe
ntati
on
that ca
n
m
on
it
or
t
he
velocit
y o
f
bl
ood i
n
t
he
arte
ries. T
he wavele
t
-
base
d
tra
nsfo
rm
l
ike
C
W
T
, P
W
T
or
S
-
tra
nsfo
rm
are
desi
gn
e
d
t
o
giv
e
good
tim
e
reso
l
ution
with
a
poor
f
reque
ncy
reso
l
ution
at
hi
gh
fr
e
qu
e
ncies
an
d
a
good
f
re
quenc
y
reso
l
ution
with
a
poor
te
m
po
r
al
reso
luti
on
at
low
f
re
qu
e
nci
es.
The
m
ajo
r
dr
a
w
back
of
t
he
se
trans
form
s
is
the
cho
ic
e
of
the
m
oth
er
wa
velet
.
Usi
ng
the
S
TFT
or
wa
vele
t
-
base
d
tra
nsfo
rm
req
uires
a
com
pr
om
ise
between
tim
e
and
fr
e
quency
reso
l
ution
s
.
F
o
r
t
he
STFT
,
na
rrower
analy
sis
wi
ndow
will
pro
vid
e
bette
r
te
m
po
ral
reso
l
ution,
but
the
con
ce
ntrat
ion
ar
ound
the
or
igi
n
of
the
Four
ie
r
tra
nsfo
rm
wil
l
necessarily
be
le
ss,
wh
ic
h
i
m
plies
a
po
orer
fr
e
quency
re
so
luti
on.
F
or
f
ur
t
her
tra
ns
f
orm
s,
the
com
pr
om
ise
is
si
m
il
a
r,
an
d
de
pe
nds
on
the
scan
f
re
qu
e
ncy
:
increasin
g
th
e
analy
sis
fr
e
quency
im
plies
i
m
pr
ovin
g
the
tim
e
reso
luti
on,
but
dec
reasin
g
th
e
fr
e
qu
e
ncy
resol
ution
[
15
]
.
The
wa
velet
-
ba
sed
tra
nsfo
rm
s
ha
ve
bee
n
de
sign
e
d
for
non
-
sta
ti
on
a
ry
sig
nals
since
they
inco
r
po
rate
the
co
nce
pt
of scal
e
to
tra
ns
f
orm
ation
, w
hic
h
giv
es
bette
r
ti
m
e
-
fr
e
qu
e
ncy reso
l
ution
:
a
c
om
pr
essed
wa
velet
to
analy
ze
the
hi
gh
f
reque
ncy
detai
l
and
a
dilat
ed
wa
velet
to
detect
unde
rly
ing
tre
nds
of
l
ow
f
reque
ncy.
In
so
no
gr
am
s
ob
t
ai
ned
by
ST
F
T,
C
W
T
,
P
W
T
an
d
S
-
tra
ns
f
or
m
are
give
n
in
F
ig
ure
3
.
The
horizo
ntal
axis
(
t
)
sh
ows
the
tim
e
an
d
t
he
fr
e
quency
(
f
)
is
s
how
n
on
t
he
ve
rtic
al
axis.T
he
gray
le
vel
i
ntensity
re
pr
ese
nts
th
e
powe
r
le
vel c
orres
pondin
g
t
o a fr
e
quency
f
or each
point i
n
t
he
ti
m
e axis [
19]
.
It
is
cl
ear
that
the
CWT,
PW
T
an
d
S
-
tra
nsf
or
m
s
cou
l
d
help
to
im
pr
ove
t
he
qu
al
it
y
of
s
onogram
s
of
Dop
pler.
Ultra
so
nic
Dop
pler
sign
al
s
sam
pled
co
ntain
a
wealt
h
of
in
form
ation
on
bl
ood
fl
ow.
T
he
m
os
t
com
pr
ehe
ns
ive
way
to
sho
w
t
his
in
form
at
io
n
is
to
pe
rfo
rm
a
tim
e
-
fr
e
qu
e
ncy
analy
sis
a
nd
prese
nt
the
resu
lt
s
as
so
no
gr
am
s.
Fo
r
t
he
S
-
tr
ansfo
rm
,
a
l
i
near
ti
m
e
-
fr
eq
uen
cy
re
prese
ntati
on
is
pre
sented
.
