I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
1
5
,
No
.
2
,
A
p
r
il 2
0
2
6
,
p
p
.
1
7
4
6
~
1
7
6
1
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
2
.
p
p
1
7
4
6
-
1
7
6
1
1746
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Deep learni
ng
f
o
r
early
det
e
ction o
f
cardio
v
a
scula
r disea
ses
via
a
usculta
tion so
un
d clas
sifica
tion
Sh
re
y
a
s
K
a
s
t
ure
1
,
Su
dh
a
n
s
h
u M
a
ury
a
2,
3
,
Am
it
K
um
a
r
S
ha
rm
a
4
,
Sa
ntho
s
h Chit
ra
j
u G
o
pa
l V
a
rma
5
,
K
a
s
his
h
M
irza
6
,
F
irdo
us
Sa
da
f
M
o
ha
m
ma
d
I
s
m
a
il
1
1
S
y
m
b
i
o
si
s
I
n
st
i
t
u
t
e
o
f
Te
c
h
n
o
l
o
g
y
,
N
a
g
p
u
r
C
a
mp
u
s
,
S
y
m
b
i
o
s
i
s I
n
t
e
r
n
a
t
i
o
n
a
l
(
D
e
e
me
d
U
n
i
v
e
r
s
i
t
y
)
,
P
u
n
e
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
M
a
n
a
v
R
a
c
h
n
a
I
n
t
e
r
n
a
t
i
o
n
a
l
I
n
st
i
t
u
t
e
o
f
R
e
s
e
a
r
c
h
a
n
d
S
t
u
d
i
e
s
(
D
e
e
me
d
t
o
b
e
U
n
i
v
e
r
si
t
y
)
,
F
a
r
i
d
a
b
a
d
,
I
n
d
i
a
3
C
h
i
t
k
a
r
a
U
n
i
v
e
r
si
t
y
I
n
st
i
t
u
t
e
o
f
En
g
i
n
e
e
r
i
n
g
a
n
d
Te
c
h
n
o
l
o
g
y
,
C
h
i
t
k
a
r
a
U
n
i
v
e
r
si
t
y
,
R
a
j
p
u
r
a
,
I
n
d
i
a
4
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
a
n
d
C
o
mm
u
n
i
c
a
t
i
o
n
En
g
i
n
e
e
r
i
n
g
,
M
a
n
i
p
a
l
U
n
i
v
e
r
si
t
y
J
a
i
p
u
r
,
Ja
i
p
u
r
,
I
n
d
i
a
5
F
i
serv
I
n
c
.
,
M
i
l
w
a
u
k
e
e
,
U
n
i
t
e
d
S
t
a
t
e
s
6
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
G
r
a
p
h
i
c
Er
a
H
i
l
l
U
n
i
v
e
r
si
t
y
,
B
h
i
m
t
a
l
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Sep
1
2
,
2
0
2
4
R
ev
is
ed
J
an
3
0
,
2
0
2
6
Acc
ep
ted
Feb
6
,
2
0
2
6
He
a
rt
d
ise
a
se
s a
r
e
o
n
e
o
f
t
h
e
m
o
st p
ro
m
i
n
e
n
t
c
a
u
se
s o
f
d
e
a
th
g
l
o
b
a
ll
y
,
w
h
ic
h
re
q
u
ires
imm
e
d
iate
a
n
d
a
c
c
u
ra
te
d
ia
g
n
o
sis.
T
h
e
a
u
sc
u
lt
a
ti
o
n
m
e
th
o
d
s
u
se
d
in
c
o
n
v
e
n
ti
o
n
a
l
m
e
d
ica
l
p
ra
c
ti
c
e
,
wh
e
re
th
e
d
o
c
t
o
r
li
ste
n
s
t
o
t
h
e
so
u
n
d
s
p
ro
d
u
c
e
d
b
y
th
e
b
o
d
y
with
o
u
t
i
n
t
e
rv
e
n
ti
o
n
is
v
e
ry
i
n
e
ffe
c
ti
v
e
b
e
c
a
u
se
o
f
th
e
li
m
it
a
ti
o
n
s
i
n
th
e
a
c
tu
a
l
sk
il
ls
a
n
d
p
e
rc
e
p
ti
o
n
o
f
t
h
e
d
o
c
t
o
r.
T
h
e
m
a
in
g
o
a
l
o
f
th
is
p
ro
jec
t
wil
l
b
e
d
e
sig
n
in
g
a
m
o
b
il
e
-
b
a
se
d
s
y
ste
m
fo
r
th
e
e
a
rl
y
d
e
tec
ti
o
n
of
c
a
rd
i
o
v
a
sc
u
lar
d
ise
a
se
(
CVD
)
by
u
ti
li
z
i
n
g
d
e
e
p
lea
rn
i
n
g
fo
r
a
u
sc
u
lt
a
ti
o
n
so
u
n
d
c
las
sifica
ti
o
n
.
Th
e
a
p
p
r
o
a
c
h
i
n
v
o
lv
e
s
th
e
u
se
o
f
d
e
e
p
lea
rn
i
n
g
stru
c
tu
re
s to
c
las
sify
c
a
rd
iac
so
u
n
d
s in
t
o
n
o
rm
a
l
a
n
d
a
b
n
o
rm
a
l
p
a
tt
e
rn
s o
n
i
ts
o
wn
.
Wav
e
let
tra
n
sfo
rm
a
ti
o
n
s,
ti
m
e
-
fre
q
u
e
n
c
y
re
p
re
se
n
tatio
n
s
,
a
n
d
M
e
l
-
fre
q
u
e
n
c
y
c
e
p
stra
l
c
o
e
fficie
n
ts
(M
F
CC)
h
a
v
e
b
e
e
n
u
se
d
i
n
fe
a
tu
re
e
x
trac
ti
o
n
.
Th
e
Re
sN
e
t1
5
2
V2
m
o
d
e
l
sh
o
we
d
h
ig
h
c
las
sifica
ti
o
n
p
e
rfo
rm
a
n
c
e
with
a
re
a
u
n
d
e
r
t
h
e
re
c
e
iv
e
r
o
p
e
ra
ti
n
g
c
h
a
ra
c
teristic
c
u
rv
e
(
AU
ROC
)
o
f
0
.
9
7
9
7
a
n
d
0
.
9
6
3
6
o
n
two
d
a
tas
e
ts.
Co
n
trar
y
to
th
a
t,
d
a
t
a
a
u
g
m
e
n
tatio
n
,
h
y
p
e
rp
a
ra
m
e
ter
o
p
ti
m
iza
ti
o
n
,
a
tt
e
n
ti
o
n
m
e
c
h
a
n
ism
s,
a
s
we
ll
a
s
in
p
u
t
-
o
u
tp
u
t
re
sid
u
a
l
c
o
n
n
e
c
ti
o
n
s,
le
d
to
b
e
tt
e
r
fu
n
c
ti
o
n
a
li
ty
a
n
d
in
terp
re
tab
il
it
y
.
T
h
is
re
se
a
rc
h
se
e
k
s
to
o
v
e
rc
o
m
e
th
e
li
m
it
a
ti
o
n
s
o
f
trad
it
io
n
a
l
ste
t
h
o
sc
o
p
e
u
se
th
r
o
u
g
h
th
e
in
c
o
r
p
o
ra
ti
o
n
o
f
so
p
h
isti
c
a
te
d
a
lg
o
rit
h
m
s
a
n
d
t
h
e
a
v
a
il
a
b
il
i
ty
o
f
m
o
b
i
le
tec
h
n
o
lo
g
y
t
h
a
t
c
o
u
ld
re
su
lt
i
n
e
a
rly
d
iag
n
o
sis a
n
d
p
re
v
e
n
ti
o
n
o
f
CVD
s
,
e
sp
e
c
ially
in
u
n
d
e
rp
ri
v
il
e
g
e
d
a
re
a
s.
K
ey
w
o
r
d
s
:
Au
s
cu
ltatio
n
class
if
icatio
n
C
ar
d
iac
d
is
ea
s
e
d
etec
tio
n
Hea
r
t so
u
n
d
d
e
n
o
is
in
g
Hy
p
er
p
ar
a
m
eter
o
p
tim
izatio
n
Sig
n
al
s
eg
m
en
tatio
n
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
:
Am
it Ku
m
ar
Sh
ar
m
a
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
a
n
d
C
o
m
m
u
n
icatio
n
E
n
g
in
ee
r
i
n
g
,
Ma
n
ip
al
Un
iv
er
s
ity
J
aip
u
r
J
aip
u
r
,
R
ajasth
an
,
I
n
d
ia
E
m
ail:
am
it.sh
ar
m
a@
jaip
u
r
.
m
an
ip
al.
ed
u
1.
I
NT
RO
D
UCT
I
O
N
W
h
en
ca
r
d
io
v
ascu
lar
d
is
ea
s
e
(
C
VDs
)
ar
e
a
to
p
-
o
n
e
ca
u
s
e
o
f
m
o
r
tal
ity
,
wh
ich
m
ea
n
s
th
e
s
p
ee
d
an
d
th
e
ac
cu
r
ac
y
o
f
th
e
d
iag
n
o
s
is
d
eter
m
in
e
th
e
p
atien
ts
'
o
u
tco
m
es.
L
is
ten
in
g
to
h
ea
r
t
s
o
u
n
d
s
with
tr
ad
itio
n
al
au
s
cu
ltatio
n
h
as
b
ec
o
m
e
a
co
m
p
l
ex
m
ed
ical
p
r
ac
tice
with
s
ig
n
if
ican
t
r
o
o
m
f
o
r
er
r
o
r
i
f
n
o
t
p
er
f
o
r
m
ed
b
y
h
ea
lth
ca
r
e
p
r
o
f
ess
io
n
als we
ll
-
t
r
ain
ed
in
s
teth
o
s
co
p
e
u
s
e.
Dee
p
lear
n
in
g
an
d
s
ig
n
al
p
r
o
ce
s
s
in
g
tech
n
iq
u
es h
av
e
b
ee
n
u
tili
s
ed
to
au
to
m
atica
ll
y
an
d
ac
cu
r
ately
ca
teg
o
r
is
e
h
ea
r
t
s
o
u
n
d
s
,
m
ak
in
g
th
e
ea
r
ly
id
en
tific
atio
n
o
f
C
VDs
f
ea
s
ib
le.
Ou
r
r
esear
ch
aim
s
to
b
u
ild
a
m
o
b
ile
-
r
ea
d
y
s
o
lu
tio
n
f
o
r
ea
r
ly
d
etec
tio
n
o
f
C
VDs
u
s
in
g
a
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
(
C
NN)
-
b
ased
au
s
cu
ltatio
n
s
o
u
n
d
class
if
icatio
n
.
T
h
e
r
aised
ap
p
r
o
ac
h
is
b
ased
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Dee
p
lea
r
n
in
g
fo
r
ea
r
ly
d
etec
t
io
n
o
f c
a
r
d
io
v
a
s
cu
la
r
d
is
ea
s
es v
ia
a
u
s
cu
lta
tio
n
s
o
u
n
d
…
(
S
h
r
ey
a
s
K
a
s
tu
r
e
)
1747
o
n
s
tate
-
of
-
th
e
-
ar
t
d
ee
p
lear
n
i
n
g
to
o
ls
,
s
u
ch
as
C
N
Ns
,
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
s
(
R
NNs
)
,
an
d
ev
en
h
y
b
r
id
m
o
d
els
lik
e
C
NN
-
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
L
STM
)
,
wh
ich
ca
n
p
er
f
o
r
m
au
to
m
at
ic
h
ea
r
tb
ea
t
s
o
u
n
d
class
if
icatio
n
,
wh
ich
h
elp
s
to
d
etec
t
p
o
ten
tial
C
VDs
[
1
]
,
[
2
]
.
Her
e,
we
im
p
lem
en
t
p
h
o
n
o
ca
r
d
io
g
r
a
m
(
PC
G
)
s
ig
n
als,
d
ig
ital
r
ec
o
r
d
s
tak
en
f
r
o
m
th
e
h
ea
r
t,
as
in
p
u
t
d
ata
f
o
r
th
e
d
ee
p
lear
n
in
g
n
etwo
r
k
s
.
T
ec
h
n
iq
u
es
s
u
ch
as
d
en
o
is
in
g
,
s
eg
m
e
n
tatio
n
,
an
d
f
ea
tu
r
e
ex
tr
ac
tio
n
ar
e
ap
p
lied
s
eq
u
en
tially
to
th
e
PC
G
s
ig
n
als.
I
t
in
cr
ea
s
es
th
e
r
eliab
ilit
y
an
d
b
r
o
a
d
er
ap
p
licatio
n
o
f
th
e
tr
ain
in
g
d
ata
u
s
e
d
b
y
th
e
class
if
ier
s
.
T
ec
h
n
iq
u
es
s
u
ch
as
wav
elet
tr
an
s
f
o
r
m
atio
n
s
,
tim
e
-
f
r
e
q
u
en
cy
r
ep
r
esen
tatio
n
s
,
an
d
M
el
-
f
r
eq
u
en
c
y
ce
p
s
tr
al
co
ef
f
icien
ts
(
MFC
C
)
,
s
u
ch
as
s
ca
lo
g
r
am
s
,
ar
e
ex
p
lo
r
ed
f
o
r
e
f
f
ec
tiv
e
f
ea
tu
r
e
ex
tr
ac
tio
n
.
T
h
e
u
tili
ze
d
s
y
s
tem
is
o
f
a
m
o
b
ile
tech
n
o
l
o
g
y
ty
p
e;
h
en
ce
,
ac
q
u
ir
in
g
h
ea
r
t
s
o
u
n
d
d
ata
is
r
elativ
ely
ea
s
y
th
r
o
u
g
h
m
o
b
ile
d
e
v
ices
o
r
p
o
r
tab
le
elec
tr
o
n
ic
s
teth
o
s
co
p
es.
Mo
r
e
o
v
er
,
ea
r
ly
d
etec
tio
n
an
d
d
iag
n
o
s
is
o
f
C
VDs
ar
e
en
ab
led
in
r
em
o
te
o
r
r
eso
u
r
ce
-
c
o
n
s
tr
ain
ed
s
ettin
g
s
.
Ap
p
ly
in
g
d
ee
p
lear
n
in
g
m
o
d
els
o
n
m
o
b
ile
d
ev
ices
m
ay
g
iv
e
r
is
e
to
a
s
i
m
p
le
an
d
r
ea
ch
a
b
le
o
p
tio
n
f
o
r
h
ea
lth
ca
r
e
p
r
o
f
ess
io
n
als
an
d
o
r
d
in
a
r
y
g
u
y
s
.
I
n
ad
d
itio
n
,
th
e
o
n
-
d
ev
ice
p
r
o
ce
s
s
s
u
p
p
o
r
ts
r
ea
l
-
tim
e
an
aly
s
is
an
d
p
r
o
v
i
d
es
u
s
er
s
with
im
m
ed
iate
f
ee
d
b
ac
k
o
n
th
eir
ca
r
d
i
o
v
ascu
lar
h
ea
lth
,
p
u
ttin
g
it
u
n
d
e
r
th
e
s
p
o
tlig
h
t.
