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.
8
,
No
.
3
,
Sep
tem
b
er
201
9
,
p
p
.
264
~
269
I
SS
N:
2
2
5
2
-
8938
,
DOI
: 1
0
.
1
1
5
9
1
/i
j
ai.
v
8
.i
3
.
p
p
264
-
2
6
9
264
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
co
r
e
.
co
m/o
n
lin
e/in
d
ex
.
p
h
p
/I
J
AI
Rev
iew
of ano
m
a
l
o
us so
und event
d
etec
tion
appro
a
ches
A
m
irul Sa
d
i
k
in M
d Af
endi
1
,
M
a
rina
Yus
o
f
f
2
1
F
a
c
u
lt
y
o
f
Co
m
p
u
ter an
d
M
a
th
e
m
a
ti
c
a
l
S
c
ien
c
e
s,
Un
iv
e
rsiti
T
e
k
n
o
lo
g
i
M
A
RA
,
S
h
a
h
A
la
m
,
S
e
lan
g
o
r,
M
a
lay
sia
2
A
d
v
a
n
c
e
d
A
n
a
l
y
ti
c
En
g
in
e
e
rin
g
Ce
n
ter (AA
EC),
F
a
c
u
lt
y
o
f
Co
m
p
u
ter an
d
M
a
th
e
m
a
ti
c
a
l
S
c
ien
c
e
s,
Un
iv
e
rsiti
T
e
k
n
o
lo
g
i
M
A
R
A
,
S
h
a
h
A
la
m
,
S
e
lan
g
o
r,
M
a
la
y
sia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
n
2
7
,
2
0
1
9
R
ev
i
s
ed
A
u
g
5
,
2
0
1
9
A
cc
ep
ted
A
u
g
2
0
,
2
0
1
9
T
h
is
p
a
p
e
r
p
re
se
n
ts
a
r
e
v
ie
w
o
f
a
n
o
m
a
lo
u
s
so
u
n
d
e
v
e
n
t
d
e
tec
ti
o
n
(S
ED)
a
p
p
ro
a
c
h
e
s.
S
ED
is
b
e
c
o
m
in
g
m
o
re
a
p
p
li
c
a
b
le
f
o
r
re
a
l
-
w
o
rld
a
p
p
li
a
c
taio
n
s
su
c
h
a
s
se
c
u
rit
y
,
f
ire
d
e
ter
m
in
a
ti
o
n
o
r
o
lt
h
e
r
e
m
e
rg
e
n
c
y
a
lar
m
s.
De
sp
it
e
m
a
n
y
re
se
a
rc
h
o
u
tco
m
e
p
re
v
io
u
sly
,
f
u
rth
e
r
re
se
a
rc
h
is
r
e
q
u
ired
t
o
re
d
u
c
e
f
a
lse
p
o
siti
v
e
s
a
n
d
i
m
p
ro
v
e
a
c
c
u
rra
c
y
.
S
ED
a
p
p
r
o
a
c
h
e
s
a
re
c
o
m
p
re
h
e
n
siv
e
l
y
o
rg
a
n
ize
d
b
y
m
e
th
o
d
s
c
o
v
e
rin
g
s
y
ste
m
p
ip
e
li
n
e
c
o
m
p
o
n
e
n
ts
o
f
a
c
o
u
stic
d
e
sc
rip
to
rs,
c
las
sif
ica
ti
o
n
e
n
g
in
e
,
a
n
d
d
e
c
isio
n
f
in
a
li
z
a
ti
o
n
m
e
th
o
d
.
T
h
e
re
v
ie
w
c
o
m
p
a
re
s
m
u
lt
ip
le
a
p
p
r
o
a
c
h
e
s
th
a
t
is
a
p
p
li
e
d
o
n
a
sp
e
c
if
ic
d
a
tas
e
t.
S
e
c
u
rit
y
re
li
e
s
o
n
a
n
o
m
a
lo
u
s
e
v
e
n
ts
in
o
r
d
e
r
to
p
re
v
e
n
t
it
o
n
e
m
u
st
f
in
d
th
e
se
a
n
o
m
a
lo
u
s
e
v
e
n
ts.
A
u
d
io
su
rv
e
i
ll
a
n
c
e
h
a
s
b
e
c
o
m
e
m
o
re
e
ff
icie
n
t
a
s
th
a
t
a
rti
f
icia
l
i
n
telli
g
e
n
c
e
h
a
s
ste
p
p
e
d
u
p
th
e
g
a
m
e
.
A
u
to
n
o
m
o
u
s
S
ED
c
o
u
ld
b
e
u
se
d
f
o
r
e
a
rl
y
d
e
tec
ti
o
n
a
n
d
p
re
v
e
n
ti
o
n
.