This
m
et
ho
d
su
r
passes
the
pro
blem
of
the
sli
ding
wi
ndow
F
ourier
tran
sform
of
fi
xed
le
ng
t
h,
a
nd
a
ddress
es
the
not
ion
of
p
ha
se
in
the
w
avelet
trans
for
m
fo
r
non
-
sta
ti
on
a
ry
signa
ls
analy
sis.
This
trans
form
pr
ov
ides
a
ve
ry
su
i
ta
ble
sp
ace
for
featu
re
e
xtracti
on
a
nd
locat
io
n
in
tim
e
and
f
requ
ency
disc
rim
in
at
ing
in
f
or
m
at
i
on
in
t
he
ultra
so
nic
Dop
pler
si
gn
al
[14
]
.
In
F
ig
ure
3,
one
ca
n
obser
ve
an
i
m
pr
ov
e
m
ent
in
qu
al
it
y
of
so
no
gr
am
s
ob
ta
ine
d
by
the
wav
el
et
trans
form
ov
er
those
obta
in
ed
by
ST
FT
.
The
s
onogram
s
ob
ta
i
ned
by
ST
FT
giv
e
a
low
-
qual
it
y
sp
ect
ral
interp
retat
ion
i
n
te
rm
s
of
loca
ti
on
of
m
ini
m
u
m
and
m
axi
m
u
m
fr
eq
uen
ci
es
.
The
ad
va
ntag
e
of
the
S
-
tra
nsfo
rm
is o
pti
m
iz
ing
the tim
e
-
fr
eq
ue
ncy res
olu
ti
on.
It
is
cl
ear
fr
om
F
igure
3
that
t
her
e
is
a
certai
n
qual
it
at
ive
im
pr
ov
em
ent
in
so
no
gr
am
s
ob
ta
ined
by
the
S
-
tra
nsfo
rm
com
par
ed
to
tho
s
e
ob
ta
ine
d
by
ST
FT
.
S
onogra
m
s
ob
ta
ined
by
ST
FT
gi
ve
f
al
se
fr
eq
ue
ncies,
an
d
sp
ect
ral
analy
s
is
by
the
ST
F
T
pro
du
ces
uncl
ear
so
no
gr
a
m
s
becau
se
of
the
distor
ti
oni
n
sp
ect
ral
est
im
at
ion
cause
d
by
sli
di
ng
wind
ow.
T
he
ad
va
ntage
of
the
S
-
tra
ns
f
or
m
co
m
par
ed
to
ST
FT
is
the
optim
iz
at
ion
of
tim
e
-
fr
e
qu
e
ncy
res
ol
ution
a
nd
the
dyn
am
ic
local
i
zat
ion
of
the
s
pectr
um
in
the
tim
e
-
fr
eq
uen
c
y
plan.
T
he
se
cond
0
0
.
5
1
1
.
5
2
2
.
5
3
3
.
5
4
-
2
.
5
-2
-
1
.
5
-1
-
0
.
5
0
0
.
5
1
1
.
5
2
2
.
5
x
1
0
4
t
i
m
e
(
s
)
x
(
t
)
0
1000
2000
3000
4000
5000
6000
0
0
.
5
1
1
.
5
2
2
.
5
3
3
.
5
4
x
1
0
6
f
r
e
q
u
e
n
c
y
(
H
z
)
a
b
s
(
X
(
f
)
)
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.
8
, N
o.
6
,
Dece
m
ber
201
8
:
4212
-
4220
4216
adv
a
ntage
of
the
S
-
tra
ns
f
orm
is
a
bette
r
loc
at
ion
of
syst
oli
c
peaks
us
e
d
t
o
dete
rm
ine
the
sp
ect
ral
br
oa
den
i
ng
ind
e
x (
SB
I
).
(a)
(b)
(c)
(d)
Figure
3. Fem
or
al
arterial
Dop
pler
s
onog
ram
s
:
(a)
us
i
ng
STF
T,
(
b)
us
in
g
C
WT,
(c) usi
ng
PWT
,
(d) usin
g
S
-
t
ra
ns
f
or
m
Partic
ularly
be
cause
of
the
i
nt
rinsic
lim
it
ati
on
s
of
t
he
s
pe
ct
rogr
am
and
par
ti
cula
rly
ti
m
e
-
fr
e
qu
e
ncy
reso
l
ution
of
pro
blem
s,
oth
er
ty
pes
of
f
requ
ency
-
ti
m
e
rep
r
esentat
ion
a
re
pr
e
ferred
.