On
e
o
f
th
e
cr
u
cial
elem
e
n
ts
o
f
th
e
class
if
ier
is
th
e
en
h
an
ce
m
e
n
t
o
f
s
o
p
h
is
ticated
d
ee
p
lear
n
in
g
al
g
o
r
ith
m
s
,
s
u
ch
as
C
NN
s
an
d
R
NN
s
,
with
in
n
o
v
ativ
e
h
y
b
r
i
d
m
o
d
els
lik
e
C
NN
-
L
STM
.
T
h
is
ar
ch
itectu
r
e
is
ad
eq
u
ate
f
o
r
ass
es
s
in
g
au
d
io
an
d
s
ig
n
al
p
r
o
ce
s
s
in
g
o
f
h
ea
r
t
s
o
u
n
d
class
if
icatio
n
[
3
]
an
d
also
f
its
th
e
task
o
f
class
if
y
in
g
th
e
h
e
ar
t
s
o
u
n
d
.
C
L
Vs
p
r
ef
er
o
b
s
er
v
in
g
th
e
p
atter
n
s
o
f
tim
e
an
d
s
p
ac
e
in
d
ata,
wh
ich
is
wh
y
th
ey
ar
e
ess
en
tial
f
o
r
in
v
esti
g
atin
g
th
e
tim
e
-
d
o
m
ain
r
ep
r
esen
tatio
n
s
o
f
h
ea
r
t
b
ea
ts
.
On
th
e
o
th
er
h
an
d
,
it
s
h
o
u
ld
b
e
n
o
ted
th
at
R
NNs
ar
e
g
o
o
d
at
h
an
d
lin
g
d
ata
s
eq
u
en
ce
s
an
d
ca
n
m
o
d
el
th
e
d
ep
en
d
en
cy
b
etwe
en
c
o
n
s
ec
u
tiv
e
h
ea
r
t
s
o
u
n
d
s
ig
n
als
[
4
]
.
T
h
e
C
NN
-
L
STM
h
y
b
r
id
m
o
d
el
p
o
o
ls
its
ad
v
an
tag
es
b
y
co
m
b
in
in
g
th
e
s
tr
en
g
t
h
s
o
f
b
o
th
ar
ch
itectu
r
es.
Seg
m
en
tatio
n
s
tr
ateg
ies
h
elp
is
o
late
th
e
co
m
p
o
n
en
ts
o
f
h
ea
r
t
s
o
u
n
d
s
,
in
clu
d
i
n
g
t
h
e
s
y
s
to
lic
an
d
d
iast
o
lic
p
h
ases
,
s
o
th
at
ea
ch
s
o
u
n
d
elem
en
t c
an
b
e
a
n
aly
ze
d
a
n
d
b
etter
class
if
ied
[
5
]
.
Ho
wev
er
,
th
er
e
ar
e
a
n
u
m
b
er
o
f
lim
itatio
n
s
in
th
is
d
o
m
ain
.
Pre
v
io
u
s
wo
r
k
s
h
av
e
s
o
m
e
d
r
awb
ac
k
s
r
eg
ar
d
in
g
th
e
v
ar
iety
o
f
d
ata
s
et
s
u
s
ed
,
th
e
clas
s
im
b
alan
ce
p
r
o
b
lem
,
a
n
d
m
o
d
el
o
p
ti
m
izatio
n
f
o
r
b
etter
class
if
icatio
n
ac
cu
r
ac
y
a
n
d
g
en
er
aliza
tio
n
.
Mo
r
eo
v
er
,
th
e
ap
p
licatio
n
o
f
s
u
ch
s
o
l
u
tio
n
s
i
n
t
h
e
m
o
b
ile
en
v
ir
o
n
m
en
t
is
cr
u
cial
to
m
ak
e
ac
ce
s
s
a
s
we
ll
as
ea
r
ly
d
iag
n
o
s
is
in
r
eg
io
n
s
with
lim
ited
r
e
s
o
u
r
ce
s
.
T
h
er
ef
o
r
e
,
th
is
s
tu
d
y
p
r
o
p
o
s
es
to
ex
ten
d
th
e
wo
r
k
f
u
r
th
er
b
y
estab
lis
h
in
g
a
r
eliab
le
an
d
p
o
r
tab
le
d
ee
p
lear
n
in
g
b
ased
C
VD
d
etec
tio
n
th
r
o
u
g
h
au
s
cu
l
tatio
n
s
o
u
n
d
class
if
icatio
n
.
T
h
e
k
ey
co
n
tr
ib
u
tio
n
s
o
f
th
is
s
tu
d
y
ar
e:
i)
E
x
p
lo
r
atio
n
o
f
ad
v
an
ce
d
d
ata
au
g
m
en
tatio
n
tec
h
n
iq
u
es,
in
clu
d
in
g
tim
e
s
tr
etch
in
g
,
p
itch
s
h
if
tin
g
,
an
d
s
p
ec
tr
o
g
r
am
a
u
g
m
en
tatio
n
,
to
ex
p
an
d
d
iv
er
s
ity
o
f
tr
ain
in
g
d
a
taset a
n
d
im
p
r
o
v
e
m
o
d
el
g
en
e
r
aliza
tio
n
.
ii)
Ap
p
licatio
n
o
f
co
m
p
r
eh
e
n
s
iv
e
h
y
p
e
r
p
ar
am
eter
o
p
tim
izati
o
n
s
tr
ateg
ies,
co
m
b
in
in
g
g
r
i
d
s
ea
r
ch
,
a
n
d
B
ay
esian
o
p
tim
izatio
n
,
to
f
i
n
e
-
tu
n
e
th
e
d
ee
p
lear
n
in
g
ar
ch
ite
ctu
r
es f
o
r
e
n
h
an
ce
d
class
if
icat
io
n
.
iii)
Per
f
o
r
m
an
ce
an
al
y
s
is
o
f
th
e
latest
d
ee
p
lear
n
in
g
m
o
d
els
lik
e
R
e
s
Net1
5
2
V2
,
Mo
b
ileNet,
an
d
Xce
p
tio
n
N
et
o
n
two
ca
r
d
io
v
as
cu
lar
s
o
u
n
d
d
atab
ases
in
m
o
b
i
le
d
ev
ice
co
m
p
atib
ilit
y
.
iv
)
Ad
o
p
tio
n
o
f
p
r
o
f
itab
le
a
r
ch
ite
ctu
r
al
am
en
d
m
e
n
ts
s
u
ch
as
atten
tio
n
m
ec
h
a
n
is
m
s
,
r
esid
u
al
c
o
n
n
ec
tio
n
s
,
as
well
as c
o
m
b
in
atio
n
m
eth
o
d
s
,
with
th
e
aim
o
f
en
h
an
cin
g
t
h
e
in
ter
p
r
etab
ilit
y
o
f
t
h
e
m
o
d
els
,
to
g
eth
er
with
th
e
m
o
d
el
r
o
b
u
s
tn
ess
.
2.
L
I
T
E
R
AT
U
RE
R
E
VI
E
W
2
.
1
.
Dee
p
lea
rning
t
ec
hn
iqu
es f
o
r
hea
rt
s
o
un
d c
la
s
s
if
ica
t
io
n a
nd
ca
rdio
v
a
s
cula
r
dis
ea
s
e
dia
g
no
s
is
T
h
e
cr
ea
tiv
e
,
lo
w
-
c
o
s
t
h
ea
r
t
b
e
at
m
ea
s
u
r
em
en
t
to
o
l
cr
ea
ted
b
y
th
e
g
r
o
u
p
o
f
r
esear
c
h
er
s
R
o
y
et
a
l.
[
6
]
aim
ed
to
a
d
d
r
ess
th
e
is
s
u
es
o
f
tr
ad
itio
n
al
s
teth
o
s
co
p
e
u
s
e
r
aised
b
y
th
e
p
a
n
d
em
ic.
B
esid
es
u
s
in
g
d
if
f
er
en
t
h
y
p
er
p
ar
am
eter
tu
n
i
n
g
tech
n
i
q
u
es,
o
u
r
m
o
d
els
also
p
er
f
o
r
m
ed
b
etter
wh
en
th
e
v
alu
es
o
f
lear
n
in
g
r
ates,
d
r
o
p
o
u
t
r
ates,
a
n
d
h
id
d
e
n
la
y
er
co
n
f
ig
u
r
atio
n
s
wer
e
ad
ju
s
ted
ac
co
r
d
in
g
ly
.
Sq
u
ee
ze
-
an
d
-
ex
citatio
n
b
lo
ck
s
wer
e
in
tr
o
d
u
ce
d
to
im
p
r
o
v
e
d
ee
p
lear
n
i
n
g
m
o
d
els,
a
n
d
ef
f
i
cien
cy
b
ec
a
m
e
a
p
r
im
a
r
y
co
n
ce
r
n
.
R
en
et
a
l.
[
7
]
h
av
e
r
ec
o
g
n
ized
th
at
tr
ad
itio
n
al
m
ac
h
in
e
lear
n
in
g
tech
n
iq
u
e
s
will
h
av
e
is
s
u
es
o
f
p
er
f
o
r
m
an
ce
lim
itatio
n
an
d
s
ca
lab
ilit
y
as
th
ey
ad
d
r
ess
b
ig
an
d
co
m
p
lex
h
ea
r
t
s
o
u
n
d
d
ata.
T
h
eir
ex
p
er
im
en
ts
d
e
m
o
n
s
tr
ate
th
at
d
ee
p
lear
n
in
g
m
o
d
els
g
en
er
ally
h
a
v
e
a
s
ig
n
if
ican
t
p
o
ten
tial
to
s
u
r
p
ass
th
e
r
esu
lts
o
f
class
ic
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
in
co
r
r
ec
tly
d
is
cr
im
in
atin
g
d
if
f
e
r
en
t
h
ea
r
t
s
o
u
n
d
co
n
d
itio
n
s
.
I
t
en
ab
les
h
ea
lt
h
ca
r
e
p
r
o
v
id
er
s
to
in
cr
ea
s
e
th
eir
u
n
d
er
s
tan
d
i
n
g
o
f
th
is
d
ec
is
io
n
-
m
ak
in
g
p
r
o
c
ess
,
th
u
s
d
ev
elo
p
in
g
th
eir
tr
u
s
t
in
th
e
p
er
s
o
n
.
Ali
et
a
l.
[
8
]
d
e
v
elo
p
ed
L
U
-
Net,
a
d
ee
p
lear
n
in
g
ar
ch
itec
tu
r
e
to
d
en
o
is
e
th
e
h
ea
r
t
s
o
u
n
d
s
ca
p
tu
r
ed
u
s
in
g
d
ig
ital
s
teth
o
s
co
p
es.
T
h
e
ai
m
was
to
b
u
ild
a
d
ee
p
en
c
o
d
er
-
d
ec
o
d
er
s
tr
u
ctu
r
e
th
at
c
o
m
b
in
ed
th
e
L
STM
m
o
d
u
les
(
b
i
-
d
ir
ec
tio
n
al)
to
c
a
p
tu
r
e
th
e
b
ea
t
p
atter
n
s
wh
ile
u
tili
zin
g
th
is
in
f
o
r
m
atio
n
f
o
r
b
ea
t
r
ec
o
n
s
tr
u
ctio
n
.
T
h
e
r
esear
ch
r
esu
lts
p
r
o
v
e
d
t
h
at,
o
n
a
v
er
ag
e,
th
er
e
was
5
.
5
7
d
B
o
f
s
o
u
n
d
S/N
atten
u
atio
n
f
o
r
all
o
f
th
e
ex
am
in
ed
r
ec
o
r
d
i
n
g
s
,
i.e
.
,
b
o
t
h
f
o
r
t
h
e
s
ig
n
als af
f
ec
ted
a
r
tific
ially
b
y
th
e
n
o
is
e
an
d
f
o
r
th
e
r
ea
l
-
wo
r
ld
d
ata.
L
i
et
a
l.
[
9
]
p
r
o
p
o
s
e
d
th
ei
r
n
e
w
a
p
p
r
o
ac
h
t
o
r
e
c
o
g
n
i
zi
n
g
h
e
a
r
t
b
e
ats
th
r
o
u
g
h
e
n
h
a
n
c
e
d
MF
C
C
f
e
at
u
r
es
u
s
i
n
g
t
h
e
d
e
e
p
r
esi
d
u
al
n
e
tw
o
r
k
(
DR
N
)
.
Si
g
n
al
p
r
e
-
p
r
o
ce
s
s
i
n
g
was
p
er
f
o
r
m
e
d
o
n
h
ea
r
t
s
o
u
n
d
s
,
a
n
d
i
m
p
r
o
v
e
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
2
,
Ap
r
il 2
0
2
6
:
1
7
4
6
-
1
7
6
1
1748
MFC
C
c
o
e
f
f
i
ci
en
ts
we
r
e
c
alc
u
la
te
d
,
w
h
i
c
h
ac
t
e
d
as
f
ea
t
u
r
es
f
o
r
d
e
ep
n
e
u
r
al
n
et
wo
r
k
(
DNN)
.
T
h
e
r
esu
lts
r
e
v
e
ale
d
th
at
th
e
b
e
tte
r
M
FC
C
f
e
at
u
r
es
p
e
r
f
o
r
m
e
d
wi
th
h
i
g
h
e
r
s
en
s
iti
v
it
y
,
s
p
ec
i
f
i
city
,
a
n
d
ac
cu
r
a
cy
,
em
p
h
asi
zi
n
g
t
h
e
ir
s
u
p
er
io
r
i
ty
i
n
c
o
n
v
e
y
i
n
g
r
ele
v
a
n
t
i
n
f
o
r
m
at
i
o
n
d
u
r
i
n
g
h
ea
r
t
s
o
u
n
d
class
if
ic
ati
o
n
p
r
o
ce
s
s
[
1
0
]
.
2
.
2
.
Aut
o
m
a
t
ed
hea
rt
s
o
un
d si
g
na
l pro
ce
s
s
ing
a
nd
deno
i
s
ing
Al
-
I
s
s
a
an
d
Alq
u
d
ah
[
1
1
]
p
r
o
p
o
s
ed
a
m
o
d
el
th
at
was
tr
ai
n
ed
an
d
e
v
alu
ated
o
n
t
h
e
O
HPDA
an
d
C
o
m
p
u
tin
g
in
C
ar
d
io
lo
g
y
C
h
allen
g
e
2
0
1
6
d
atasets
.
I
n
th
e
f
iv
e
-
class
task
in
v
o
lv
in
g
h
ea
r
t
s
o
u
n
d
s
,
an
ex
p
lo
r
ed
d
ataset
is
o
p
en
.
T
h
e
r
esear
ch
er
s
also
ex
p
lo
ited
th
e
f
r
eq
u
en
c
y
d
o
m
ain
in
p
u
t
s
ig
n
al,
o
u
tp
u
t
tin
g
th
e
o
p
en
-
h
ea
r
t
s
o
u
n
d
d
ata
class
if
icatio
n
r
ate
o
f
9
9
.
7
3
%
an
d
9
0
.
6
5
%
f
o
r
th
e
Ph
y
s
io
Net/C
o
m
p
u
tin
g
i
n
C
ar
d
io
lo
g
y
2
0
1
6
ch
allen
g
e
d
ataset.