It
is
f
o
u
n
d
t
h
a
t
t
h
e
sta
te
o
f
th
e
a
r
t
m
e
th
o
d
v
iab
le
u
se
d
in
S
ED
u
si
n
g
f
e
a
tu
re
s
o
f
lo
g
-
m
e
l
e
n
e
rg
ie
s
in
c
o
n
v
o
lu
ti
o
n
a
l
re
c
u
rre
n
t
n
e
u
ra
l
n
e
tw
o
rk
(CR
NN
)
w
it
h
l
o
n
g
sh
o
rt
term
m
e
m
o
r
y
(
L
S
T
M
)
w
it
h
a
v
e
ri
f
ica
ti
o
n
ste
p
o
f
th
re
sh
o
ld
i
n
g
h
a
s
o
b
tai
n
e
d
9
3
.
1
%
F
1
sc
o
re
a
n
d
0
.
1
3
0
7
ER.
It
is
f
o
u
n
d
t
h
a
t
f
e
a
tu
re
e
x
trac
ti
o
n
o
f
lo
g
m
e
l
e
n
e
rg
ies
a
re
h
ig
h
l
y
re
li
a
b
le
m
e
th
o
d
sh
o
w
in
g
p
ro
m
isin
g
re
su
lt
s
o
n
m
u
lt
ip
le ex
p
e
rime
n
ts.
K
ey
w
o
r
d
s
:
A
co
u
s
tic
Fea
tu
r
es
Au
d
io
P
r
o
ce
s
s
in
g
P
atter
n
R
ec
o
g
n
itio
n
So
u
n
d
E
v
e
n
t D
etec
tio
n
Co
p
y
rig
h
t
©
2
0
1
9
In
stit
u
te o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
rin
g
a
n
d
S
c
ien
c
e
.
Al
l
rig
h
ts re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Ma
r
in
a
Yu
s
o
f
f
,
A
d
v
an
ce
d
An
al
y
t
ic
E
n
g
i
n
ee
r
in
g
C
en
ter
(
AA
E
C
)
,
Facu
lt
y
o
f
C
o
m
p
u
ter
an
d
Ma
t
h
e
m
a
tical
Scie
n
ce
s
,
Un
i
v
er
s
iti T
ek
n
o
lo
g
i M
A
R
A
,
Sh
a
h
A
la
m
,
Selan
g
o
r
,
Ma
la
y
s
i
a
.
E
m
ail:
m
ar
in
a
y
@
t
m
s
k
.
u
it
m
.
e
d
u
.
m
y
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
So
u
n
d
ev
e
n
t
d
etec
tio
n
(
SED)
r
esear
ch
f
ield
h
as
b
ee
n
an
ac
tiv
e
r
ec
e
n
t
l
y
[1
-
4
]
.
Au
to
n
o
m
o
u
s
a
u
d
io
s
u
r
v
ei
lla
n
ce
h
as
b
ec
o
m
e
m
o
r
e
e
f
f
ic
ien
t
a
s
t
h
at
ar
ti
f
icial
in
te
lli
g
en
ce
h
a
s
s
tep
p
ed
u
p
th
e
g
a
m
e
[
5
]
.
Ma
ch
in
e
lea
r
n
in
g
h
a
s
b
ec
o
m
e
v
er
y
p
o
w
er
f
u
l
to
allo
w
n
e
w
ap
p
r
o
ac
h
es
to
b
e
f
ac
ilit
ated
in
co
u
n
tle
s
s
d
o
m
ai
n
[
6
]
.
A
co
u
s
t
i
c
s
u
r
v
ei
llan
ce
s
p
ec
if
ical
l
y
i
n
s
ec
u
r
it
y
ap
p
licatio
n
is
s
t
ill
n
e
w
to
th
e
w
o
r
ld
an
d
r
eq
u
ir
es
r
esear
ch
to
allo
w
b
ett
er
p
er
f
o
r
m
a
n
ce
s
[
4
,
7
,
8
]
.
T
h
e
d
etec
tio
n
o
n
a
n
o
m
alo
u
s
a
u
d
i
o
ev
en
t
s
ef
f
ec
ti
v
el
y
w
h
ile
a
v
o
id
in
g
f
alse
p
o
s
i
tiv
e
s
is
cr
u
cia
l
in
s
ec
u
r
it
y
[
3
]
.
T
w
o
ca
teg
o
r
ies
o
f
s
o
u
n
d
r
ec
o
g
n
i
t
io
n
,
n
o
n
-
s
p
ee
ch
t
h
e
d
eter
m
in
at
io
n
o
f
th
e
s
o
u
n
d
e
v
en
t
s
o
u
r
ce
an
d
s
p
ee
ch
r
ec
o
g
n
i
tio
n
o
f
v
er
b
al
lan
g
u
ag
e
[9
-
10]
.