The
tim
e
-
fr
equ
e
nc
y
and
tim
e
-
scal
e
analy
sis’s
ha
ve
be
en
dev
el
op
e
d
to
m
ee
t
a
need
f
or
dem
onstrat
ing
phen
om
ena
that
are
very
local
iz
ed
in
ti
m
e
and
f
reque
ncy.
U
nlike
th
e
sp
ect
r
ogram
,
the
res
olu
ti
on
of
t
he
tim
e
-
fr
e
qu
e
ncy
re
pr
e
se
ntati
on
ob
ta
ine
d
by
th
e
S
-
tra
ns
f
or
m
or
wa
velet
-
tra
nsfo
rm
is
dep
en
den
t
on
fr
e
que
ncy
and
ti
m
e.
Both
well
local
iz
ed
in
tim
e and
f
re
qu
ency, the
S
-
tra
ns
f
or
m
o
r wa
ve
le
t t
ran
sf
or
m
h
as
prop
e
rtie
s
of
"z
oom
"
m
aking
it
an
i
deal too
l
for
detect
ing
phe
nom
ena
of
hi
gh
fr
e
quen
cy
an
d
s
hort
durati
on.
It
is
w
or
t
h
rem
ind
ing
t
hat
wa
velet
tra
nsfo
rm
s
po
s
sess
a
good
te
m
po
ral
res
ol
utio
n
at
hi
gh
f
re
qu
e
ncies
and
the
vice
-
ve
rsa
[
19]
.
T
he
wa
velet
trans
form
s
util
iz
e
a
set
of
analy
ti
cal
fu
nctions
buil
t
by
exp
a
ns
io
n
/
com
pr
essio
n
an
d
translat
io
n
of
a
functi
on
c
al
le
d
m
oth
er
wa
velet
.
I
n
ou
r
stu
dy
,
we
ch
os
e
t
he
Morlet
wav
el
et
that
has
a
s
i
m
i
l
ar
form
to
D
oppler
ultra
so
un
d
sign
al
s
[
20
]
.
4.
CA
L
CU
L
ATI
NG SBI
Sp
ee
ds
of
t
he
red
bloo
d
cel
ls
in
a
ve
ssel
ar
e
determ
ined
f
ro
m
the
Ultras
ound
D
opple
r
sign
al
s.
Th
e
tem
po
ral
e
vo
l
ut
ion
of
these
e
cho
e
s
is
s
onog
ram
s.
A
var
ia
ti
on
of
the
se
s
pe
eds
tr
anslat
es
directl
y
a
va
riat
ion
fr
e
qu
e
ncy,
or
tem
po
ral
le
vel
s
onogram
.
In
fa
ct
,
su
c
h
sit
uati
on
s
exist
wh
e
r
e
struct
ur
es
are
fou
nd
in
t
he
a
rteries
(car
otid
or
fe
m
or
al
).
These
directl
y
aff
ect
the
bloo
d
fl
ow
that
bec
om
es
non
-
unif
or
m
i
n
their
neig
hb
orh
ood.
This
cause
s
enl
arg
em
ent
of
th
e
Dopp
le
r
sig
na
l
sp
ect
ru
m
near
the
syst
olic
peak
quantifi
e
d
by
w
hat
we
c
al
l
the
t
i
m
e
(
s
)
f
r
e
q
u
e
n
c
y
(
H
z
)
0
0
.
5
1
1
.
5
2
2
.
5
3
3
.
5
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
f
r
e
q
u
e
n
c
y
(
H
z
)
t
i
m
e
(
s
e
c
)
0
0
.
5
1
1
.
5
2
2
.
5
3
3
.
5
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
5500
f
r
e
q
u
e
n
c
y
(
H
z
)
t
i
m
e
(
s
)
0
0
.
5
1
1
.
5
2
2
.
5
3
3
.
5
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
T
i
m
e
(
s
)
f
r
e
q
u
e
n
c
y
(
H
z
)
0
0
.
2
0
.
4
0
.
6
0
.
8
1
1
.
2
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
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
Packets W
avel
et
s and S
t
ockwel
l
Transfor
m An
alysis of F
e
mo
r
al
Dop
pler
U
lt
ra
s
ound
Si
gnals (M.