T
h
ey
r
e
p
o
r
ted
th
at
p
atien
ts
co
u
ld
b
en
ef
it
f
r
o
m
th
e
m
o
d
el's
f
ea
tu
r
e,
wh
ich
s
h
o
ws
ca
r
d
iac
d
is
ea
s
es
a
t
th
e
in
itial
s
tag
es
o
f
d
ev
elo
p
m
e
n
t,
an
y
wh
e
r
e
in
th
e
wo
r
ld
wh
er
e
a
d
o
cto
r
is
u
n
av
ailab
le.
J
o
s
h
i
et
a
l.
[
1
2
]
s
u
g
g
est
an
AI
-
b
ased
s
y
s
tem
th
at
also
p
er
m
it
s
au
to
m
atic,
alb
eit
r
ea
l
-
tim
e,
d
iag
n
o
s
is
o
f
C
VD
s
.
Pro
m
is
in
g
r
esu
lts
wer
e
o
b
tain
ed
f
r
o
m
ev
alu
atin
g
t
h
e
s
u
g
g
es
ted
s
o
lu
tio
n
:
ap
p
ly
i
n
g
f
iv
e
-
f
o
l
d
cr
o
s
s
-
v
alid
atio
n
,
th
e
1
D
-
C
NN
cla
s
s
if
ier
wo
r
k
ed
o
u
t
r
ig
h
t
f
o
r
ca
r
d
iac
illn
ess
with
an
ac
cu
r
ac
y
o
f
ab
o
u
t
9
6
.
9
5
%,
an
d
f
o
r
th
e
2D
-
C
NN,
it
was
9
7
.
8
5
%.
As
a
r
esu
lt,
b
o
th
class
if
ier
s
s
h
o
w
th
at
th
e
an
aly
s
is
o
f
elec
tr
o
ca
r
d
io
g
r
a
m
(
E
C
G
)
s
ig
n
als
b
ased
o
n
d
ee
p
lear
n
in
g
tech
n
o
l
o
g
y
is
a
p
r
o
m
i
s
in
g
m
eth
o
d
f
o
r
d
iag
n
o
s
in
g
ca
r
d
iac
d
is
ea
s
es.
B
aik
u
v
ek
o
v
et
a
l.
[
1
3
]
d
ev
is
e
d
an
au
to
m
atic
class
if
icatio
n
s
y
s
tem
f
o
r
d
if
f
er
en
t b
o
r
d
er
s
o
f
h
ea
r
t d
is
ea
s
es u
s
in
g
d
ig
ital
PC
G
s
ig
n
als
an
d
d
ee
p
lear
n
in
g
m
et
h
o
d
s
.
Sp
ec
if
ical
ly
,
th
e
p
r
o
p
o
s
ed
d
ee
p
C
NN
m
o
d
el
s
aw
r
eg
u
lar
h
ea
r
tb
ea
ts
co
r
r
ec
tly
id
en
tifie
d
9
3
.
5
0
%
o
f
t
h
e
tim
e,
an
d
ab
n
o
r
m
al
h
ea
r
t
s
o
u
n
d
s
wer
e
id
en
tifi
ed
alm
o
s
t
clo
s
e
to
9
3
.
2
5
%.
Un
d
en
iab
ly
,
th
e
f
astes
t
s
ce
n
e
was
th
e
o
n
e
wh
er
e
we
d
escr
ib
ed
th
e
n
ew
s
teth
o
s
co
p
e
th
at
co
u
l
d
co
m
p
lete
th
e
test
in
o
n
l
y
1
5
s
ec
o
n
d
s
,
an
d
th
is
is
a
s
ig
n
if
ican
t
d
if
f
er
en
ce
f
r
o
m
th
e
p
r
ev
io
u
s
m
ec
h
a
n
is
m
.
B
o
n
d
ar
ev
a
et
a
l.
[
1
4
]
s
tag
e
a
b
lo
ck
s
eg
m
e
n
tatio
n
-
f
r
ee
tech
n
iq
u
e
f
o
r
class
if
y
in
g
h
ea
r
t
s
o
u
n
d
s
in
to
n
o
r
m
al
a
n
d
m
u
r
m
u
r
g
r
o
u
p
s
.
I
t
r
elied
o
n
th
e
d
is
cr
ete
wav
elet
tr
a
n
s
f
o
r
m
atio
n
f
o
r
n
o
is
e
ca
n
ce
llin
g
,
f
ea
tu
r
e
e
x
tr
ac
tio
n
,
f
ea
tu
r
e
s
elec
tio
n
,
an
d
class
if
icatio
n
o
f
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
es
(
SVMs)
an
d
DNN
s
.
Fu
r
th
er
m
o
r
e,
th
e
y
o
b
tain
ed
th
at
th
e
am
o
u
n
t
o
f
tr
ain
in
g
d
ata
an
d
th
e
p
atien
t
-
in
d
ep
e
n
d
en
t
s
ettin
g
h
av
e
d
ir
ec
tly
ca
u
s
ed
th
e
u
p
r
is
in
g
o
f
th
e
m
o
d
el's
p
er
f
o
r
m
an
ce
o
n
s
ev
er
al
ev
al
u
atio
n
s
tan
d
ar
d
s
.
T
h
e
m
o
d
el
h
a
d
8
1
%
o
f
t
h
e
tr
u
e
p
o
s
itiv
es in
th
e
two
class
es a
n
d
9
6
% o
f
th
e
tr
u
e
p
o
s
itiv
es in
th
e
p
atien
t
-
d
ep
e
n
d
en
t
s
ettin
g
.
2
.
3
.
M
a
chine
lea
rning
a
nd
deep
lea
rning
f
o
r
ca
rdio
v
a
s
cula
r
dis
ea
s
e
risk
predict
io
n
Su
b
r
am
an
i
et
a
l.
[
1
5
]
d
ev
el
o
p
ed
m
ac
h
in
e
lear
n
in
g
m
o
d
el
s
r
elatin
g
to
C
VD
r
is
k
ass
e
s
s
m
en
t
an
d
f
u
r
th
er
lo
o
k
ed
at
h
o
w
th
ey
p
er
f
o
r
m
e
d
s
tatis
tical
ly
co
m
p
ar
ed
with
tr
ad
itio
n
al
s
tatis
t
ical
m
eth
o
d
s
.
T
h
e
h
ea
r
t
d
is
ea
s
e
d
ataset
was
u
s
ed
to
tr
ai
n
th
e
m
o
d
els
an
d
e
v
alu
ate
th
eir
p
er
f
o
r
m
an
ce
.
T
h
r
o
u
g
h
t
h
e
p
r
ec
is
e
ap
p
r
aisal
o
f
p
atien
ts
at
h
i
g
h
-
r
i
s
k
s
tatu
s
,
h
o
s
p
itals
an
d
h
ea
lth
ca
r
e
p
r
o
v
id
e
r
s
ca
n
s
tr
iv
e
to
i
m
p
lem
en
t
p
r
o
ac
tiv
e
clin
ical
ac
tio
n
s
th
at
ca
n
f
u
r
th
er
lim
it
th
e
r
is
k
s
an
d
e
n
h
an
c
e
p
atien
t
r
esu
lts
.
Ob
a
y
y
a
et
a
l.
[
1
6
]
d
esig
n
e
d
a
n
au
to
m
ated
ca
r
d
i
o
v
ascu
lar
d
is
e
ase
d
iag
n
o
s
is
u
s
in
g
h
o
n
ey
b
a
d
g
er
o
p
tim
izatio
n
with
a
m
o
d
if
ied
d
ee
p
lear
n
in
g
(
AC
VD
-
HB
OM
DL
)
alg
o
r
ith
m
,
s
ig
n
if
ican
tly
im
p
r
o
v
in
g
C
VD
d
iag
n
o
s
is
's
s
p
ee
d
an
d
p
r
ec
is
io
n
.
Gr
o
win
g
o
n
th
e
p
r
ev
alen
ce
o
f
th
e
AC
VD
-
HB
OM
DL
m
eth
o
d
f
o
r
ev
al
u
atin
g
h
ea
lth
co
n
d
itio
n
s
will
p
r
o
v
e
th
at
it
h
as
a
b
etter
p
er
f
o
r
m
an
ce
th
an
b
ag
g
i
n
g
,
J
4
8
,
an
d
o
th
er
s
,
in
clu
d
in
g
s
im
p
le
class
if
icatio
n
an
d
r
eg
r
ess
io
n
tr
ee
s
(
SC
)
,
r
ed
u
ce
d
er
r
o
r
p
r
u
n
in
g
tr
ee
(
R
E
PTr
ee
)
,
ar
tific
ial
n
eu
r
al
n
etwo
r
k
(
ANN
)
,
a
n
d
SVM,
with
an
im
p
r
o
v
ed
ac
cu
r
ac
y
o
f
9
9
.
3
9
%.
M
ek
a
h
lia
et
a
l.
[
1
7
]
p
r
esen
ted
a
d
ee
p
lear
n
in
g
a
p
p
r
o
ac
h
f
o
r
class
if
y
in
g
h
ea
r
t
s
o
u
n
d
ca
teg
o
r
ies
r
elate
d
to
C
VD
s
.
T
h
r
o
u
g
h
th
eir
s
tu
d
y
,
t
h
ey
b
r
o
u
g
h
t
ab
o
u
t
th
e
i
d
ea
th
at
th
e
s
ca
lo
g
r
am
r
ep
r
esen
tatio
n
o
f
h
ea
r
t
s
o
u
n
d
s
ig
n
als
co
u
p
led
with
th
e
C
NN
-
d
ir
ec
t
ac
y
clic
g
r
ap
h
(
DA
G
)
m
o
d
el,
wh
ich
is
s
u
p
er
io
r
to
o
t
h
er
m
eth
o
d
s
in
class
if
icatio
n
ac
cu
r
ac
y
,
co
m
es f
i
r
s
t.
W
an
g
et
a
l.
[
1
8
]
p
u
t f
o
r
wa
r
d
a
co
m
p
u
ter
ized
d
iag
n
o
s
tic
m
eth
o
d
,
wh
ich
is
b
ased
o
n
h
ea
r
t
s
o
u
n
d
wav
e
s
to
class
if
y
ca
r
d
io
v
ascu
lar
ab
n
o
r
m
alities
.
T
h
e
y
in
tr
o
d
u
ce
d
a
n
ew
q
u
ality
o
f
h
ea
r
tb
ea
t
wav
e
d
ata
r
ec
o
r
d
f
o
r
p
atien
ts
with
h
y
p
er
te
n
s
io
n
.
T
h
e
p
r
o
p
o
s
ed
s
tr
ateg
y
h
as
ex
h
ib
ited
b
etter
p
er
f
o
r
m
a
n
ce
th
an
th
e
o
th
e
r
b
aselin
es,
lik
e
L
STM
an
d
C
NN,
wh
ich
wer
e
ca
p
ab
le
o
f
o
n
ly
r
ea
ch
in
g
lo
wer
ac
cu
r
ac
y
o
n
t
h
e
s
am
e
task
.
W
e
ca
m
e
u
p
with
th
e
r
esear
ch
f
in
d
in
g
s
f
r
o
m
Z
h
o
u
et
a
l.
[
1
9
]
r
eg
ar
d
in
g
th
e
ef
f
ec
t
o
f
d
if
f
er
e
n
t
d
ata
a
u
g
m
e
n
tatio
n
tech
n
iq
u
es
o
n
th
e
class
if
y
in
g
p
o
wer
o
f
a
C
NN
m
o
d
el
th
at
u
tili
ze
s
s
p
ec
tr
o
g
r
am
s
to
d
if
f
er
en
tiate
b
etwe
en
h
ea
lth
y
an
d
a
b
n
o
r
m
al
h
ea
r
t
s
o
u
n
d
s
.
T
h
e
attr
ib
u
ted
s
u
cc
ess
o
f
a
s
p
ec
if
ic
au
g
m
en
tatio
n
tech
n
iq
u
e
is
b
ec
au
s
e
it
s
ati
s
f
ies
b
o
th
th
e
cr
iter
ia
f
o
r
th
e
"p
h
y
s
io
lo
g
i
ca
l
co
n
s
tr
ain
t"
an
d
"sp
ec
tr
o
m
etr
ic
co
n
s
tr
ain
t,"
w
h
ich
en
s
u
r
es
th
at
th
e
a
r
tific
ially
m
ad
e
d
ata
r
e
m
ain
s
clin
ically
d
ep
en
d
a
b
le
an
d
co
n
tain
s
o
n
ly
t
h
e
s
p
ec
tr
u
m
p
ar
ts
th
at
ar
e
n
ec
ess
ar
y
f
o
r
h
ea
r
t
s
o
u
n
d
an
al
y
s
is
.
3.
M
E
T
H
O
D
Fig
u
r
e
1
d
ep
icts
th
e
wo
r
k
f
lo
w
f
o
r
an
aly
zin
g
h
ea
r
t
s
o
u
n
d
s
b
y
d
ee
p
lear
n
in
g
m
eth
o
d
s
.
I
t
b
eg
in
s
with
s
p
litt
in
g
th
e
s
ig
n
als
o
f
v
en
tr
icu
lar
an
d
atr
ial
co
n
tr
ac
tio
n
in
to
two
s
ec
tio
n
s
:
P1
a
n
d
P2
.
T
h
e
n
,
th
ese
s
ig
n
als
ar
e
k
ep
t
s
ep
ar
ate
an
d
p
ass
ed
o
n
to
a
s
et
o
f
alr
ea
d
y
-
tr
ain
ed
f
ea
tu
r
e
class
if
ier
s
an
d
s
p
ec
tr
o
g
r
ap
h
s
.
T
h
e
attr
ib
u
te
-
g
en
er
ate
d
f
ea
tu
r
es
h
av
e
b
ee
n
u
s
ed
to
tr
ain
a
m
o
d
el
co
m
p
o
s
ed
o
f
m
an
y
h
id
d
en
lay
er
s
in
d
ee
p
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Dee
p
lea
r
n
in
g
fo
r
ea
r
ly
d
etec
t
io
n
o
f c
a
r
d
io
v
a
s
cu
la
r
d
is
ea
s
es v
ia
a
u
s
cu
lta
tio
n
s
o
u
n
d
…
(
S
h
r
ey
a
s
K
a
s
tu
r
e
)
1749
lear
n
in
g
an
d
to
ca
teg
o
r
ize
th
e
h
ea
r
t
s
o
u
n
d
s
in
to
lab
els
lik
e
S1
,
S2
,
ex
tr
a
s
y
s
to
le,
m
u
r
m
u
r
,
d
iast
o
lic
m
u
r
m
u
r
,
an
d
n
o
r
m
al
h
ea
r
t so
u
n
d
s
.
T
h
is
DNN
o
r
ien
ts
its
elf
o
n
th
e
in
p
u
t sets
to
m
ak
e
th
ese
class
if
ica
tio
n
s
.
Fig
u
r
e
1
.
Dee
p
lear
n
in
g
p
ip
eli
n
e
f
o
r
h
ea
r
t so
u
n
d
class
if
icatio
n
3
.
1
.