Ma
n
y
w
o
r
k
s
e
m
b
r
ac
e
th
e
m
et
h
o
d
o
f
m
u
lt
i
-
lab
el
class
if
icatio
n
to
d
etec
t
p
o
ly
p
h
o
n
ic
ac
o
u
s
tic
ev
e
n
ts
w
it
h
a
w
o
r
ld
w
id
e
li
m
i
t
to
d
etec
t
ac
tiv
e
ac
o
u
s
tic
e
v
e
n
ts
[
2
,
7
,
10]
.
T
h
e
lack
o
f
a
cc
u
r
ac
y
a
n
d
h
i
g
h
f
alse
p
o
s
it
i
v
es
o
n
cu
r
r
en
t
SED
ap
p
r
o
ac
h
es
is
h
o
ld
in
g
au
to
n
o
m
o
u
s
a
u
d
io
s
u
r
v
eilla
n
ce
to
b
e
ap
p
lied
f
o
r
r
ea
l
-
w
o
r
ld
ap
p
licatio
n
s
[
1
1
]
.
I
n
t
h
e
r
ac
e
to
w
ar
d
s
m
a
k
i
n
g
a
u
d
io
s
u
r
v
eil
lan
ce
a
r
ea
lit
y
m
an
y
h
a
v
e
a
id
b
y
p
r
o
v
id
i
n
g
d
ata
f
o
r
tr
ai
n
i
n
g
an
d
test
i
n
g
n
e
w
m
o
d
els
i
n
s
o
l
v
i
n
g
th
e
p
r
o
b
le
m
.
Data
s
e
ts
o
n
SE
D
ca
n
b
e
f
o
u
n
d
a
v
ailab
le
i
n
m
an
y
v
ar
ietie
s
as
t
h
e
in
ter
n
e
t o
f
th
in
g
s
ar
e
b
ec
o
m
i
n
g
m
o
r
e
p
o
p
u
lar
[
1
,
1
0
,
1
2
].
T
h
e
r
ar
e
ter
m
r
elate
s
to
t
h
e
p
o
s
s
ib
ilit
y
th
at
an
y
tar
g
et
s
o
u
n
d
ev
e
n
t
to
b
e
id
e
n
t
if
ied
m
a
y
h
ap
p
en
at
m
o
s
t
o
n
ce
w
ith
in
h
al
f
a
m
i
n
u
te
[
4
,
1
2
]
.
Dete
ctin
g
r
ar
e
s
o
u
n
d
ev
e
n
ts
ar
e
o
n
e
o
f
th
e
k
e
y
f
ea
tu
r
es
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
R
ev
iew
o
f a
n
o
ma
lo
u
s
s
o
u
n
d
e
ve
n
t d
etec
tio
n
a
p
p
r
o
a
c
h
es
(
A
mir
u
l S
a
d
ikin
Md
A
fen
d
i
)
265
f
o
r
s
ec
u
r
it
y
[
1
0
]
.
T
h
e
ap
p
r
o
ac
h
es
u
s
ed
in
p
r
ev
io
u
s
s
t
u
d
ies,
p
ar
ticu
lar
l
y
i
n
d
etec
ti
o
n
o
f
r
ar
e
ev
e
n
ts
ca
n
b
e
s
i
m
p
l
if
ied
b
y
th
e
co
m
b
i
n
ati
o
n
s
o
f
ce
r
tai
n
p
ar
ts
o
f
th
e
w
h
o
le
d
et
ec
tio
n
s
y
s
te
m
[
8
]
.
P
ar
ts
o
f
th
e
d
etec
tio
n
s
y
s
te
m
th
at
co
n
s
is
ts
o
f
th
e
a
u
d
io
ch
a
n
n
el
in
p
u
t
d
ata,
f
ea
t
u
r
e
e
x
tr
ac
tio
n
,
clas
s
if
icatio
n
,
an
d
d
ec
is
io
n
m
a
k
i
n
g
[
8
]
.
T
h
is
p
ap
er
p
r
esen
ts
a
r
ev
ie
w
o
f
a
n
o
m
alo
u
s
s
o
u
n
d
ev
e
n
t
d
et
ec
tio
n
ap
p
r
o
ac
h
es
b
y
o
r
g
a
n
izi
n
g
a
n
u
m
b
er
o
f
r
esear
ch
d
o
n
e
o
n
a
n
o
m
a
lo
u
s
SED
r
ec
e
n
tl
y
.
C
o
m
p
ar
in
g
t
h
e
m
o
n
t
h
e
b
asi
s
o
f
th
e
w
h
o
le
s
y
s
te
m
co
m
p
o
s
i
t
io
n
to
w
ar
d
s
th
e
r
es
u
lt
s
ac
h
ie
v
ed
.