L
atfaoui)
4217
sp
ect
ral
broa
de
ning
ind
e
x
(
SB
I
)
.
H
ow
e
ver,
this
ind
ex
is
highly
cor
r
el
at
ed
with
the
na
ture
of
the
en
velo
pe
fr
e
qu
e
ncy
(
f
max
an
d
f
mean
).
In
fact,
the
f
requ
ency
ge
ne
rated
fro
m
so
nogra
m
s
is
e
m
bed
ded
in
noise
em
anati
ng
from
d
iffer
e
nt
so
urces
r
e
flect
ing (t
he wal
l of
the artery,
s
kin, etc
.)
that ca
n see
in Fi
gu
r
e
4.
(a)
(b)
(c)
(d)
Figure
4. Ma
xi
m
u
m
an
d
m
ea
n fr
e
qu
e
ncy e
nvel
opes
of the
fem
or
al
arteria
l extract
ed
from
(
a) S
TFT,
(b)
C
WT,
(c) P
WT, (
d) S
-
t
ransform
so
nogra
m
s
A
filt
er
app
li
e
d
to
the
sp
ect
r
al
env
el
opes
is
req
ui
red
to
de
te
rm
ine
the
sy
stoli
c
peak
an
d
est
i
m
a
te
the
SBI
in
de
x.
T
he
F
igu
re
5
il
lustrate
the
syst
olic
peak
s
that
ar
e
cl
early
def
ined
an
d
thu
s
al
l
ow
i
ng
b
et
te
r
te
m
po
ral
locat
ion
of
syst
olic evo
l
utio
n.
The
se
ver
it
y o
f
the sten
os
is (
gi
ven
in pe
rcen
t
age %)
is e
xpr
essed by t
he
ra
ti
o
of
the d
ia
m
et
er r
e
du
ce
d by the
st
en
osi
s and t
he a
ct
ual d
ia
m
et
er
of the
artery.
(a)
(b)
Figure
5. Fil
te
red
e
nv
el
op
e
s fm
ax
an
d fm
ea
n of t
he
a
rtery
fem
or
al
snogra
m
m
s:
(a)
u
sin
g STFT
,
(b) usin
g
C
WT
0
0
.
5
1
1
.
5
2
2
.
5
3
3
.
5
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
t
i
m
e
(
s
)
f
r
e
q
u
e
n
c
y
(
H
z
)
0
0
.
5
1
1
.
5
2
2
.
5
3
3
.
5
4
0
1000
2000
3000
4000
5000
6000
t
i
m
e
(
s
)
f
r
e
q
u
e
n
c
y
(
H
z
)
0
0
.
5
1
1
.
5
2
2
.
5
3
3
.
5
4
-
5
5
0
0
-
5
0
0
0
-
4
5
0
0
-
4
0
0
0
-
3
5
0
0
-
3
0
0
0
-
2
5
0
0
-
2
0
0
0
-
1
5
0
0
-
1
0
0
0
-
5
0
0
f
r
e
q
u
e
n
c
y
(
H
z
)
t
i
m
e
(
s
)
0
0
.
2
0
.
4
0
.
6
0
.
8
1
1
.
2
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
f
r
e
q
u
e
n
c
y
(
H
z
)
t
i
m
e
(
s
)
0
0
.
5
1
1
.
5
2
2
.
5
3
3
.
5
4
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
t
i
m
e
(
s
)
f
r
e
q
u
e
n
c
y
(
H
z
)
0
0
.
5
1
1
.
5
2
2
.
5
3
3
.
5
4
1000
1500
2000
2500
3000
3500
4000
t
i
m
e
(
s
)
f
r
e
q
u
e
n
c
y
(
H
z
)
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.
8
, N
o.
6
,
Dece
m
ber
201
8
:
4212
-
4220
4218
(c)
(d)
Figure
5. Fil
te
red
e
nv
el
op
e
s fm
ax
an
d fm
ea
n o
f
the a
rtery
fem
or
al
snogra
m
m
s:
(c)
P
W
T
, (d) u
si
ng S
-
trans
form
In
Fig
ur
e
6,
th
e
upstream
flow
of
t
he
ste
nos
is
is
lam
inar
an
d
re
d
bloo
d
cel
ls
flo
w
wit
h
a
sp
ee
d
cal
le
d
aver
a
ge
s
pee
d
V
mean
.