Da
t
a
a
cquis
it
io
n a
nd
pr
e
-
pro
ce
s
s
ing
PC
G
s
ig
n
als,
th
e
d
ig
ital
r
ec
o
r
d
in
g
s
o
f
h
ea
r
t
s
o
u
n
d
s
,
wer
e
co
llected
th
r
o
u
g
h
m
o
b
ile
d
ev
ices
(
L
ittma
n
n
3
,
2
0
0
elec
tr
o
n
ic
s
teth
o
s
co
p
e
at
4
,
0
0
0
Hz
s
am
p
lin
g
r
ate)
,
elec
tr
o
n
ic
s
teth
o
s
c
o
p
es,
an
d
ex
is
tin
g
d
atab
ases
(
Ph
y
s
io
Net
ch
allen
g
e
2
0
1
6
,
PASC
AL
h
ea
r
t so
u
n
d
d
atab
ase)
.
T
h
e
r
ec
o
r
d
in
g
s
wer
e
s
tan
d
ar
d
ized
to
a
d
u
r
atio
n
o
f
1
0
-
2
0
s
ec
o
n
d
s
p
e
r
s
am
p
le,
with
a
m
i
n
im
u
m
q
u
ality
th
r
esh
o
ld
o
f
1
6
-
b
it
r
eso
lu
tio
n
.
Data
q
u
ality
cr
iter
ia
in
clu
d
ed
clea
r
ca
r
d
iac
cy
cles,
m
in
im
al
a
m
b
ien
t
n
o
i
s
e,
an
d
p
r
o
p
e
r
s
en
s
o
r
p
lace
m
en
t.
Fil
tr
atio
n
was
r
ep
o
r
ted
to
ac
h
iev
e
th
e
r
em
o
v
al
o
f
b
ac
k
g
r
o
u
n
d
n
o
is
e
an
d
a
r
tifa
cts
f
r
o
m
th
e
PC
G
s
ig
n
als
[
2
0
]
.
T
h
e
l
o
w
-
lev
el
B
u
tter
wo
r
th
d
ig
ital
f
ilter
in
g
alg
o
r
ith
m
(
4
th
o
r
d
er
,
cu
t
o
f
f
f
r
e
q
u
en
cies:
2
0
-
4
0
0
Hz)
was
u
s
ed
to
elim
in
ate
th
e
h
ig
h
-
f
r
eq
u
e
n
cy
n
o
is
e
co
m
p
o
n
en
ts
in
th
e
s
y
s
tem
.
W
av
elet
d
en
o
is
in
g
tech
n
iq
u
es,
th
at
is
,
wav
elet
th
r
esh
o
ld
in
g
(
u
s
in
g
Dau
b
ec
h
ies
-
4
wav
elet,
s
o
f
t
th
r
esh
o
ld
in
g
with
u
n
iv
e
r
s
al
th
r
esh
o
ld
)
,
wer
e
also
em
b
r
ac
ed
to
ac
h
iev
e
d
en
o
is
in
g
[
2
1
]
.
Mo
r
e
in
ter
esti
n
g
ly
,
L
U
-
Net
was
u
s
ed
as
an
in
v
esti
g
atio
n
with
b
lo
ck
-
n
este
d
L
STM
m
o
d
u
les,
co
n
f
ig
u
r
ed
with
3
L
STM
lay
er
s
(
1
2
8
,
6
4
,
an
d
3
2
u
n
its
,
r
esp
ec
tiv
ely
)
.
Seg
m
en
tatio
n
w
as
d
o
n
e
to
d
etac
h
in
d
iv
id
u
al
h
ea
r
tb
ea
t
co
m
p
o
n
e
n
ts
,
lik
e
s
y
s
to
lic
an
d
d
iast
o
lic
p
ar
ts
,
em
p
lo
y
in
g
en
v
elo
p
e
d
etec
tio
n
ap
p
r
o
ac
h
es
s
u
ch
as
Sh
an
n
o
n
en
er
g
y
e
n
v
elo
p
e
esti
m
atio
n
(
win
d
o
w
s
ize:
0
.
0
2
s
,
o
v
e
r
lap
:
5
0
%)
an
d
d
ee
p
lear
n
in
g
f
r
am
ewo
r
k
s
[
2
2
]
.
PC
G
ch
ar
ac
ter
is
tic
co
m
p
o
n
en
ts
wer
e
id
en
tifie
d
f
r
o
m
th
e
s
ig
n
als o
f
s
eg
m
en
ted
an
d
d
en
o
is
ed
PC
G
[
2
3
]
.
MFC
C
s
wer
e
co
m
p
u
ted
u
s
in
g
m
el
-
s
ca
le
f
ilter
b
an
k
s
(
4
0
f
ilter
s
)
an
d
d
is
cr
e
te
co
s
in
e
tr
an
s
f
o
r
m
m
eth
o
d
s
.
W
e
d
er
iv
ed
tim
e
-
f
r
eq
u
e
n
cy
r
ep
r
esen
tatio
n
s
,
n
am
ely
s
ca
lo
g
r
am
s
,
f
r
o
m
co
n
tin
u
o
u
s
wav
elet
tr
an
s
f
o
r
m
(
C
W
T
)
alg
o
r
ith
m
s
(
Mo
r
let
wav
elet,
s
ca
les
1
-
1
2
8
)
[
2
4
]
.
Featu
r
e
ex
tr
ac
tio
n
o
f
a
u
to
m
atic
ty
p
e
was
also
d
em
o
n
s
tr
ated
b
y
p
e
r
f
o
r
m
in
g
th
e
C
NN
alg
o
r
ith
m
s
d
ir
ec
tly
o
n
th
e
PC
G
s
ig
n
als o
r
s
p
ec
tr
o
g
r
am
s
.
3
.
2
.
Da
t
a
a
ug
m
ent
a
t
io
n
Data
au
g
m
e
n
tatio
n
tec
h
n
iq
u
es
wer
e
u
s
ed
to
in
cr
ea
s
e
th
e
v
ar
iatio
n
lev
el
an
d
th
e
d
ata
in
t
h
e
tr
ain
in
g
s
et.
T
im
e
s
tr
etch
in
g
(
±
1
0
% sp
ee
d
m
o
d
if
icatio
n
)
an
d
p
itch
ad
ju
s
tm
en
t a
lg
o
r
ith
m
s
(
±
2
s
em
ito
n
es)
wer
e
u
tili
ze
d
to
ch
an
g
e
th
e
au
d
i
o
s
ig
n
als'
p
ac
e
an
d
f
r
eq
u
en
cy
[
2
5
]
.
Dif
f
er
en
t
n
o
is
es
(
Gau
s
s
ian
,
p
in
k
,
an
d
en
v
ir
o
n
m
en
tal
n
o
is
es
at
s
ig
n
al
-
to
-
n
o
is
e
r
atio
(
SNR
)
lev
els
o
f
5
-
1
5
d
B
)
wer
e
ad
d
ed
to
th
e
o
r
ig
in
al
s
ig
n
al
u
s
in
g
n
o
is
e
in
jectio
n
alg
o
r
ith
m
s
.
B
esid
es
th
e
e
x
am
p
les
m
en
tio
n
ed
ab
o
v
e,
Sp
ec
Au
g
m
en
t
was
also
o
n
e
o
f
th
e
m
eth
o
d
s
ex
p
lo
r
ed
.
T
im
e
war
p
i
n
g
(
W
=8
0
)
,
tim
e
m
ask
in
g
(
T
=
4
0
)
,
an
d
f
r
eq
u
e
n
cy
m
ask
in
g
(
F
=3
0
)
alg
o
r
ith
m
s
wer
e
ap
p
lied
to
th
e
s
p
ec
tr
o
g
r
a
m
s
.
3
.
3
.
Dee
p
lea
rning
mo
del dev
elo
pm
ent
Deep
-
lear
n
in
g
n
etwo
r
k
s
wer
e
estab
lis
h
ed
to
d
etec
t
a
n
o
m
a
lies
in
h
ea
r
t
s
o
u
n
d
s
.
C
NN
a
r
e
p
r
ec
is
e
ar
ch
itectu
r
es
f
o
r
tim
e
-
f
r
e
q
u
e
n
cy
ca
r
d
iac
s
o
u
n
d
r
e
p
r
esen
t
atio
n
s
[
2
6
]
.
Su
cc
ess
f
u
l
C
NN
ar
ch
itectu
r
es
ar
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
2
,
Ap
r
il 2
0
2
6
:
1
7
4
6
-
1
7
6
1
1750
co
m
p
r
is
ed
o
f
R
esNet
(
1
8
lay
er
s
with
r
esid
u
al
c
o
n
n
ec
tio
n
s
)
,
Mo
b
ileNetV2
(
1
.
4
ex
p
a
n
s
io
n
f
ac
to
r
)
,
a
n
d
C
NN
-
DAG.
I
n
p
u
t
d
im
e
n
s
io
n
s
wer
e
s
tan
d
ar
d
ized
t
o
2
2
4
×
2
2
4
p
ix
els
f
o
r
s
p
ec
tr
o
g
r
am
s
with
b
atch
n
o
r
m
aliza
tio
n
af
ter
ea
ch
co
n
v
o
lu
tio
n
al
lay
er
.
R
NNs,
in
clu
d
in
g
L
STM
(
3
lay
er
s
,
2
5
6
u
n
i
ts
ea
ch
)
an
d
g
ated
r
ec
u
r
r
en
t
u
n
it
(
GR
U
)
(
2
lay
er
s
,
1
2
8
u
n
its
ea
ch
)
wer
e
th
e
m
o
d
els
wid
ely
u
s
ed
to
tac
k
le
s
eq
u
en
tial
d
at
a
m
o
d
ellin
g
a
n
d
p
r
eser
v
in
g
te
m
p
o
r
al
d
ep
en
d
en
cies
in
h
ea
r
t
s
o
u
n
d
s
ig
n
als
[
2
7
]
.
Du
al
ty
p
es
o
f
C
NN
-
L
STM
m
o
d
els
,
s
u
ch
as
C
NN
-
b
id
ir
ec
t
io
n
al
g
ated
r
ec
u
r
r
en
t
u
n
it
(
B
i
GR
U)
,
wer
e
co
n
s
tr
u
cted
to
e
n
s
u
r
e
th
e
u
s
e
o
f
t
h
e
s
p
atial
f
ea
tu
r
e
ex
tr
ac
tio
n
p
o
w
er
o
f
C
NNs a
n
d
th
e
tem
p
o
r
al
m
o
d
elin
g
ca
p
ac
ity
o
f
L
STM
s
.
3
.
4
.
M
o
del
t
ra
ini
ng
a
nd
o
pti
m
iza
t
io
n
W
e
s
p
lit
th
e
d
ataset
in
to
th
e
v
alid
atio
n
(
2
0
%),
tr
ain
in
g
(
6
0
%),
an
d
test
in
g
s
ets
(
2
0
%).
T
h
e
d
ee
p
lear
n
in
g
m
o
d
els
wer
e
tr
ain
e
d
u
s
in
g
au
g
m
en
te
d
tr
ain
in
g
d
ata
b
o
o
s
ted
with
p
ar
am
o
u
n
t
h
y
p
er
p
ar
am
eter
s
.
T
r
ain
in
g
was
co
n
d
u
cted
u
s
in
g
Ad
am
o
p
tim
izer
(
in
itial
lear
n
in
g
r
ate:
0
.
0
0
1
,
β1
=0
.
9
,
β2
=0
.
9
9
9
)
with
a
b
atch
s
ize
o
f
3
2
an
d
ea
r
l
y
s
to
p
p
i
n
g
(
p
atien
ce
=1
0
e
p
o
ch
s
)
.
T
ec
h
n
i
q
u
es
s
u
ch
as
g
r
id
s
ea
r
ch
,
B
ay
esian
o
p
tim
izatio
n
,
an
d
ev
o
lu
tio
n
ar
y
alg
o
r
ith
m
s
lik
e
h
o
n
ey
b
a
d
g
er
o
p
tim
izatio
n
(
HB
O)
wer
e
u
tili
ze
d
f
o
r
th
e
h
y
p
er
p
ar
am
ete
r
tu
n
in
g
[
2
8
]
.
T
h
e
h
y
p
e
r
p
ar
am
eter
r
an
g
es
in
clu
d
e
d
lear
n
in
g
r
ates
(
1
0
-
5
to
1
0
-
2
)
,
d
r
o
p
o
u
t
r
ates
(
0
.
1
-
0
.
5
)
,
a
n
d
n
etwo
r
k
wid
th
(
3
2
-
2
5
6
u
n
its
)
.
Atten
tio
n
m
ec
h
an
is
m
s
(
s
elf
-
a
tten
tio
n
with
8
h
ea
d
s
)
,
r
esid
u
a
l
co
n
n
ec
tio
n
s
,
an
d
en
s
em
b
le
m
eth
o
d
s
(
b
ag
g
in
g
with
5
m
o
d
els)
wer
e
th
en
e
m
b
ed
d
e
d
in
th
e
m
o
d
el
t
o
en
s
u
r
e
ef
f
icien
cy
an
d
en
h
an
ce
d
in
ter
p
r
eta
b
ilit
y
.
3
.
5
.
E
v
a
lua
t
i
o
n
a
nd
deplo
y
m
ent
W
e
f
ir
s
t
ev
alu
ate
th
e
tr
ain
ed
m
o
d
els
b
y
co
m
p
a
r
in
g
t
h
eir
a
cc
u
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
F1
-
s
co
r
e,
an
d
ar
ea
u
n
d
er
th
e
r
ec
eiv
er
o
p
e
r
atin
g
ch
ar
ac
te
r
is
tic
cu
r
v
e
(
AUR
OC
)
o
n
th
e
test
d
ataset.
Sin
ce
th
e
s
tu
d
y
f
o
llo
wed
a
r
ep
ea
ted
m
ea
s
u
r
es
d
esig
n
,
p
air
ed
t
-
test
s
(
p
<0
.
0
5
)
wer
e
u
s
ed
to
ass
ess
th
e
s
tatis
tical
s
ig
n
if
ican
ce
,
wh
ile
co
n
f
id
en
ce
in
ter
v
als we
r
e
esti
m
ated
th
r
o
u
g
h
b
o
o
ts
tr
ap
r
esam
p
lin
g
with
1
,
0
0
0
iter
atio
n
s
.
C
o
m
p
ar
is
o
n
s
o
f
d
ee
p
lear
n
in
g
ar
ch
itectu
r
es
an
d
f
e
atu
r
e
ex
tr
ac
tio
n
m
eth
o
d
s
wer
e
also
m
ad
e.
T
h
e
p
er
f
o
r
m
an
ce
o
f
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
was
esti
m
ated
u
s
in
g
cr
o
s
s
v
alid
atio
n
o
f
th
e
5
-
f
o
ld
ty
p
e.
Acc
o
r
d
in
g
to
th
e
f
in
d
in
g
s
o
f
th
is
s
tu
d
y
,
th
e
b
est p
er
f
o
r
m
an
ce
o
f
th
e
m
o
d
el
was a
ch
iev
ed
wh
en
r
u
n
n
in
g
it
o
n
m
o
b
ile
d
ev
ices (
in
f
er
e
n
ce
tim
e
<1
s
)
o
r
clo
u
d
p
latf
o
r
m
s
s
u
ch
as
AW
S
E
C
2
t2
.
lar
g
e
in
s
tan
ce
s
to
ad
d
r
ess
h
ea
r
t
s
o
u
n
d
s
cr
ee
n
i
n
g
an
d
C
VD
s
d
etec
tio
n
in
r
ea
l
-
tim
e
[
2
9
]
.