Ob
s
er
v
atio
n
s
ar
e
m
ai
n
l
y
o
n
th
e
co
m
p
o
s
it
io
n
o
f
th
e
t
y
p
ical
SED
p
ip
elin
e
co
m
p
o
n
e
n
t
s
.
T
h
e
p
er
f
o
r
m
an
c
e
o
f
th
e
s
y
s
te
m
ar
e
ev
al
u
ate
d
o
n
a
s
in
g
le
an
d
m
u
lti
-
clas
s
b
asis
.
2.
SO
UND
E
V
E
N
T
DE
T
E
CT
I
O
N
T
h
e
j
o
b
o
f
id
en
ti
f
y
in
g
s
o
u
n
d
ev
en
t
s
i
n
cl
u
d
es
id
e
n
ti
f
y
i
n
g
a
n
d
class
i
f
y
i
n
g
s
o
u
n
d
s
i
n
a
u
d
io
f
o
r
ec
asti
n
g
an
d
o
f
f
s
et
s
o
u
n
d
s
f
o
r
d
if
f
er
e
n
t
ca
s
es
o
f
s
o
u
n
d
ev
e
n
ts
a
n
d
o
f
f
er
i
n
g
a
te
x
tu
a
l
d
escr
ip
to
r
f
o
r
ea
ch
[
1
3
-
14]
.
C
o
m
m
o
n
cla
s
s
i
f
icat
io
n
m
eth
o
d
u
s
ed
o
n
SED
ar
e
as
th
e
co
n
v
e
n
tio
n
al
s
p
ee
ch
r
ec
o
g
n
iti
o
n
[
1
2
]
.
A
lo
n
g
t
h
e
p
ip
elin
e
o
f
a
t
y
p
ical
SED
it
co
n
s
is
t
o
f
b
asicall
y
f
ea
tu
r
e
ex
tr
ac
tio
n
f
o
llo
w
ed
b
y
th
e
class
i
f
icatio
n
[
1
5
]
.
C
las
s
i
f
icatio
n
m
et
h
o
d
th
at
i
s
p
o
p
u
lar
r
ec
en
tl
y
co
n
s
is
t
o
f
C
o
n
v
o
lu
tio
n
Ne
u
r
al
Net
w
o
r
k
(
C
NN)
[
1
6
]
.
I
n
s
u
r
v
eil
lan
ce
SED
ar
c
h
itect
u
r
es
co
n
s
is
t
o
f
ad
d
itio
n
al
b
ac
k
g
r
o
u
n
d
s
u
b
tr
ac
tio
n
,
o
b
j
ec
t
tr
ac
k
in
g
&
s
it
u
atio
n
a
l
an
al
y
s
is
p
r
o
ce
s
s
es
i
n
t
h
e
p
ip
elin
e
[
8
]
.
L
ate
s
t
r
e
s
ea
r
ch
o
f
a
n
i
m
p
r
o
v
ed
p
ip
elin
e
s
u
g
g
est
a
v
er
if
ica
tio
n
s
tep
to
r
ed
u
ce
f
alse p
o
s
iti
v
es a
f
ter
t
h
e
SED
p
r
o
ce
s
s
[
3
]
.
Fig
u
r
e
1
s
h
o
w
s
t
h
e
ex
ten
d
ed
SED
p
ip
elin
e.
I
npu
t
t
ype
S
a
m
pl
i
ng
R
a
t
e
F
e
a
t
ur
e
E
xt
r
a
c
t
i
on
M
e
t
hod
B
a
c
kgr
oun
d
S
ubr
a
c
t
i
on
C
l
a
s
s
i
f
i
c
a
t
i
on
M
e
t
hod
D
e
c
i
s
i
on
M
a
ki
ng
Fig
u
r
e
1
.
So
u
n
d
ev
e
n
t d
etec
tio
n
p
ip
elin
e
I
n
th
e
f
ir
s
t
s
ta
g
e
in
p
u
t
t
y
p
e,
s
a
m
p
li
n
g
r
ate
an
d
f
ea
t
u
r
e
ex
tr
ac
tio
n
is
t
h
e
au
d
io
s
p
ec
if
i
ca
tio
n
s
o
f
p
r
ep
r
o
ce
s
s
in
g
s
tep
.
T
h
e
in
p
u
t
t
y
p
e
a
n
d
s
a
m
p
lin
g
r
ate
s
p
ec
if
icatio
n
ar
e
g
r
o
u
p
ed
w
i
th
f
ea
t
u
r
e
ex
tr
ac
tio
n
m
et
h
o
d
b
ec
au
s
e
t
h
e
y
ar
e
co
-
r
elate
d
,
in
p
u
t
t
y
p
e
o
f
m
o
n
o
o
r
s
ter
eo
ch
a
n
n
el
an
d
s
a
m
p
li
n
g
r
ate
w
ill
r
es
u
lt
o
f
d
if
f
er
e
n
t
a
m
o
u
n
t
o
f
f
ea
t
u
r
es
e
x
tr
ac
ted
.