At
the
st
enosi
s,
the
re
d
blood
cel
l
velo
ci
ty
increases
because
of
a
rterial
co
ns
tric
ti
on
,
a
nd
this
m
ai
ntains
a
con
sta
nt
flo
w.
I
n
this
case
,
the
sp
ee
d
is
m
axi
m
u
m
and
is
cal
le
d
V
max
.
Im
m
ediat
el
y
dow
n
stream
of
the
ste
no
sis
,
wh
e
n
the
diam
et
er
increases
,
s
udden
ly
a
ppears
vortexe
d
c
om
plete
ly
disrupting
the
flo
w,
this
al
lo
ws
the
re
d
bl
ood
cel
ls
to
ta
ke
m
ulti
ple
sp
eeds
an
d
in
al
l
directi
ons.
T
he
aver
a
ge
value
of
these
sp
ee
ds
giv
es
the
a
ver
a
ge
flo
w
velocit
y
V
mean
[15].
Si
nce
t
he
s
pee
d
V
max
and
V
mean
res
pe
ct
ively
repres
ent
the
fr
e
qu
e
ncies
f
ma
x
an
d
f
mean
D
oppler
s
pectr
um
, o
ne
can
exp
res
s the
de
gr
ee
of
s
te
no
sis
acco
rdi
ng
t
o
f
max
an
d
f
mean
:
m
a
x
m
a
x
f
f
f
SB
I
m
e
a
n
(7)
Accor
ding
to
t
his
eq
uatio
n,
t
he
SB
I
is
the
n
us
e
d
in
our
s
tud
y
to
cal
c
ulate
the
de
gr
ee
of
ste
nosis.
Since
the
SB
I
i
s
cal
culat
ed
by
the
rati
o
of
(
f
m
ax
-
f
mean
)
an
d
f
max
,
it
is
necess
ary
to
us
e
la
r
ge
r
val
ues
of
f
ma
x
and
f
mean
by
the
fact
that
us
ing
sm
al
l
values
may
intro
duce
a
i
m
po
rtant
error
s
in
the
SB
I
cal
culat
ion
.
F
or
thi
s
reason,
t
he
SB
I
is
cal
culat
ed
at
the
syst
olic
peak
,
wh
e
re
the
flow
rates
(
fr
e
qu
e
ncies
of
the
Dop
pler
s
pectru
m
)
are
m
axi
m
a
l.
The
degree
of
sever
it
y
of
the
ste
no
sis,
(
give
n
in
per
ce
nta
ge
%)
is
e
xpre
ssed
by
the
di
a
m
et
er
reduce
d by the
ste
no
sis
and t
he
r
eal
arte
ry d
i
a
m
et
er [
15]
.
%
100
%
100
)
(
m
a
x
V
m
e
a
n
V
m
a
x
V
A
B
A
s
t
e
n
o
s
i
s
of
D
e
g
r
e
e
(8)
(a)
(b)
Figure
6. (a
) D
ia
gr
am
il
lustratin
g
t
he varia
ble involve
d
i
n
t
he defi
niti
on
of
SBI
.
f
max
is t
he
m
axi
m
u
m
fr
e
qu
e
ncy at
pe
ak
syst
ole,
f
mean
is t
he
m
ean
fr
e
qu
e
ncy,
S is
syst
olic pea
k,
D
is e
nd
diasto
li
c h
ei
gh
t
.
(b)
the
ef
fect
of sten
os
is
on th
e flow
of b
l
ood
in the
arte
ries
0
0
.
5
1
1
.
5
2
2
.
5
3
3
.
5
4
-
4
5
0
0
-
4
0
0
0
-
3
5
0
0
-
3
0
0
0
-
2
5
0
0
-
2
0
0
0
-
1
5
0
0
-
1
0
0
0
f
r
e
q
u
e
n
c
y
(
H
z
)
t
i
m
e
(
s
)
0
0
.
2
0
.
4
0
.
6
0
.
8
1
1
.
2
1
.
4
1000
1500
2000
2500
3000
3500
4000
4500
5000
f
r
e
q
u
e
n
c
y
(
H
z
)
t
i
m
e
(
s
)
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
Packets W
avel
et
s and S
t
ockwel
l
Transfor
m An
alysis of F
e
mo
r
al
Dop
pler
U
lt
ra
s
ound
Si
gnals (M.