T
h
er
e
we
r
e
e
asy
-
to
-
r
ea
d
o
u
tp
u
t
i
n
ter
f
ac
es
an
d
v
is
u
al
r
ep
r
esen
tatio
n
s
m
ad
e
f
o
r
b
o
th
t
h
e
h
ea
lth
ca
r
e
p
r
o
f
ess
io
n
al
an
d
t
h
e
p
er
s
o
n
,
an
d
th
ey
d
is
p
lay
e
d
th
e
f
in
al
o
u
tco
m
es.
T
h
e
d
ep
lo
y
m
en
t
p
ip
elin
e
in
v
o
lv
ed
m
o
d
el
q
u
an
tizatio
n
f
o
r
m
o
b
ile
d
ev
ices
(
8
-
b
it)
an
d
r
ep
r
esen
tatio
n
al
s
tate
t
r
an
s
f
er
ap
p
licatio
n
p
r
o
g
r
a
m
m
in
g
in
ter
f
ac
e
(
R
E
ST
API
s
)
f
o
r
clo
u
d
d
e
p
lo
y
m
e
n
t.
3
.
6
.
Sig
na
l
pro
ce
s
s
ing
pip
eli
ne
I
n
o
u
r
s
ig
n
al
p
r
o
ce
s
s
in
g
p
ip
el
in
e,
th
er
e
wer
e
s
ev
er
al
s
tep
s
t
o
im
p
r
o
v
e
th
e
q
u
ality
o
f
th
e
d
ata
g
o
in
g
in
to
an
aly
s
is
.
I
n
itial
s
ig
n
al
p
r
o
ce
s
s
in
g
in
v
o
lv
ed
h
ig
h
-
p
ass
f
ilter
in
g
with
a
cu
to
f
f
f
r
eq
u
en
cy
o
f
2
0
Hz
t
o
ex
clu
d
e
lo
w
-
f
r
e
q
u
en
c
y
n
o
is
e,
af
ter
wh
ich
th
e
s
ig
n
al
was
s
h
if
ted
to
h
av
e
a
m
ea
n
o
f
ze
r
o
.
T
h
e
d
ata
was
th
en
n
o
r
m
alize
d
th
r
o
u
g
h
am
p
litu
d
e
n
o
r
m
aliza
tio
n
b
y
s
ca
lin
g
ea
ch
r
ec
o
r
d
in
g
to
f
all
b
etwe
en
-
1
a
n
d
+1
.
T
o
im
p
r
o
v
e
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
s
ig
n
al
f
ilter
in
g
,
th
e
ad
a
p
tiv
e
f
ilter
in
g
was
d
o
n
e
as
f
o
llo
ws:
least
m
ea
n
s
q
u
ar
es
(
LMS
)
alg
o
r
ith
m
,
th
e
v
alu
es
o
f
s
tep
s
ize
u
s
ed
wer
e
0
.
0
1
,
th
e
n
u
m
b
er
o
f
tap
s
u
s
ed
was
3
2
,
an
d
f
in
ally
,
th
e
co
n
v
er
g
en
ce
th
r
esh
o
ld
u
s
ed
w
as
1
0
-
6
t
h
u
s
,
th
is
h
elp
ed
to
m
i
n
im
ize
th
e
lev
el
o
f
n
o
is
e
as
well
as
m
ain
tain
in
g
th
e
p
u
r
ity
o
f
th
e
o
r
ig
in
al
s
ig
n
a
l.
3
.
7
.
E
t
hica
l
co
ns
idera
t
io
ns
a
nd
da
t
a
g
o
v
er
na
nce
Var
io
u
s
eth
ical
s
tan
d
ar
d
n
o
r
m
s
,
an
d
d
ata
g
o
v
er
n
a
n
ce
p
o
li
cies
wer
e
f
o
llo
wed
in
th
e
p
r
o
ce
s
s
o
f
th
e
s
tu
d
y
.
Priv
ac
y
m
ea
s
u
r
es
we
r
e
in
itiated
th
r
o
u
g
h
d
ata
an
o
n
y
m
izatio
n
p
r
o
ce
d
u
r
es,
ac
ce
s
s
co
n
tr
o
l
m
ea
s
u
r
es,
au
d
it
tr
ailin
g
p
r
o
ce
d
u
r
es,
en
c
r
y
p
tio
n
r
eq
u
ir
em
en
ts
,
a
n
d
d
o
c
u
m
en
ted
d
ata
r
ete
n
tio
n
p
o
licies.
T
o
en
s
u
r
e
eth
ical
p
r
ac
tice
,
th
e
f
o
llo
win
g
s
af
eg
u
ar
d
s
wer
e
u
s
ed
;
g
ettin
g
ap
p
r
o
v
al
f
r
o
m
th
e
in
s
titu
tio
n
al
r
ev
iew
b
o
ar
d
(
I
R
B
)
,
in
f
o
r
m
in
g
p
ar
ticip
an
ts
ab
o
u
t
th
e
s
tu
d
y
,
s
ig
n
in
g
d
ata
u
s
ag
e
ag
r
ee
m
en
ts
,
d
o
in
g
p
r
iv
ac
y
i
m
p
ac
t
ass
es
s
m
en
ts
,
an
d
h
av
in
g
th
e
s
tu
d
y
g
o
t
h
r
o
u
g
h
eth
ics
co
m
m
ittees.
Su
ch
m
ea
s
u
r
es
allo
wed
m
in
im
izin
g
u
n
eth
ical
b
e
h
av
io
r
d
u
r
in
g
th
e
r
esear
ch
p
r
o
ce
d
u
r
e
an
d
co
n
s
id
er
in
g
th
e
co
n
f
i
d
en
ti
ality
o
f
p
ar
ticip
a
n
ts
an
d
th
eir
i
n
f
o
r
m
atio
n
.
4.
P
RO
P
O
SE
D
M
O
D
E
L
4
.
1
.
Da
t
a
pre
-
pro
ce
s
s
ing
B
o
th
h
ig
h
-
an
d
lo
w
-
p
ass
f
ilter
s
ar
e
u
s
ed
f
o
r
n
o
is
e
co
m
p
o
n
en
t
r
em
o
v
al
f
r
o
m
th
e
h
ea
r
tb
e
at
s
ig
n
als.
T
h
e
f
ir
s
t
is
f
o
r
r
em
o
v
in
g
th
e
h
ig
h
est
-
f
r
eq
u
e
n
cy
n
o
is
e,
an
d
th
e
s
ec
o
n
d
is
f
o
r
r
em
o
v
in
g
lo
wer
-
f
r
e
q
u
en
c
y
co
m
p
o
n
en
ts
.
T
h
e
s
am
p
lin
g
r
ate
f
o
r
au
d
io
r
ec
o
r
d
in
g
s
was
u
s
ed
to
d
ef
in
e
Ny
q
u
is
t
f
r
eq
u
en
cy
.
C
alcu
lated
ac
co
r
d
in
g
to
Ny
q
u
is
t's
th
eo
r
y
,
it
h
as
a
cu
to
f
f
f
r
eq
u
en
cy
o
f
n
o
r
m
al.
A
d
ig
ital
B
u
tter
wo
r
t
h
f
ilter
was
d
esig
n
e
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Dee
p
lea
r
n
in
g
fo
r
ea
r
ly
d
etec
t
io
n
o
f c
a
r
d
io
v
a
s
cu
la
r
d
is
ea
s
es v
ia
a
u
s
cu
lta
tio
n
s
o
u
n
d
…
(
S
h
r
ey
a
s
K
a
s
tu
r
e
)
1751
an
d
im
p
lem
en
ted
u
s
in
g
s
u
itab
le
f
ilter
d
esig
n
tech
n
iq
u
es.
T
h
e
B
u
tter
wo
r
th
f
ilter
is
an
in
f
i
n
ite
im
p
u
ls
e
r
esp
o
n
s
e
(
I
I
R
)
f
ilter
,
co
m
m
o
n
ly
u
s
ed
f
o
r
its
co
n
tin
u
o
u
s
ly
s
m
o
o
th
a
n
d
f
lat
f
r
e
q
u
en
cy
r
esp
o
n
s
e
in
t
h
e
p
ass
b
an
d
.
T
h
e
an
alo
g
f
ilter
p
ar
am
eter
s
ar
e
tu
n
ed
to
ef
f
icien
tly
r
em
o
v
e
th
e
h
ig
h
f
r
eq
u
e
n
cies
th
at
co
n
tain
th
e
m
o
s
t
n
o
is
e
wh
ile
p
r
eser
v
i
n
g
t
h
e
h
ea
r
t
s
o
u
n
d
c
o
m
p
o
n
en
ts
wi
th
lo
w
f
r
e
q
u
en
cies
th
at
ca
r
r
y
m
o
s
t
o
f
th
e
v
alu
ab
le
in
f
o
r
m
atio
n
.
T
h
e
f
ilter
d
esig
n
p
r
o
ce
s
s
ca
lcu
lated
th
e
n
u
m
er
ato
r
an
d
d
en
o
m
in
ato
r
p
o
l
y
n
o
m
ial
co
ef
f
icien
ts
d
ef
in
in
g
th
e
B
u
tter
wo
r
th
f
ilter
'
s
tr
an
s
f
er
f
u
n
ctio
n
.
T
h
e
h
ea
r
t
s
o
u
n
d
s
ig
n
als
tr
ea
ted
b
y
wav
e
let
d
en
o
is
in
g
wer
e
tu
r
n
ed
m
o
r
e
ac
cu
r
ate
an
d
s
m
o
o
th
er
b
y
r
em
o
v
i
n
g
th
e
n
o
is
e
.
T
h
e
r
eg
is
tr
atio
n
was
ca
r
r
ied
o
u
t
b
y
s
p
ec
ialized
s
y
s
tem
s
p
r
o
ce
s
s
in
g
th
e
s
ig
n
als
in
to
wav
elet
co
ef
f
icien
ts
b
a
s
ed
o
n
th
e
d
is
cr
ete
wav
elet
tr
an
s
f
o
r
m
alg
o
r
ith
m
s
.
I
n
th
e
en
s
u
in
g
s
tag
e,
s
o
f
t
o
r
h
ar
d
th
r
esh
o
ld
in
g
tec
h
n
iq
u
es
wer
e
em
p
lo
y
e
d
wh
ile
d
ea
lin
g
with
th
e
wav
elet
co
ef
f
icien
ts
.
T
h
e
s
ig
n
als
wer
e
r
ec
o
n
s
tr
u
cted
f
r
o
m
t
h
r
esh
o
ld
wav
elet
co
ef
f
icien
ts
in
th
e
d
e
co
m
p
o
s
itio
n
s
tag
e,
wh
ich
f
in
ally
r
ed
u
ce
d
th
e
u
n
d
esire
d
co
m
p
o
n
e
n
t
n
o
is
e
wh
ile
s
ig
n
al
f
ea
tu
r
es
wer
e
p
r
eser
v
ed
.
T
h
e
th
r
esh
o
ld
was
ap
p
r
o
ac
h
e
d
u
s
in
g
b
o
th
s
o
f
t
an
d
co
m
p
le
x
tech
n
iq
u
es.
So
f
t
th
r
esh
o
ld
i
n
g
im
p
lied
a
s
h
r
i
n
k
in
g
v
alu
e
f
o
r
th
e
wav
elet
co
ef
f
icien
ts
clo
s
e
to
ze
r
o
,
wh
ile
h
ar
d
t
h
r
esh
o
ld
in
g
d
r
o
p
p
e
d
th
e
co
ef
f
icien
ts
b
elo
w
th
e
th
r
esh
o
ld
lev
el
to
ze
r
o
e
n
tire
ly
.
T
o
tr
ain
th
e
L
U
-
Net
m
o
d
el,
we
d
esig
n
ed
a
d
ata
s
et
co
n
s
is
tin
g
o
f
n
o
is
y
an
d
clea
n
h
ea
r
t
s
o
u
n
d
s
ig
n
als,
s
er
v
in
g
as
tr
ain
in
g
d
ata.
T
h
e
n
o
is
y
s
ig
n
als
wer
e
s
im
u
lated
b
y
ar
tific
ially
ad
d
in
g
b
r
ea
th
in
g
s
o
u
n
d
s
,
am
b
ien
t h
o
s
p
ital n
o
is
es,
an
d
wh
ite
Gau
s
s
ian
n
o
is
e
with
th
e
clea
n
h
ea
r
t so
u
n
d
r
ec
o
r
d
in
g
s
as th
e
b
ac
k
g
r
o
u
n
d
.
T
h
e
tr
ain
in
g
p
r
o
ce
s
s
in
clu
d
ed
th
e
C
NN
s
'
m
in
im
izatio
n
o
f
th
e
m
is
m
atch
b
etwe
en
th
e
d
en
o
is
ed
o
u
tp
u
t
s
ig
n
als
p
r
o
v
id
e
d
b
y
th
e
L
U
-
N
et
m
o
d
el
a
n
d
th
e
clea
n
tar
g
et
s
ig
n
als.
T
o
th
is
en
d
,
we
u
s
ed
th
e
m
ea
n
s
q
u
ar
ed
er
r
o
r
lo
s
s
an
d
s
elec
ted
p
er
ce
p
tu
al
ev
alu
atio
n
m
etr
ics
to
d
r
iv
e
th
e
m
o
d
el
tr
ain
in
g
a
n
d
s
ec
u
r
e
th
e
ac
cu
r
ate
r
ec
o
n
s
tr
u
ctio
n
o
f
th
e
s
ig
n
al
p
ar
am
eter
s
.
Fig
u
r
e
2
is
th
e
g
r
a
p
h
o
f
th
e
s
ig
n
al
am
p
litu
d
e
f
o
r
th
e
o
r
ig
in
al
h
ea
r
t
s
o
u
n
d
a
n
d
th
e
wav
elet,
L
U
-
N
et,
an
d
lo
w
-
p
ass
f
ilter
ed
at
a
r
eg
u
lar
s
p
ee
d
o
f
0
.
1
s
ec
o
n
d
s
an
d
is
p
an
n
ed
o
v
e
r
2
s
ec
o
n
d
s
.
T
h
e
x
-
ax
is
is
tim
e
(
s
)
,
wh
ich
g
i
v
es
th
e
tim
e
p
e
r
io
d
o
f
2
s
ec
o
n
d
s
o
v
er
wh
ich
t
h
e
h
ea
r
t
s
o
u
n
d
s
ig
n
al
is
tak
en
.
T
h
is
ax
is
s
h
o
ws
t
h
e
r
eg
u
lar
in
te
r
v
als
o
f
0
.
1
s
ec
o
n
d
s
.
T
h
e
y
-
ax
is
in
th
e
g
iv
en
d
iag
r
am
s
illu
s
tr
ates
s
ig
n
al
am
p
litu
d
e
.
T
h
is
r
ep
r
esen
ts
s
ig
n
al
am
p
litu
d
e
o
r
s
tr
en
g
th
in
r
elatio
n
to
t
h
e
h
ea
r
t
s
o
u
n
d
s
in
a
p
ar
ticu
la
r
tim
e
f
r
am
e.
T
h
e
y
-
ax
is
s
ca
le
g
o
es
f
r
o
m
-
0
.