Hen
ce
,
it
is
co
m
b
in
e
d
as
it
is
o
n
e
p
r
o
ce
s
s
i
n
t
h
e
p
i
p
elin
e
w
h
er
e
p
r
o
p
er
tu
n
in
g
co
u
ld
b
e
d
o
n
e.
T
h
e
a
u
d
io
is
p
r
o
ce
s
s
ed
to
t
y
p
ical
l
y
2
0
-
2
0
0
m
s
b
r
ief
f
r
a
m
e
s
to
o
b
tain
t
h
e
a
u
d
io
ch
ar
ac
ter
i
s
tic
s
o
f
c
h
o
ice
[
1
0
]
.
S
m
aller
f
r
a
m
es
w
i
ll
m
ak
e
u
p
m
o
r
e
f
ea
t
u
r
es c
o
u
ld
lead
to
o
v
er
f
itti
n
g
in
ar
ti
f
icia
l
in
telli
g
e
n
ce
s
y
s
te
m
s
,
w
h
ich
c
an
r
es
u
lt
i
n
lo
w
er
p
r
ec
is
io
n
d
u
r
in
g
cla
s
s
i
f
icatio
n
[
1
7
]
.
R
ep
r
esen
tatio
n
s
o
f
th
e
s
ig
n
al
r
a
n
g
e,
in
cl
u
d
in
g
s
u
c
h
Me
l
-
f
r
eq
u
en
c
y
co
e
f
f
icien
ts
(
MFC
C
)
,
Me
l
e
n
er
g
ies,
o
r
j
u
s
t
t
h
e
a
m
p
l
itu
d
e
o
r
en
er
g
y
r
an
g
e
ar
e
o
f
ten
u
s
ed
i
n
s
o
u
n
d
ev
e
n
t id
e
n
ti
f
icatio
n
[
13
-
14
].
2
.
1
.
Aco
us
t
ic
f
ea
t
ures
A
m
aj
o
r
co
m
p
o
n
e
n
t
i
n
t
h
e
S
E
D
p
ip
elin
e
is
th
e
f
ea
t
u
r
e
e
x
t
r
ac
tio
n
m
et
h
o
d
.
T
h
er
e
ar
e
m
a
n
y
f
ea
t
u
r
es
ca
n
b
e
e
x
tr
ac
ted
f
r
o
m
a
n
au
d
i
o
s
o
u
r
ce
s
ep
ar
ated
in
to
t
w
o
m
ain
t
y
p
e
s
o
f
f
ea
tu
r
e
s
p
ec
tr
al
f
ea
tu
r
es
an
d
r
h
y
t
h
m
f
ea
t
u
r
es.
C
o
m
m
o
n
s
p
ec
tr
al
f
ea
tu
r
es a
r
e
as lis
ted
b
elo
w
.
1.
C
o
m
p
u
te
a
ch
r
o
m
a
g
r
a
m
f
r
o
m
a
w
a
v
ef
o
r
m
o
r
p
o
w
er
s
p
ec
tr
o
g
r
a
m
.
2.
C
o
n
s
tan
t
-
Q
c
h
r
o
m
a
g
r
a
m
3.
Mel
-
s
ca
led
s
p
ec
tr
o
g
r
a
m
.
4.
Mel
-
f
r
eq
u
en
c
y
ce
p
s
tr
al
co
ef
f
i
cien
ts
(
MF
C
C
s
)
5.
r
o
o
t
-
m
ea
n
-
s
q
u
ar
e
(
R
MS)
v
alu
e
6.
C
o
m
p
u
te
th
e
s
p
ec
tr
al
ce
n
tr
o
id
.
7.
P
’
th
-
o
r
d
er
s
p
ec
tr
al
b
an
d
w
id
th
.
8.
Sp
ec
tr
al
co
n
tr
ast
9.
Sp
ec
tr
al
f
lat
n
ess
10.
R
o
ll
-
o
f
f
f
r
eq
u
e
n
c
y
.
11.
co
ef
f
icie
n
t
s
o
f
f
itti
n
g
an
n
th
-
o
r
d
er
p
o
ly
n
o
m
ia
l to
th
e
co
lu
m
n
s
o
f
a
s
p
ec
tr
o
g
r
a
m
12.
C
o
m
p
u
te
th
e
ze
r
o
-
cr
o
s
s
in
g
r
at
e
o
f
an
au
d
io
ti
m
e
s
er
ies.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
A
I
Vo
l.
8
,
No
.