L
atfaoui)
4219
Ther
e
a
re
two
m
et
ho
ds
of
cal
culat
ion
:
The
f
irst
m
et
ho
d
is
to
cal
culat
e
the
aver
a
ge
pa
ra
m
et
ers
(
f
max
and
f
mean
)
of
ea
ch
syst
olic
peak
,
an
d
the
n
to
deduce
the
SB
I
.
The
seco
nd
m
et
ho
d
is
to
aver
a
ge
the
SB
I
fou
nd
from
the
par
a
m
et
ers
(
f
max
an
d
f
m
ean
)
of
each
syst
olic
peak
.
The
SBI
dif
fer
e
nt
values
c
al
culat
ed
by
va
rio
us
m
et
ho
ds
a
ppli
ed
on
ST
F
T
,
S
-
t
ran
s
f
or
m
,
CWT
an
d
PWT
m
od
el
in
g
tra
nsfo
rm
te
chn
iqu
es
are
il
lustrate
d
on
t
he
F
igure
7
.
Figure
7.
SB
I
m
agn
it
ud
e
of
di
ff
ere
nts
Fem
or
al
Do
pp
le
r
sig
nals
us
in
g:
ST
FT
,
CW
T,
P
W
T
an
d
S
-
tra
nsfo
rm
m
et
hods
These
res
ults
sh
ow
that
the
m
easur
em
ent
of
s
pectral
bro
aden
i
ng
qu
a
nti
fied
by
broa
de
ning
sp
ect
ra
l
ind
e
x
m
ay
be
an
in
dicat
ion
of
ste
nosis
se
ver
it
y
at
the
f
e
m
or
al
arterie
s
.
The
SB
I
was
cal
culat
ed
f
rom
the
ST
FT
,
S
-
t
ran
s
f
or
m
,
CWT
a
nd
PWT
son
ogra
m
s.
On
e
obser
ved
a
st
ron
g
c
orrelat
ion
bet
ween
the
val
ue
of
th
e
SBI
ob
ta
ine
d
by
ST
FT
an
d
t
ha
t
ob
ta
ine
d
by
S
-
tra
nsfo
rm
,
CWT
an
d
PWT
.
The
res
ults
of
this
stu
dy
prov
e
d
that
in
sp
it
e
of
th
e
qu
al
it
at
ive
im
pr
ov
em
ent
of
the
di
ff
e
ren
t
so
no
gr
am
s;
it
has
no
qu
a
nt
it
at
ive
adv
anta
ge
in
e
m
plo
yi
ng
the
S
-
trans
f
or
m
,
CWT
and
PW
T
com
par
ed
to
the
ST
FT
f
or
the
determ
inatio
n
of
SB
I
due
to
it
s
weak va
riance
and w
it
h t
he a
ddit
ion
al
nu
m
erical
r
eq
uirem
ents.
5.
CONCL
US
I
O
N
The
ti
m
e
-
fr
eq
uen
cy
a
naly
sis
m
et
ho
ds
us
e
d
in
this
w
ork
aim
to
sh
ow
the
s
onogram
s
of
ultras
on
ic
Dop
pler
sig
nal
s,
wa
velet
-
ba
s
ed
m
et
ho
ds
s
uc
h
as
CWT
,
P
WT
an
d
S
-
tra
nsfr
om
and
the
cl
assic
al
ST
FT
m
et
ho
d
hav
e
bee
n
c
ompare
d
in
te
rm
s
of
t
heir
fr
e
que
ncy
-
re
so
l
ving
pow
e
r
an
d
t
heir
ef
fects
in
det
erm
ining
the
s
pectral
broa
den
i
ng in
de
x
in
the
prese
nce
of the ste
nosis i
n t
he ult
ra
so
un
d Dop
pler
sig
nals of t
he fem
or
al
arterie
s.
REFERE
NCE
S
[1]
M.
Jianpi
ng
an
d
J.
Jin,
“
Anal
ysis
and
design
of
m
odifi
ed
wi
ndow
shape
s
for
S
-
tra
nsform
to
impro
ve
ti
m
e
-
fre
quency
lo
ca
l
i
za
t
ion
,
”
Me
chan
ic
al
Syste
ms
and
Signal P
roc
essing
,
v
ol
.