0
4
to
0
.
0
4
,
s
h
o
win
g
th
e
am
p
litu
d
e
f
o
r
d
if
f
e
r
en
t
s
ig
n
al
p
r
o
ce
s
s
in
g
tech
n
iq
u
es
u
n
d
er
c
o
m
p
ar
is
o
n
.
T
h
e
g
r
a
p
h
r
ep
r
esen
t
s
th
e
r
aw
o
r
in
itial
h
ea
r
t
s
o
u
n
d
s
ig
n
al
a
n
d
th
en
th
e
s
ig
n
als
af
ter
ap
p
ly
i
n
g
wav
elet
d
en
o
is
e,
L
U
-
N
et
d
en
o
is
e
,
a
n
d
th
e
n
a
lo
w
-
p
ass
f
ilter
.
I
t
also
allo
ws
v
er
if
y
in
g
h
o
w
ea
c
h
d
en
o
is
in
g
m
eth
o
d
im
p
ac
ts
th
e
g
en
e
r
al
f
ea
tu
r
es
o
f
th
e
s
ig
n
al
am
p
litu
d
es
o
v
e
r
tim
e,
m
ak
in
g
it
p
o
s
s
ib
le
t
o
ass
es
s
th
e
ch
an
g
es in
th
e
c
h
ar
ac
ter
is
tics
o
f
th
e
h
ea
r
t so
u
n
d
wav
ef
o
r
m
.
Fig
u
r
e
2
.
Sig
n
al
a
m
p
litu
d
e
v
alu
es f
o
r
th
e
r
aw
h
ea
r
ts
o
u
n
d
s
ig
n
al
an
d
th
e
d
en
o
is
ed
s
ig
n
als
T
h
e
tr
ain
in
g
p
r
o
ce
s
s
r
ep
r
esen
ts
th
e
m
o
s
t
s
ig
n
if
ican
t
co
m
p
o
n
en
t
o
f
o
u
r
ap
p
r
o
ac
h
,
wh
ic
h
in
v
o
lv
es
m
in
im
izin
g
th
e
d
if
f
er
e
n
ce
b
et
wee
n
th
e
m
o
d
el'
s
p
r
ed
ictio
n
s
an
d
th
e
g
r
o
u
n
d
tr
u
th
an
n
o
tatio
n
s
.
T
h
e
C
NN
-
b
ased
s
eg
m
en
tatio
n
m
o
d
el
m
eth
o
d
u
s
ed
ar
ch
itectu
r
es
th
at
co
u
ld
ac
cu
r
ately
d
r
aw
th
e
h
ea
r
t
s
o
u
n
d
s
'
lo
ca
l
tem
p
o
r
al
an
d
s
p
ec
tr
al
f
ea
tu
r
es.
Fig
u
r
e
3
d
em
o
n
s
tr
ates
th
e
en
v
elo
p
e
o
f
th
e
en
er
g
y
o
f
th
e
d
e
n
o
is
ed
h
ea
r
t
s
o
u
n
d
s
ig
n
al
v
alu
es
at
s
et
in
ter
v
als.
Dep
ict
ed
o
v
e
r
tim
e,
o
n
th
e
x
-
a
x
is
,
a
0
-
3
s
ec
onds
o
f
th
e
h
ea
r
t
s
o
u
n
d
s
ig
n
al
h
as
b
ee
n
u
s
ed
at
in
ter
v
als
o
f
0
.
1
s
ec
o
n
d
s
.
On
th
e
y
-
ax
is
,
we
ca
n
s
ee
th
e
en
er
g
y
e
n
v
elo
p
e
o
f
th
e
d
en
o
is
ed
h
ea
r
t
s
o
u
n
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
2
,
Ap
r
il 2
0
2
6
:
1
7
4
6
-
1
7
6
1
1752
s
ig
n
al.
T
h
is
r
ep
r
esen
ts
th
e
o
v
er
all
en
er
g
y
o
r
am
p
litu
d
e
o
f
t
h
e
h
ea
r
t
s
o
u
n
d
wav
ef
o
r
m
o
v
e
r
tim
e.
B
y
d
r
awin
g
th
e
en
er
g
y
en
v
el
o
p
e,
th
is
g
r
a
p
h
f
ac
ilit
ates
v
is
u
aliza
tio
n
o
f
v
ar
iatio
n
s
in
th
e
en
er
g
y
d
is
tr
ib
u
tio
n
an
d
in
ten
s
ity
o
f
th
e
d
en
o
is
ed
h
ea
r
t
s
o
u
n
d
s
ig
n
al
d
u
r
in
g
th
e
an
aly
ze
d
ti
m
e
p
er
io
d
.
T
h
is
ca
n
h
elp
o
b
t
ain
in
f
o
r
m
atio
n
o
r
ch
ar
ac
ter
is
tic
s
ab
o
u
t th
e
p
r
o
ce
s
s
ed
h
ea
r
t so
u
n
d
d
ata
,
as we
ll a
s
th
eir
b
eh
av
io
r
.
Fig
u
r
e
3
.
T
h
e
en
er
g
y
e
n
v
elo
p
e
o
f
th
e
d
en
o
is
ed
h
ea
r
t so
u
n
d
s
ig
n
al
4
.
2
.
Da
t
a
a
ug
m
ent
a
t
io
n
A
s
ca
lin
g
tim
e
alg
o
r
ith
m
was
u
s
ed
in
t
h
e
d
ata
co
n
ce
r
n
in
g
h
u
m
an
h
ea
r
t
ac
tiv
ity
,
wh
ich
m
ea
n
s
th
at
th
e
s
o
u
n
d
tem
p
o
o
r
d
u
r
atio
n
was
ad
ju
s
t
ed
s
o
th
at
p
itch
an
d
f
r
eq
u
e
n
cy
co
m
p
o
n
en
ts
r
em
ain
ed
u
n
af
f
ec
te
d
.
I
n
ad
d
itio
n
to
p
r
o
lo
n
g
ati
o
n
,
t
h
e
d
ig
ital
au
d
io
wo
r
k
s
tatio
n
(
DAW)
tech
n
iq
u
e
was
em
p
lo
y
ed
t
o
m
o
d
if
y
th
e
p
itch
o
f
th
e
h
ea
r
t
s
o
u
n
d
s
ig
n
als.
T
h
e
s
tr
ateg
ies
d
ev
elo
p
ed
e
m
p
lo
y
ed
a
m
eth
o
d
o
f
f
r
e
q
u
en
c
y
r
es
am
p
lin
g
,
wh
e
r
e
th
e
s
am
p
lin
g
r
ate
was
tr
an
s
f
o
r
m
e
d
,
r
esu
ltin
g
in
a
c
h
an
g
e
t
o
th
e
au
d
ib
le
to
n
e
o
r
f
r
eq
u
e
n
cy
co
n
ten
t.
W
e
d
ev
elo
p
ed
s
am
p
les
th
at
u
tili
s
e
d
if
f
er
en
t
f
r
eq
u
e
n
cy
s
p
ec
tr
u
m
r
a
n
g
es
,
e
n
ab
l
in
g
u
s
to
p
o
r
tr
a
y
th
e
in
t
er
f
er
en
ce
with
th
e
r
h
y
th
m
o
f
v
ar
io
u
s
f
ac
to
r
s
,
in
clu
d
in
g
a
g
e,
g
e
n
d
er
,
an
d
h
ea
r
t
d
is
ea
s
e.
W
e
co
m
b
in
ed
t
h
e
m
eth
o
d
s
o
f
tim
e
s
tr
etch
in
g
an
d
p
itch
s
h
if
tin
g
t
o
u
s
e.
T
h
e
in
jectio
n
s
co
n
s
is
tin
g
o
f
n
o
is
e
wer
e
v
a
r
ied
an
d
d
i
v
er
s
e
f
o
r
au
g
m
en
te
d
s
am
p
les.
On
e
o
f
th
e
m
o
s
t
co
m
m
o
n
ly
u
s
ed
p
ar
am
eter
s
in
co
m
m
u
n
icatio
n
s
y
s
tem
s
im
u
latio
n
s
is
th
e
SNR
.
T
h
is
v
alu
e
r
ep
r
esen
ts
th
e
r
atio
o
f
s
ig
n
al
p
o
wer
an
d
n
o
is
e
en
er
g
y
.
T
h
er
ef
o
r
e
,
we
c
an
ex
p
er
im
e
n
t
with
v
ar
io
u
s
co
m
m
u
n
icati
o
n
s
y
s
tem
s
b
y
s
elec
tin
g
d
if
f
er
en
t
n
o
is
e
co
n
d
itio
n
s
an
d
s
ig
n
al
d
eg
r
a
d
a
tio
n
lev
el
s
.
Als
o
,
s
y
n
th
etic
s
am
p
les
wer
e
o
b
tain
ed
in
v
ar
io
u
s
SN
R
v
alu
es
,
with
wh
ich
th
e
h
i
g
h
SNR
(
lo
w
n
o
is
e)
lev
el
was
co
m
p
ar
ed
to
t
h
e
lo
w
SNR
(
h
ig
h
n
o
is
e)
o
n
e
.
T
h
is
p
o
licy
c
o
u
ld
ad
ap
t
th
e
m
o
d
el
to
d
ea
l
with
th
e
in
p
u
t
n
o
is
es
o
f
d
if
f
er
e
n
t
in
ten
s
ities
d
u
r
in
g
tr
ain
in
g
,
wh
ich
,
in
tu
r
n
,
is
r
eg
ar
d
e
d
as o
n
e
o
f
th
e
r
ec
ip
es
f
o
r
o
v
er
co
m
i
n
g
g
e
n
er
aliza
tio
n
lim
itatio
n
s
an
d
g
ettin
g
b
etter
r
esu
lts
.
W
e
o
b
tain
ed
s
p
ec
tr
o
g
r
am
s
f
r
o
m
th
e
h
ea
r
t
s
o
u
n
d
s
ig
n
als
u
s
in
g
th
e
s
h
o
r
t
-
tim
e
Fo
u
r
ier
tr
an
s
f
o
r
m
(
STFT
)
an
d
th
e
C
W
T
.
W
e
p
er
f
o
r
m
e
d
th
e
tim
e
war
p
in
g
o
n
th
e
g
en
er
ate
d
s
p
ec
tr
o
g
r
am
s
in
th
e
tim
e
d
o
m
ain
,
wh
ich
in
clu
d
es
s
tr
etch
in
g
an
d
s
h
r
in
k
in
g
th
e
tim
e
d
i
r
ec
tio
n
wh
ile
k
ee
p
i
n
g
t
h
e
f
r
eq
u
e
n
cy
co
n
ten
t
u
n
alter
ed
.
B
y
b
en
d
in
g
th
e
tim
e
d
im
en
s
io
n
,
we
p
r
o
d
u
ce
d
au
g
m
en
ted
s
p
ec
tr
o
g
r
am
s
am
p
les
with
d
if
f
er
e
n
t
tem
p
o
r
al
s
eq
u
en
ce
s
,
s
im
u
latin
g
v
a
r
io
u
s
h
ea
r
tb
ea
ts
o
r
p
u
ls
es.
T
h
is
p
r
o
v
id
ed
th
e
tem
p
o
r
al
v
ar
iety
in
t
h
e
h
ea
r
t
s
o
u
n
d
d
ata
th
at
im
p
r
o
v
e
d
th
e
d
i
v
er
s
ity
o
f
th
e
tr
ain
in
g
d
ataset.
I
n
ad
d
itio
n
,
th
e
d
ata
was
en
h
a
n
ce
d
b
y
ti
m
e
m
ask
in
g
o
n
th
e
s
p
ec
tr
o
g
r
am
s
.
Fig
u
r
e
4
s
h
o
ws
th
e
d
en
o
is
in
g
p
er
f
o
r
m
an
ce
o
f
h
ea
r
t so
u
n
d
s
ig
n
als
.
C
o
n
s
eq
u
en
tl
y
,
s
o
m
e
te
m
p
o
r
al
i
n
f
o
r
m
ati
o
n
was
m
ask
e
d
b
y
ze
r
o
i
n
g
o
u
t
s
p
ec
if
i
c
ti
m
e
s
t
ep
s
i
n
th
e
s
p
e
ct
r
o
g
r
a
m
.
Al
o
n
g
w
it
h
ti
m
e
m
as
k
i
n
g
,
w
e
u
s
e
d
f
r
eq
u
e
n
cy
m
as
k
i
n
g
,
w
h
i
c
h
i
n
v
o
l
v
ed
r
e
m
o
v
i
n
g
t
h
e
f
r
e
q
u
en
ci
es
f
r
o
m
th
e
s
p
ec
t
r
o
g
r
am
s
.
T
h
ese
au
g
m
e
n
te
d
s
a
m
p
les
c
o
m
p
o
s
ed
o
f
v
a
r
ia
ti
o
n
s
n
o
t o
n
l
y
i
n
th
e
ti
m
e
d
o
m
ai
n
b
u
t a
ls
o
in
t
h
e
f
r
e
q
u
e
n
c
y
d
o
m
a
in
h
el
p
e
d
th
e
m
o
d
els
b
ec
o
m
e
f
am
ili
ar
wit
h
v
a
r
i
o
u
s
co
n
d
it
io
n
s
a
n
d
p
a
tte
r
n
s
.
Fig
u
r
e
5
s
h
o
ws
th
e
c
o
n
n
ec
tio
n
b
etwe
en
th
e
h
ea
r
t
s
o
u
n
d
s
ig
n
als
tim
e
-
war
p
in
g
f
ac
to
r
an
d
th
e
s
ig
n
al
d
u
r
atio
n
.
E
ac
h
d
ata
p
o
i
n
t
r
ep
r
esen
ts
a
s
ep
ar
ate
h
ea
r
t
s
o
u
n
d
s
ig
n
al,
wh
er
e
t
h
e
s
am
p
le
I
D
s
h
o
ws
th
e
ty
p
e
o
f
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Dee
p
lea
r
n
in
g
fo
r
ea
r
ly
d
etec
t
io
n
o
f c
a
r
d
io
v
a
s
cu
la
r
d
is
ea
s
es v
ia
a
u
s
cu
lta
tio
n
s
o
u
n
d
…
(
S
h
r
ey
a
s
K
a
s
tu
r
e
)
1753
s
ig
n
al
(
n
o
r
m
al,
m
u
r
m
u
r
,
o
r
a
b
n
o
r
m
al
)
.
T
h
e
tim
e
war
p
in
g
am
o
u
n
ts
r
a
n
g
in
g
f
r
o
m
.
6
(
ti
m
e
co
m
p
r
ess
io
n
)
to
1
.
4
(
tim
e
s
tr
etch
in
g
)
f
o
r
m
th
e
x
-
ax
is
,
s
h
o
win
g
th
at
th
ese
lev
els
o
f
war
p
in
g
alter
th
e
tem
p
o
r
al
p
atter
n
s
an
d
th
e
h
ea
r
t
r
ate
s
en
s
atio
n
s
co
n
s
id
er
a
b
ly
.
T
h
e
s
ig
n
al
d
u
r
atio
n
alo
n
g
th
e
y
-
a
x
is
s
h
o
ws
th
e
len
g
th
c
h
an
g
es
d
u
e
to
tim
e
war
p
in
g
.