3
,
Sep
tem
b
er
201
9
:
2
6
4
–
2
6
9
266
As
ce
p
s
tr
al
ch
ar
ac
ter
i
s
tics
ar
e
ca
lcu
la
ted
b
y
tak
in
g
th
e
tr
an
s
f
o
r
m
a
tio
n
o
f
th
e
d
is
to
r
ted
lo
g
ar
ith
m
ic
s
p
ec
tr
u
m
f
r
o
m
Fo
u
r
ier
,
th
e
y
co
n
tai
n
d
at
a
o
n
th
e
r
ate
ch
an
g
e
s
in
th
e
d
i
s
tin
c
t
s
p
ec
tr
u
m
b
an
d
s
.
B
ec
au
s
e
o
f
th
eir
ca
p
ac
ity
to
s
ep
ar
ate
t
h
e
e
f
f
ec
t
o
f
s
o
u
r
ce
an
d
f
ilter
i
n
a
v
o
ice
s
i
g
n
a
l
,
C
ep
s
tr
al
ch
ar
ac
ter
is
tics
ar
e
b
en
ef
icia
l
f
o
r
SED
.
Featu
r
es
co
u
ld
b
e
e
m
p
lo
y
ed
b
est
o
n
t
h
eir
s
p
ec
if
icall
y
d
esi
g
n
ed
p
u
r
p
o
s
es.
T
h
u
s
s
elec
tio
n
o
f
f
ea
tu
r
e
s
p
la
y
s
a
v
ital
r
o
le
o
n
SE
D
s
y
s
te
m
s
[
9
]
.
A
lter
n
at
iv
e
s
f
o
r
tailo
r
ed
m
ad
e
f
ea
t
u
r
e
e
x
tr
ac
tio
n
w
o
u
ld
b
e
th
e
r
a
w
w
av
e
f
o
r
m
s
ac
o
u
s
tic
m
o
d
ell
in
g
[
1
8
]
.
2
.
2
.
F
e
a
t
ured
da
t
a
s
et
T
h
e
d
ataset
f
ea
tu
r
ed
i
n
th
is
r
ev
ie
w
ar
e
b
ased
o
n
t
h
e
De
tec
tio
n
C
la
s
s
i
f
icatio
n
A
co
u
s
tic
Scen
e
a
n
d
E
v
en
ts
(
D
C
ASE)
2
0
1
7
T
am
p
er
e
U
n
i
v
er
s
iti
T
ec
n
o
lo
g
y
(
T
UT
)
R
ar
e
So
u
n
d
E
v
en
ts
2
0
1
7
d
ev
elo
p
m
en
t
d
ata
s
et
[
6
]
.
T
h
e
r
esu
lt
s
o
f
ap
p
r
o
ac
h
es
co
m
p
ar
e
d
ar
e
b
ased
o
n
th
is
d
ataset.
C
o
m
p
o
s
itio
n
o
f
au
d
io
m
ater
ials
s
o
u
n
d
ev
e
n
ts
w
h
ic
h
ar
e
k
n
o
w
n
ar
e
o
f
t
h
e
f
o
llo
w
i
n
g
clas
s
es:
1.
B
ab
y
cr
y
i
n
g
(
1
0
6
in
s
tan
ce
s
f
o
r
tr
ain
in
g
,
4
2
in
s
ta
n
ce
s
f
o
r
test
)
2.
Glass
b
r
ea
k
i
n
g
(
9
6
in
s
tan
ce
s
f
o
r
tr
ain
in
g
,
4
3
in
s
tan
ce
s
f
o
r
te
s
t)
3.
Gu
n
s
h
o
t (
1
3
4
in
s
ta
n
ce
s
f
o
r
tr
ain
i
n
g
,
5
3
in
s
tan
ce
s
f
o
r
test
)
T
h
e
s
tatis
tical
d
ata
o
f
t
h
e
s
o
u
n
d
ev
en
t d
ev
elo
p
m
e
n
t d
ataset
s
i
s
as s
h
o
w
n
i
n
T
ab
le
1
.
T
ab
le
1.
Featu
r
ed
d
ataset
s
tatis
tics
C
l
a
ss
B
a
b
y
C
r
y
G
l
a
ss B
r
e
a
k
G
u
n
S
h
o
t
U
sag
e
T
r
a
i
n
T
e
st
T
r
a
i
n
T
e
st
T
r
a
i
n
T
e
st
M
e
a
n
2
.
4
1
1
.
8
5
1
.
3
6
0
.
7
2
1
.
4
3
1
.
0
4
M
a
x
5
.
1
4
.
2
4
4
.
5
4
0
.
1
8
2
4
.
4
3
.
6
8
M
i
n
0
.
6
6
0
.
7
8
0
.
2
6
0
.
3
0
.
2
4
0
.
3
M
e
d
i
a
n
2
.