58
,
p
p
.
2
71
-
284,
2015
.
[2]
S.
Kara
,
“
Cla
ss
ifi
cation
of
m
it
ra
l
stenosis
from
Doppler
signal
s
using
short
ti
m
e
Fourier
tra
nsfor
m
and
art
ifi
c
ial
neur
al ne
twor
ls
,
”
E
xpe
rt
Syst
ems
wit
n
applicati
on
s
,
pp
.
468
-
475
,
2
007.
[3]
S.
Dhan
y
a
r
and
V.
S.
K
.
Rosh
ni,
“
Com
par
ison
of
var
ious
texture
class
ifi
c
ation
m
et
hods
us
ing
m
ult
ire
solut
ion
ana
l
y
sis
and li
ne
ar
reg
ression
m
odel
li
ng
,
”
Spring
erPl
us,
2016
.
[4]
F.
Dirge
nali,
e
t
al.
,
“
Esti
m
ation
of
wave
le
t
and
short
-
ti
m
e
Fourier
trans
form
so
nogra
m
s
of
nor
m
al
and
dia
be
tic
subjec
ts’
el
e
ct
ro
gastrogr
am
,
”
Co
mputers i
n
B
iol
o
gy
and
M
edicine
,
vol
.
36
,
p
p.
128
9
–
1302,
2006
.
[5]
J.
S.
As
hwin
and
N.
Manoha
ran
,
“
Audio
Denoising
Based
on
Sh
ort
Ti
m
e
Fourier
Tra
nsform
,
”
In
donesian
Journa
l
of
E
le
c
tric
al
En
gine
ering
and
C
omputer
Scienc
e,
vol
.
9
,
pp
.
89
-
9
2,
2018
.
[6]
V.
Thiy
ag
ara
j
an
and
N.
P.
Subram
ani
am,
“
Anal
y
sis
and
Esti
m
at
ion
of
Har
m
onic
s
Us
ing
W
ave
let
T
e
chni
que
,
”
TEL
KOMNIKA
Indone
sian J
ourn
al
of
Elec
tric
al
Engi
ne
ering,
vol
/i
ss
ue:
13(2)
,
pp
.
305
-
313,
2015.
[7]
K.
Sat
y
anara
y
an
a,
e
t
al
,
“
Ide
nti
f
i
ca
t
ion
of
Fault
s
in
HV
DC
Sy
st
e
m
using
W
ave
let
Anal
y
sis
,
”
Inter
nati
onal
Journ
al
of
E
le
c
tric
al
and
Computer
Eng
i
nee
ring
,
vol
.
2
,
pp.
175
-
182
,
20
12.
[8]
M.
Vata
ni
,
“
Tr
ansie
nt
Ana
l
y
s
i
s
of
Sw
it
chi
ng
the
Distribu
ted
Gene
ration
Units
in
Distri
buti
on
Network
s,”
Inte
rnational
Jo
urnal
of Appl
i
ed Powe
r E
ng
ine
eri
ng,
vol
.
5
,
pp
.
13
0
-
136,
2016
.
1
2
3
4
0
0
.
1
0
.
2
0
.
3
0
.
4
0
.
5
0
.
6
0
.
7
SBI
S
T
FT
C
W
T
P
W
T
S
-
t
r
a
n
s
f
o
r
m
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.
8
, N
o.
6
,
Dece
m
ber
201
8
:
4212
-
4220
4220
[9]
P.
Soundirara
ju
and
N.
Log
an
at
han
,
“
W
ave
let
Tra
nsform
s
Based
Pow
er
Tra
n
sform
erProte
ct
io
n
from
Magne
tic
Inrush Curre
nt
,
”
TEL
KOMNIKA
Indone
sian J
ourn
al
of
Elec
tric
al
Engi
ne
ering,
vol
.
14
,
pp
.
381
-
38
7,
2015
.
[10]
Z.
Guo,
“
Ti
m
e
-
fre
quency
r
epr
e
senta
ti
on
and
p
at
t
ern
re
cogni
tion
of
Doppler
Blood
Flow
Signal
for
Stenos
is
Cla
ss
ifi
c
at
ion
,
”
P
HD
the
sis,
Mc
Gill
Uni
ve
rs
it
y
,
Montreal
,
1993.