T
h
e
h
ig
h
er
th
e
tim
e
-
war
p
in
g
f
ac
t
o
r
,
th
e
l
o
n
g
e
r
o
r
s
h
o
r
ter
th
e
d
u
r
atio
n
.
Fig
u
r
e
4
.
Den
o
is
in
g
p
er
f
o
r
m
a
n
ce
o
f
h
ea
r
t so
u
n
d
s
ig
n
als
Fig
u
r
e
5
.
T
im
e
-
wr
ap
p
i
n
g
f
ac
t
o
r
s
an
d
ch
a
n
g
es in
tem
p
o
r
al
p
atter
n
s
4
.
3
.
Dee
p
lea
rning
mo
del dev
elo
pm
ent
T
h
e
R
esNet
ar
ch
itectu
r
e
u
s
es
r
esid
u
al
co
n
n
ec
tio
n
s
to
en
a
b
le
in
f
o
r
m
atio
n
to
f
lo
w
f
r
o
m
lo
wer
to
h
ig
h
er
lay
er
s
,
cr
u
cial
in
ad
d
r
ess
in
g
th
e
v
an
is
h
in
g
g
r
a
d
ie
n
t
is
s
u
es
in
d
ee
p
n
etwo
r
k
s
.
T
h
ese
h
id
d
en
p
at
h
s
ass
is
ted
th
e
d
ee
p
C
NN
m
o
d
els
in
tr
ain
in
g
f
o
r
h
ea
r
t
s
o
u
n
d
an
al
y
s
is
.
W
e
co
n
s
id
er
ed
th
e
Mo
b
ileNetV2
ar
ch
itectu
r
e
,
co
n
s
is
tin
g
o
f
d
e
p
th
-
wis
e
s
ep
ar
ab
le
co
n
v
o
lu
ti
o
n
s
an
d
in
v
er
ted
r
esid
u
al
b
l
o
ck
s
.
T
h
is
t
y
p
e
o
f
ar
ch
itectu
r
e
is
co
m
p
u
tatio
n
all
y
ef
f
icien
t
an
d
aim
ed
at
d
ep
l
o
y
in
g
o
n
m
o
b
ile
an
d
em
b
e
d
d
ed
d
ev
ices.
So
,
it
is
m
ea
n
t
to
b
e
im
p
lem
en
ted
in
o
u
r
m
o
b
ile
C
VD
d
etec
tio
n
s
y
s
tem
.
I
n
ad
d
itio
n
,
we
tr
ied
o
u
t
th
e
C
NN
-
DAG
ar
ch
itectu
r
e,
wh
ich
co
n
tain
s
DAG
s
tr
u
ctu
r
e
in
s
tead
o
f
a
s
eq
u
en
tial
o
n
e.
T
h
is
m
ec
h
a
n
is
m
ac
h
iev
es
b
o
th
th
e
ef
f
icien
cy
o
f
p
r
o
ce
s
s
in
g
an
d
t
h
e
ef
f
ec
tiv
e
n
ess
o
f
co
m
p
u
tatio
n
al
r
eso
u
r
ce
u
tili
za
tio
n
,
wh
ic
h
ca
n
ac
ce
ler
ate
t
h
e
s
p
ee
d
o
f
tr
ain
i
n
g
an
d
p
er
f
o
r
m
an
ce
o
f
m
o
d
els.
I
n
o
u
r
C
NN
a
r
ch
itectu
r
e
d
esig
n
,
we
co
n
s
id
er
ed
th
e
f
ea
tu
r
es
o
f
th
e
in
p
u
t
d
ata,
s
u
ch
as
s
p
ec
tr
o
g
r
am
o
r
s
ca
lo
g
r
a
m
,
s
o
th
at
t
h
e
m
o
d
els
co
u
ld
ac
q
u
ir
e
th
e
r
eq
u
ir
ed
s
p
atial
an
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
2
,
Ap
r
il 2
0
2
6
:
1
7
4
6
-
1
7
6
1
1754
tem
p
o
r
al
p
atter
n
s
in
th
e
tim
e
-
f
r
eq
u
e
n
cy
r
ep
r
esen
tatio
n
s
.
A
n
o
th
er
ar
e
a
we
lo
o
k
ed
in
to
was
u
s
in
g
L
STM
ce
lls
,
a
v
ar
ian
t
o
f
R
NN
th
at
is
ef
f
e
ctiv
ely
ap
p
lied
to
aid
th
e
d
y
n
am
ics
an
aly
s
is
o
f
s
eq
u
en
tial
d
ata
o
v
er
tim
e.
T
h
is
lets
y
o
u
m
o
d
el
lo
n
g
-
ter
m
d
e
p
en
d
en
cies
an
d
t
h
e
tim
e
p
atter
n
s
em
b
ed
d
ed
in
h
ea
r
t
s
o
u
n
d
s
ig
n
als.
R
ath
er
th
a
n
ce
n
ter
in
g
o
n
t
h
e
is
s
u
es
ass
o
ci
ated
with
v
an
is
h
in
g
g
r
ad
ien
t
p
r
o
b
lem
s
,
w
h
ich
m
a
k
e
th
em
u
n
f
it
f
o
r
lo
n
g
-
ter
m
r
elatio
n
s
h
ip
m
o
d
elin
g
,
th
e
y
o
p
t
f
o
r
th
e
atten
tio
n
m
ec
h
an
is
m
.
T
h
e
im
p
lem
e
n
tatio
n
o
f
th
e
m
o
d
el
em
b
r
ac
ed
th
e
u
s
e
o
f
an
L
STM
n
e
u
r
al
n
etwo
r
k
,
wh
ich
b
r
in
g
s
th
e
r
e
q
u
ir
e
d
d
y
n
am
ics
an
d
p
atter
n
s
to
th
e
h
ea
r
t
s
o
u
n
d
s
ig
n
als
b
y
p
e
r
f
o
r
m
in
g
t
h
em
as
s
eq
u
en
ce
s
.
T
h
e
L
STM
s
th
at
we
i
n
OURi
e
h
av
e
em
p
lo
y
ed
u
til
ize
th
e
p
r
o
ce
s
s
o
f
s
elec
tiv
e
s
to
r
in
g
an
d
f
o
r
g
ettin
g
in
f
o
r
m
atio
n
,
wh
o
s
e
r
esp
o
n
s
ib
ilit
y
lies
in
m
em
o
r
izin
g
tem
p
o
r
al
f
ea
tu
r
es
o
f
h
ea
r
t
s
o
u
n
d
s
eq
u
e
n
ce
s
.
W
e
tr
ied
d
if
f
er
en
t
L
STM
ar
ch
itect
u
r
es
o
r
co
n
f
i
g
u
r
atio
n
s
th
er
e
,
an
d
th
e
n
u
m
b
er
o
f
lay
er
s
was
ch
an
g
ed
m
an
y
tim
es.
W
e
also
u
s
ed
th
e
n
u
m
b
er
o
f
h
id
d
e
n
u
n
its
(
n
eu
r
o
n
s
)
an
d
b
id
ir
ec
tio
n
al
R
NN
lay
er
s
in
th
is
ca
s
e.
T
o
ef
f
ec
ti
v
ely
r
e
p
r
o
d
u
ce
th
e
co
n
s
ec
u
tiv
e
n
atu
r
e
o
f
h
ea
r
t
s
o
u
n
d
s
ig
n
als
an
d
th
e
tim
e
d
e
p
en
d
en
cies,
we
p
ass
ed
th
e
s
ig
n
als
as
s
eq
u
en
ce
s
o
r
th
eir
f
ea
tu
r
es
as
s
eq
u
en
ce
s
to
th
e
R
NN
s
.
W
e
tr
ea
ted
th
e
h
ea
r
t
-
s
o
u
n
d
d
ata
as
s
eq
u
en
ce
s
o
f
o
b
s
er
v
atio
n
s
an
d
o
r
g
an
ize
d
th
em
in
to
s
eq
u
en
ce
f
o
r
m
.
Seq
u
en
tially
g
iv
en
in
p
u
ts
wer
e
th
en
p
r
o
v
id
ed
to
eith
e
r
L
STM
o
r
GR
U
ar
ch
itectu
r
es
o
f
R
N
N
m
o
d
els.
T
h
e
R
NN
m
o
d
els
l
ea
r
n
an
d
m
e
m
o
r
iz
e
tem
p
o
r
al
d
ep
e
n
d
en
cies
an
d
p
atter
n
s
in
th
e
p
r
o
ce
s
s
in
g
s
eq
u
en
ce
s
.
R
ec
u
r
r
in
g
co
n
n
ec
tio
n
s
ac
r
o
s
s
R
NN
u
n
it
s
en
ab
led
in
f
o
r
m
atio
n
to
m
o
v
e
th
r
o
u
g
h
th
e
s
eq
u
en
ce
,
wh
er
e
m
o
d
els
co
u
l
d
lear
n
lo
n
g
-
r
an
g
e
d
ep
en
d
en
cies
an
d
m
o
d
el
tem
p
o
r
al
d
y
n
am
ics
in
th
e
d
ata.
T
h
e
g
atin
g
m
ec
h
a
n
is
m
s
in
L
STM
an
d
GR
U
u
n
its
ad
v
an
ce
d
t
h
is
ca
p
ab
ilit
y
,
s
elec
tiv
ely
r
em
em
b
er
in
g
o
r
f
o
r
g
ettin
g
d
ata
as
r
eq
u
ir
ed
.
T
h
r
o
u
g
h
th
e
R
NN
tr
ain
in
g
,
th
e
m
o
d
els
lear
n
ed
to
ex
tr
ac
t
c
o
m
p
eten
t
r
ep
r
esen
tatio
n
s
f
r
o
m
th
e
s
eq
u
e
n
tial
d
ata
b
y
ca
p
tu
r
i
n
g
th
e
p
at
ter
n
s
ch
ar
ac
ter
is
tic
o
f
d
if
f
e
r
en
t
h
ea
r
t
s
o
u
n
d
c
o
n
d
it
io
n
s
an
d
C
VD
s
.
T
h
e
lear
n
ed
r
ep
r
esen
tatio
n
f
r
o
m
th
e
R
NN
m
o
d
els
was
f
ed
in
to
class
if
icatio
n
task
s
f
o
llo
win
g
tr
ain
in
g
.
Su
ch
r
ep
r
esen
tatio
n
s
co
u
ld
b
e
co
n
n
ec
ted
in
to
t
h
e
f
u
lly
co
n
n
ec
ted
lay
er
s
o
r
co
m
b
in
ed
with
o
th
er
m
o
d
el
co
m
p
o
n
en
ts
lik
e
co
n
v
o
lu
tio
n
al
la
y
er
s
to
g
iv
e
th
e
f
i
n
al
class
if
icatio
n
o
f
h
ea
r
t so
u
n
d
s
in
t
o
ca
teg
o
r
ies s
u
ch
as n
o
r
m
al
a
n
d
a
b
n
o
r
m
al
o
r
ca
r
d
io
v
ascu
lar
c
o
n
d
itio
n
s
d
iag
n
o
s
is
.
4
.
4
.
M
o
del
t
ra
ini
ng
a
nd
o
pti
m
iza
t
io
n
W
e
d
iv
id
ed
th
e
en
tire
h
ea
r
t
s
o
u
n
d
d
ataset
in
to
th
r
ee
s
u
b
s
ets:
v
ar
io
u
s
attr
ib
u
tes
lik
e
tr
ain
in
g
,
v
alid
atio
n
,
an
d
test
in
g
.
T
h
e
t
r
ain
in
g
d
ata
s
ets
tr
ain
th
e
d
e
ep
lear
n
in
g
m
o
d
els
b
y
lear
n
i
n
g
th
e
p
atter
n
s
an
d
r
ep
r
esen
tatio
n
s
to
class
if
y
d
ee
p
s
o
u
n
d
s
.
T
h
e
v
alid
atio
n
s
u
b
s
et
was
p
ar
am
o
u
n
t
wh
en
ch
o
o
s
in
g
th
e
m
o
d
el
an
d
f
in
e
-
tu
n
in
g
th
e
p
ar
am
eter
s
.
T
h
e
r
eg
u
lar
izatio
n
tech
n
iq
u
e
was f
r
eq
u
en
tly
ap
p
lied
d
u
r
i
n
g
th
e
tr
ain
in
g
p
r
o
ce
s
s
to
p
r
ev
en
t
o
v
er
f
itti
n
g
.
T
h
e
m
o
d
els
wer
e
ev
alu
ated
u
s
in
g
th
ei
r
p
er
f
o
r
m
a
n
ce
o
n
th
e
v
alid
ati
o
n
s
et
af
ter
ev
er
y
iter
atio
n
,
wh
ich
was
u
s
ed
to
a
d
ju
s
t
th
e
h
y
p
e
r
p
ar
am
ete
r
s
an
d
s
elec
t
th
e
b
est
-
p
er
f
o
r
m
in
g
m
o
d
el
c
o
n
f
ig
u
r
atio
n
.
T
h
is
s
lice
en
ab
led
th
e
m
o
d
el
t
o
ev
alu
ate
its
ab
ilit
ies
o
n
u
n
k
n
o
wn
d
ata,
th
u
s
g
iv
i
n
g
a
le
v
e
led
p
lay
in
g
g
r
o
u
n
d
to
d
eter
m
in
e
g
en
er
ality
.
Usi
n
g
th
e
g
r
id
-
s
ea
r
ch
m
eth
o
d
,
we
co
n
s
tr
u
cted
a
g
r
id
o
f
elem
en
ts
o
f
th
e
h
y
p
er
p
ar
am
eter
s
p
ac
e
to
b
e
i
n
v
esti
g
ated
.
B
esid
es
th
at,
we
n
o
ticed
th
at
th
e
h
y
p
e
r
p
ar
am
et
er
s
th
at
m
o
s
t
af
f
ec
t
th
e
m
o
d
el
p
er
f
o
r
m
a
n
ce
-
lear
n
in
g
r
ate,
b
atch
s
ize,
n
u
m
b
er
o
f
h
id
d
en
la
y
er
s
,
d
r
o
p
o
u
t,
a
n
d
L
2
r
eg
u
lar
izatio
n
-
wer
e
th
e
m
o
s
t
cr
itical.
W
e
tr
i
ed
th
e
m
o
d
el
a
n
d
its
p
er
f
o
r
m
an
ce
b
y
tr
ain
in
g
an
d
ev
alu
ati
n
g
th
e
m
o
d
el
o
n
th
e
v
alid
atio
n
s
et
b
ased
o
n
th
e
f
ac
to
r
s
we
h
ad
in
th
e
d
ef
in
ed
g
r
id
.
Af
ter
th
e
ex
h
a
u
s
tiv
e
g
r
id
s
ea
r
ch
o
v
er
all
h
y
p
er
p
ar
am
eter
c
o
m
b
in
atio
n
s
,
we
p
ick
th
e
co
m
b
i
n
atio
n
o
f
p
ar
am
eter
s
th
at
af
f
o
r
d
s
u
s
th
e
b
est
r
esu
lts
o
n
th
e
v
alid
atio
n
s
et.