3
3
1
.
6
7
1
.
2
9
0
.
7
1
.
2
1
0
.
7
6
S
t
a
n
d
a
r
d
D
e
v
i
a
t
i
o
n
0
.
9
8
0
.
8
3
0
.
7
5
0
.
3
0
.
8
6
0
.
8
3
T
h
e
len
g
t
h
o
f
t
h
e
tar
g
et
s
ce
n
a
r
io
is
g
en
er
all
y
b
r
ief
r
elat
iv
e
t
o
th
e
3
0
-
s
ec
o
n
d
b
ac
k
g
r
o
u
n
d
n
o
is
e
[
1
3
]
.
T
r
ain
in
g
an
d
test
i
n
g
d
ata
s
ets
ar
e
p
r
o
d
u
ce
d
r
esp
ec
tiv
el
y
.
C
o
m
b
i
n
atio
n
s
ar
e
p
r
o
d
u
ce
d
b
y
s
o
f
t
w
ar
e
p
ac
k
ag
e
w
it
h
s
o
u
r
ce
in
f
o
r
m
atio
n
,
in
cl
u
d
in
g
a
m
b
ien
t
n
o
is
e
a
n
d
s
p
ec
if
i
ed
o
cc
u
r
r
en
ce
s
[
1
2
]
.
2.
3
.
Sed e
v
a
lua
t
io
n
m
et
ric
T
h
e
o
f
f
icial
m
etr
ic
i
n
t
h
e
D
C
ASE2
0
1
7
C
h
allen
g
e
t
h
at
i
s
E
R
is
t
h
e
s
u
m
o
f
in
s
er
tio
n
,
d
el
etio
n
an
d
s
u
b
s
t
itu
t
io
n
r
ates
[
7
,
1
0
]
.
E
R
is
ex
p
lai
n
ed
th
o
r
o
u
g
h
l
y
i
n
[
10
]
.
T
h
e
ev
alu
atio
n
i
n
v
o
lv
e
s
ca
lcu
latio
n
o
f
tr
u
e
p
o
s
itiv
e
(
T
P
)
,
f
alse
p
o
s
iti
v
e
(
FP
)
,
an
d
f
alse
n
e
g
ati
v
e
s
(
FN)
.
I
f
th
e
o
u
tp
u
t
o
f
th
e
s
y
s
te
m
c
o
r
r
ec
tly
p
r
ed
icts
t
h
e
p
r
esen
ce
an
d
o
n
s
et
o
f
a
ca
s
e,
it
w
il
l
b
e
ca
lcu
lated
as
T
P
.
On
l
y
w
h
e
n
it
is
p
r
ed
icted
w
it
h
i
n
5
0
0
m
illi
s
ec
o
n
d
s
(
m
s
)
o
f
th
e
ac
t
u
al
s
tar
ti
n
g
ti
m
e
is
d
etec
tio
n
as
tr
u
e
[
19
]
.
F
P
d
em
o
n
s
tr
ate
s
th
at
i
f
th
er
e
is
n
o
ev
en
t,
th
e
s
y
s
te
m
w
r
o
n
g
l
y
d
etec
t
s
th
e
e
x
is
te
n
ce
o
f
an
e
v
en
t.
I
f
t
h
e
o
u
tp
u
t
o
f
th
e
s
tr
ateg
y
m
is
s
es
t
h
e
e
v
e
n
t,
an
FN
w
il
l b
e
co
n
s
id
er
ed
.
T
h
ese
m
etr
ic
s
ar
e
u
s
ed
in
t
h
e
f
in
a
l
p
h
ase
to
ca
lcu
late
t
h
e
er
r
o
r
r
a
te
an
d
th
e
F
-
s
co
r
e.
=
+
(
1
)
=
2
+
(
2
)
W
h
er
e
N
in
d
icate
s
th
e
to
tal
a
m
o
u
n
t
o
f
s
a
m
p
les
i
n
t
h
e
as
s
es
s
m
e
n
t
d
atase
t
an
d
w
h
er
e
P
an
d
R
d
en
o
te
ac
cu
r
ac
y
a
n
d
r
ec
all
as d
escr
ib
ed
b
elo
w
.
=
+
(
3
)
=
2
+
(
4
)
3.
P
AST
AP
P
RO
ACH
E
S R
E
S
UL
T
S
T
r
a
d
itio
n
all
y
SED
p
o
p
u
lar
ap
p
r
o
ac
h
es
w
er
e
Ga
u
s
s
ian
Mix
t
u
r
e
Mo
d
el(
GM
M)
an
d
Hid
d
e
n
Ma
r
k
o
v
Mo
d
el(
HM
M)
b
u
t
h
a
s
m
o
v
ed
to
th
e
n
e
w
d
ee
p
lear
n
i
n
g
-
b
a
s
ed
m
et
h
o
d
s
[
7
,
2
0
]
.