[11]
G.
Serbe
s,
e
t
al.
,
“
Dire
ct
ion
al
dual
-
tr
ee
comple
x
wave
le
t
pa
ck
et
tra
nsform
s
fo
r
proc
essing
quadr
at
ur
e
signal
s
,
”
Me
dic
a
l
&
Bi
o.
Eng.
&
Co
mputing
,
v
ol
.
54
,
pp
.
29
5
-
313,
2014
.
[12]
Z.
Zha
ng
,
et
al.
,
“
Ti
m
e
Freque
nc
y
W
ave
num
ber
Anal
y
sis
of
Su
rfa
ce
W
ave
s
an
d
Signal
Enha
n
ce
m
ent
Us
ing
S
-
tra
nsform
,
”
J. C
omp.
Ac
ous
,
vol
.
23,
2015.
[13]
Z.
Bouguil
a
,
et
al.
,
“
A
new
optim
iz
ed
Stockwel
l
tra
nsform
appl
i
ed
on
s
y
nthetic
and
rea
l
non
-
st
a
ti
onar
y
signa
ls
,
”
Digit
al
S
ignal P
roce
ss
ing,
v
ol
.
4
6,
pp
.
226
-
238
,
2015.
[14]
H.
K.
V
y
dana
and
A.
K.
Vuppala,
“
Detect
ion
of
fricat
ive
s
us
ing
S
-
t
ran
sform
,
”
The
Journal
of
the
Ac
oust
ical
Soci
e
ty
o
f Ameri
ca,
v
ol
/i
ss
ue:
14
0
(
5
)
,
pp
.
3895
-
3
907,
2016
.
[15]
S.
Kara
,
et
al
.
,
“
Spect
ra
l
broa
d
e
ning
of
lower
e
xtre
m
ity
venous
Doppler
signa
ls
using
STF
T
an
d
AR
m
odel
li
ng
,
”
Digit
al
S
ignal P
roce
ss
ing
,
vol
.
1
8,
pp
.
669
–
676
,
2008.
[16]
P.
I.
J.
Kee
ton
and
F.
S.
Sc
hli
ndwein
,
“
Spect
ra
l
broa
deni
n
g
of
cl
ini
c
al
Doppler
signal
s
using
F
F
T
and
aut
ore
gr
essive
m
odel
li
ng
,
”
Europ
ean
Journal
of
Ultrasound
,
vol
.
7
,
p
p
.
209
–
218
,
1
998.
[17]
F.
S.
Schli
ndwein
and
D.
H.
Eva
ns,
“
Sele
ct
ion
of
orde
r
of
aut
ore
gre
ss
ive
m
odel
for
spec
tra
l
an
aly
s
is
of
Doppler
ult
rasound
sign
a
ls
,
”
Ul
trasound Me
d
B
iol,
vol
/i
ss
ue:
16(1)
,
pp
.
81
-
91,
1990
.
[18]
Y.
Zha
ng
,
e
t
al
.
,
“
Corre
ct
ion
for
broa
den
ing
in
Doppler
blood
f
l
ow
spec
trum
est
imate
d
using
wa
vel
e
t
tr
ansform
,
”
Me
dic
a
l
Eng
inee
ring
&
Phy
sics
,
vol.
28
,
pp
.
596
–
603,
2006
.
[19]
X.
S.
Li,
e
t
al
.
,
“
Anal
y
sis
and
Si
m
pli
fic
a
ti
on
of
Thre
e
-
Dim
ensio
nal
Spac
e
Vec
to
r
PW
M
for
Thre
e
-
Phase
Four
-
Leg
Inve
rte
rs
,
”
I
EEE
Tr
ansacti
ons on
Industrial Elect
ronics
,
vol
.
58
,
p
p.
450
-
464
,
201
1.
[20]
K.
Kaluzy
nski
a
nd
T.
Palko,
“
Ef
fec
t
of
m
et
hod
a
nd
par
amet
ers
of
spec
tr
al
ana
l
y
si
s
on
select
ed
ind
ic
e
s
of
sim
ula
te
d
Doppler
spec
tra
,
”
Me
d
.
Bi
ol
.
Eng
.
Comput.
,
vo
l.
3
1,
pp
.
249
–
256
,
1993.
Evaluation Warning : The document was created with Spire.PDF for Python.