T
h
is
ap
p
r
o
ac
h
was
co
m
p
lem
en
ted
b
y
th
e
B
ay
esian
o
p
tim
izatio
n
ap
p
r
o
a
ch
,
wh
ich
m
a
k
es
it
p
o
s
s
ib
le
to
d
ea
l
with
h
y
p
e
r
p
ar
am
eter
tu
n
in
g
p
r
o
b
lem
s
ef
f
icien
tly
.
B
ay
esian
o
p
tim
izatio
n
ex
p
l
o
its
th
e
p
r
ev
io
u
s
k
n
o
w
-
h
o
w
a
n
d
a
n
o
r
g
an
ized
s
ea
r
ch
i
n
th
e
p
ar
a
m
eter
s
p
ac
e,
r
esu
ltin
g
in
a
b
ig
co
s
t
r
e
d
u
ctio
n
co
m
p
ar
ed
to
a
n
o
n
-
p
ar
a
m
etr
ic
g
r
id
s
ea
r
ch
.
Fig
u
r
e
6
d
e
p
icts
th
at
th
er
e
ar
e
o
f
ten
h
u
n
d
r
ed
s
o
f
c
o
m
b
in
atio
n
s
f
o
r
ea
ch
h
y
p
e
r
p
ar
am
ete
r
th
at
tak
e
astro
n
o
m
ic
v
alu
es,
im
p
l
y
in
g
t
h
e
lin
k
b
etwe
en
th
e
m
o
d
el'
s
p
er
f
o
r
m
a
n
ce
an
d
th
e
h
y
p
er
p
ar
am
eter
tu
n
i
n
g
p
r
o
ce
d
u
r
e.
E
v
er
y
tick
m
ar
k
lab
eled
o
n
th
e
x
-
ax
is
is
alig
n
ed
with
a
c
er
tain
co
m
b
in
atio
n
o
f
h
y
p
er
p
ar
am
eter
s
o
f
th
e
m
o
d
el,
in
clu
d
i
n
g
t
h
e
r
ate
o
f
lear
n
in
g
,
th
e
n
u
m
b
er
o
f
lay
er
s
,
th
e
s
ize
o
f
a
b
atch
a
n
d
o
th
er
s
.
T
h
e
y
-
ax
is
o
f
th
e
g
r
ap
h
d
is
p
lay
s
two
im
p
o
r
tan
t
p
er
f
o
r
m
an
ce
m
etr
ics:
v
alid
at
io
n
ac
cu
r
ac
y
an
d
v
alid
atio
n
F1
-
s
co
r
e
.
T
h
ese
m
e
tr
ics
in
d
icate
d
h
o
w
th
e
d
ee
p
l
ea
r
n
in
g
m
o
d
el
f
ar
e
d
in
th
e
v
a
lid
atio
n
d
ataset
f
o
r
ea
ch
o
f
th
e
h
y
p
er
p
ar
a
m
eter
s
th
at
wer
e
b
ein
g
tu
n
ed
.
W
e
p
lan
n
ed
to
ad
o
p
t
a
n
en
s
em
b
le
ap
p
r
o
ac
h
th
at
co
llected
s
ev
er
al
d
ee
p
-
lear
n
in
g
s
o
lu
tio
n
s
f
o
r
b
etter
tr
ain
in
g
an
d
wid
e
r
s
tab
ilit
y
.
Mo
r
eo
v
er
,
we
also
an
aly
ze
d
th
e
m
o
d
els
o
f
b
ag
g
i
n
g
,
b
o
o
s
tin
g
,
an
d
s
tack
in
g
th
at
p
r
o
d
u
ce
d
m
u
ltip
le
m
o
d
els,
an
d
af
ter
th
at,
a
co
llectio
n
o
f
th
e
ir
f
o
r
ec
asts
was
d
is
tr
ib
u
ted
.
T
h
e
s
tack
in
g
p
r
o
ce
s
s
was
ap
p
lied
to
th
e
tr
ain
in
g
m
eta
-
m
o
d
el,
wh
ic
h
r
ec
eiv
e
d
s
o
m
e
o
f
th
e
b
ase
m
o
d
els'
p
r
ed
ictio
n
s
as
in
p
u
ts
.
T
h
e
ty
p
e
–
th
e
jo
in
i
n
g
o
f
d
if
f
er
e
n
t
tech
an
d
s
k
ills
o
f
th
e
m
o
d
els
–
g
ets
b
etter
r
esu
lts
,
an
d
th
e
s
h
o
r
tco
m
in
g
s
ar
e
in
th
e
s
en
s
e
o
f
th
e
m
eth
o
d
o
f
e
n
s
em
b
le
m
itig
atio
n
.
T
ab
le
1
p
r
esen
ts
th
e
p
er
f
o
r
m
a
n
ce
s
co
r
es
o
f
d
if
f
er
en
t
m
o
d
els
f
o
r
th
e
d
is
cr
im
in
an
t
a
n
aly
s
is
o
f
a
h
ea
r
t
s
o
u
n
d
o
n
th
e
test
s
et.
T
h
e
r
ep
o
r
ted
p
er
f
o
r
m
an
ce
cr
it
er
io
n
th
at
m
ea
s
u
r
es
g
o
o
d
r
esu
lts
is
ac
cu
r
ac
y
,
p
r
ec
i
s
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e.
T
h
e
ac
cu
r
ac
y
r
ate
th
en
is
th
e
to
tal
co
u
n
t
o
f
t
r
u
e
class
if
icatio
n
s
an
d
th
e
r
elativ
e
p
o
r
tio
n
of
th
o
s
e.
Ho
wev
er
,
p
r
ec
is
io
n
r
ef
er
s
to
th
e
p
r
o
p
o
r
tio
n
o
f
p
o
s
itiv
e
p
r
ed
ictio
n
s
,
an
d
it
is
ca
lcu
lated
in
th
e
o
p
p
o
s
ite
wa
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Dee
p
lea
r
n
in
g
fo
r
ea
r
ly
d
etec
t
io
n
o
f c
a
r
d
io
v
a
s
cu
la
r
d
is
ea
s
es v
ia
a
u
s
cu
lta
tio
n
s
o
u
n
d
…
(
S
h
r
ey
a
s
K
a
s
tu
r
e
)
1755
C
u
m
u
lativ
e
r
ec
all
p
o
in
ts
o
u
t
h
o
w
m
an
y
r
ea
l
p
o
s
itiv
es
wer
e
co
r
r
ec
tly
esti
m
ated
b
y
t
h
e
m
o
d
el,
a
n
d
th
e
F1
-
s
co
r
e
is
th
e
h
ar
m
o
n
ic
av
e
r
ag
e
o
f
b
o
th
p
r
ec
is
io
n
an
d
r
e
ca
ll
m
etr
ics.
T
h
e
d
iag
n
o
s
tic
m
o
d
el
d
ip
PIN
h
as
p
r
o
v
e
d
to
m
o
d
el
d
ip
PIN
class
if
icatio
n
,
an
d
it
is
en
v
is
ag
ed
t
h
at
p
er
f
o
r
m
a
n
ce
im
p
r
o
v
em
e
n
t
will
b
e
b
ased
o
n
th
e
ac
tu
al
d
ata
af
ter
u
s
in
g
th
e
en
s
em
b
le
ap
p
r
o
ac
h
.
T
h
is
m
o
d
el
en
s
u
r
e
d
th
e
ac
tu
al
wea
k
n
ess
es
,
as
th
ey
ar
e
ea
ch
m
o
d
el
ass
im
ilate
s
th
e
s
tr
en
g
th
s
o
f
th
e
o
th
er
.
Fig
u
r
e
6
.
Hy
p
er
p
a
r
am
eter
co
m
b
in
atio
n
s
an
d
th
eir
co
r
r
esp
o
n
d
in
g
v
alid
atio
n
ac
c
u
r
ac
y
a
n
d
F1
-
s
co
r
e
T
ab
le
1
.
Per
f
o
r
m
an
ce
co
m
p
a
r
is
o
n
o
f
d
i
f
f
er
en
t
m
o
d
els
M
o
d
e
l
A
c
c
u
r
a
c
y
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
sc
o
r
e
C
N
N
mo
d
e
l
0
.
8
7
0
.
8
5
0
.
8
9
0
.
8
7
R
N
N
mo
d
e
l
0
.
8
9
0
.
9
1
0
.
8
7
0
.
8
9
C
N
N
-
LSTM
h
y
b
r
i
d
m
o
d
e
l
0
.
9
1
0
.
9
2
0
.
9
0
0
.
9
1
En
se
mb
l
e
mo
d
e
l
0
.
9
4
0
.
9
5
0
.
9
3
0
.
9
4
4
.
5
.
E
v
a
lua
t
i
o
n
a
nd
deplo
y
m
ent
T
h
e
p
er
f
o
r
m
a
n
ce
s
f
o
r
m
u
ltip
le
class
s
co
r
es
ar
e
ev
alu
ated
with
o
u
t
tr
u
e
lab
els
p
e
r
tin
en
t
to
th
ese
s
p
ec
if
ic
class
es
b
ased
o
n
F1
-
s
co
r
e
,
r
ec
all,
an
d
p
r
ec
is
io
n
m
e
tr
ics.
Acc
u
r
ac
y
m
ea
s
u
r
es
th
e
d
eg
r
ee
o
f
co
r
r
ec
tly
class
if
ied
d
ata
s
am
p
les
am
o
n
g
th
e
en
tire
d
ataset.
T
h
e
m
etr
ics
wer
e
g
iv
en
o
u
t
to
f
u
n
ctio
n
as
a
y
ar
d
s
tick
to
m
ea
s
u
r
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
m
o
d
el.
I
t
is
also
o
n
e
o
f
th
e
tr
u
e
o
n
es,
an
d
it
was
co
r
r
ec
tly
co
n
clu
d
ed
.
T
h
e
ap
p
r
o
ac
h
with
th
e
m
in
im
u
m
n
u
m
b
er
o
f
f
alse
n
eg
ativ
es
is
g
o
o
d
b
ec
au
s
e
it
co
r
r
ec
tly
lab
els
th
e
class
.
Sin
ce
we
v
alid
ated
th
at
th
e
F1
-
s
co
r
e
c
r
iter
io
n
p
er
f
o
r
m
a
n
ce
is
co
r
r
ec
t,
we
ca
n
f
r
ee
ly
s
witch
to
r
e
ca
ll
an
d
p
r
ec
is
io
n
,
wh
ich
ar
e
th
e
d
r
iv
i
n
g
f
o
r
ce
s
o
f
th
e
d
ev
elo
p
ed
m
o
d
el.
T
h
e
d
ee
p
lear
n
in
g
m
o
d
el
(
Fig
u
r
e
7
)
f
o
r
h
ea
r
t
s
o
u
n
d
an
aly
s
is
was
te
s
ted
an
d
ev
alu
a
ted
u
s
in
g
v
ar
io
u
s
d
o
s
ag
es
ac
r
o
s
s
th
e
h
ea
r
t
s
o
u
n
d
ca
teg
o
r
ies.
Ou
r
m
o
d
el
h
a
d
an
o
v
er
all
ac
cu
r
ac
y
o
f
0
.
9
2
f
o
r
t
h
e
n
o
r
m
al
h
ea
r
t
s
o
u
n
d
s
ca
teg
o
r
y
,
im
p
ly
in
g
a
lik
en
ess
b
etw
ee
n
its
d
iag
n
o
s
is
an
d
ac
tu
al
ca
s
es.
I
n
e
v
alu
atin
g
th
e
m
u
r
m
u
r
s
id
en
tific
atio
n
,
b
o
a
s
tin
g
a
p
r
ec
is
io
n
o
f
0
.
9
1
,
th
e
m
o
d
e
y
ield
s
a
lo
w
r
ate
o
f
f
alse
p
o
s
itiv
es
with
in
th
e
r
esu
lts
.
On
th
e
o
th
er
h
an
d
,
th
e
p
r
ec
is
io
n
o
f
m
u
r
m
u
r
s
r
ec
all
was
lo
wer
at
0
.
8
5
,
in
d
icatin
g
th
at
m
u
r
m
u
r
s
m
ig
h
t
b
e
co
n
f
u
s
ed
with
s
o
m
e
in
s
tan
ce
s
o
f
m
u
r
m
u
r
s
.
T
h
e
p
r
ec
is
io
n
v
alu
e
f
o
r
th
e
m
o
d
el'
s
ex
tr
asy
s
to
les
class
,
wh
ich
m
ea
n
s
ac
ce
p
ta
b
le
b
u
t
c
u
r
tailin
g
er
r
o
r
s
,
i
n
d
icate
d
r
o
o
m
f
o
r
im
p
r
o
v
em
e
n
t
f
o
r
ca
s
es
wh
er
e
f
alse
p
o
s
itiv
es
an
d
n
eg
ativ
e
r
ea
ctio
n
s
wer
e
u
n
ac
ce
p
tab
le.
T
h
e
ar
r
h
y
th
m
ia,
an
ir
r
eg
u
lar
h
ea
r
t
b
ea
t,
was
d
iag
n
o
s
ed
with
h
ig
h
f
id
elity
with
a
p
r
ec
is
io
n
m
etr
ic
o
f
0
.
9
3
an
d
a
r
ec
all
v
alu
e
o
f
0
.
9
0
.
T
h
e
F1
-
s
co
r
e
h
a
d
s
o
lid
p
o
in
ts
r
is
in
g
f
r
o
m
0
.
8
4
f
o
r
e
x
tr
asy
s
to
le
to
0
.
9
2
f
o
r
ar
r
h
y
th
m
ia,
an
d
th
o
s
e
s
tead
y
p
o
in
ts
im
p
ly
th
at
th
e
m
o
d
el
h
a
s
a
b
alan
ce
d
p
e
r
f
o
r
m
an
ce
ac
r
o
s
s
all
clas
s
es.
Mo
b
ile
d
ev
ices
u
tili
s
e
o
u
r
d
e
ep
lear
n
in
g
m
o
d
els
f
o
r
h
ea
r
t
s
o
u
n
d
class
if
icatio
n
,
en
a
b
lin
g
i
m
m
ed
iate
h
ea
r
t
d
is
ea
s
e
d
iag
n
o
s
tics
.
W
e
f
ir
s
t
o
p
tim
ized
an
d
co
m
p
r
e
s
s
ed
th
e
n
eu
r
al
n
etwo
r
k
s
to
d
ec
r
em
en
t
m
em
o
r
y
f
o
o
tp
r
i
n
t
an
d
co
m
p
u
te
d
em
an
d
s
to
d
ep
lo
y
th
e
m
o
d
els
to
s
m
ar
tp
h
o
n
es.
W
e
a
p
p
li
ed
tech
n
iq
u
es
lik
e
q
u
an
tizatio
n
,
p
r
u
n
in
g
,
an
d
m
o
d
el
d
is
till
atio
n
to
d
ev
elo
p
th
e
r
ed
u
ce
d
-
s
ize
m
o
d
els with
o
u
t a
p
p
r
ec
iab
le
wo
r
k
i
n
g
ca
p
ac
ity
lo
s
s
.
W
e
lev
er
ag
e
m
o
b
ile
ap
p
d
ev
elo
p
m
en
t
u
s
in
g
cr
o
s
s
-
p
latf
o
r
m
f
r
am
ewo
r
k
s
a
n
d
lib
r
ar
ies
th
at
ex
ce
l
Evaluation Warning : The document was created with Spire.PDF for Python.