R
ec
en
t
a
p
p
r
o
ac
h
u
s
es
d
ee
p
lear
n
in
g
-
b
ased
m
et
h
o
d
s
t
h
at
a
r
e
p
o
p
u
lar
l
y
u
s
ed
co
n
s
i
s
t
o
f
Dee
p
Neu
r
al
Net
w
o
r
k
(
DNN)
,
R
ec
u
r
r
en
t
Ne
u
r
al
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
R
ev
iew
o
f a
n
o
ma
lo
u
s
s
o
u
n
d
e
ve
n
t d
etec
tio
n
a
p
p
r
o
a
c
h
es
(
A
mir
u
l S
a
d
ikin
Md
A
fen
d
i
)
267
Net
w
o
r
k
(
R
N
N)
,
C
o
n
v
o
lu
tio
n
al
R
ec
u
r
r
en
t
Ne
u
r
al
Net
w
o
r
k
(
C
R
NN)
[
21
]
.
T
ab
le
2
s
h
o
w
s
p
ast
ap
p
r
o
ac
h
es
b
ased
o
n
a
co
llectio
n
o
f
t
h
e
ex
p
er
i
m
e
n
ts
co
n
d
u
cted
o
n
t
h
e
T
U
T
R
ar
e
So
u
n
d
E
v
en
t
2
0
1
7
Data
s
et
is
co
m
p
ar
ed
b
ased
o
n
th
eir
r
esp
ec
tiv
e
ac
cu
r
ac
y
,
f
ea
tu
r
es,
cla
s
s
i
f
ier
a
n
d
d
ec
is
io
n
m
a
k
i
n
g
s
p
ec
if
ic
atio
n
s
.
T
h
e
co
lu
m
n
f
ea
t
u
r
es
ar
e
t
h
e
c
h
o
ice
o
f
f
ea
t
u
r
es
e
x
tr
ac
tio
n
m
eth
o
d
t
h
at
will
b
e
f
ed
in
to
th
e
clas
s
i
f
ier
e
n
g
in
e.
T
h
e
C
la
s
s
i
f
ier
co
lu
m
n
d
e
n
o
tes
t
h
e
m
et
h
o
d
o
f
d
etec
tio
n
e
m
p
lo
y
ed
in
th
e
s
y
s
te
m
.
T
h
e
co
lu
m
n
d
ec
is
io
n
m
ak
in
g
r
e
f
er
s
to
t
h
e
f
i
n
al
v
er
if
icat
io
n
s
tep
in
th
e
SED
P
ip
elin
e.
A
d
d
itio
n
all
y
,
F1
m
u
lticla
s
s
i
n
co
n
s
is
ten
c
y
s
h
o
w
th
e
d
ev
iatio
n
o
f
ea
ch
clas
s
p
er
f
o
r
m
an
ce
.
T
h
e
ap
p
r
o
ac
h
b
est
w
as
t
h
e
co
m
b
i
n
atio
n
o
f
a
1
Di
m
e
n
s
io
n
al
C
NN
an
d
R
NN
w
i
th
lo
n
g
s
h
o
r
t te
r
m
m
e
m
o
r
y
u
n
it
s
(
L
ST
M)
w
it
h
th
e
a
v
er
ag
e
ac
c
u
r
ac
y
o
f
9
3
.
1
%
[
19
].
T
ab
le
2.
R
esu
lts
o
f
p
ast
ap
p
r
o
ac
h
es o
f
SED
co
n
d
u
cted
o
n
T
UT
R
ar
e
SED
2
0
1
7
R
e
se
a
r
c
h
e
r
s
O
v
e
r
a
l
l
F
1
M
u
l
t
i
-
C
l
a
ss
I
n
c
o
n
si
st
e
n
c
y
F
e
a
t
u
r
e
s
C
l
a
ssi
f
i
e
r
D
e
c
i
si
o
n
M
a
k
i
n
g
ER
F1
[
1
9
]
0
.
1
3
0
7
9
3
.
1
4
.
1
l
o
g
-
m
e
l
e
n
e
r
g
i
e
s
C
R
N
N
-
L
S
T
M
t
h
r
e
sh
o
l
d
i
n
g
[
7
]
0
.
1
7
3
3
91
3
.
7
sp
e
c
t
r
o
g
r
a
m
C
N
N
maj
o
r
i
t
y
v
o
t
e
[
3
]
0
.
2
7
7
3
8
5
.
3
3
.
6
l
o
g
-
m
e
l
e
n
e
r
g
i
e
s
C
R
N
N
me
d
i
a
n
f
i
l
t
e
r
i
n
g
e
n
se
mb
l
e
[
2
4
]
0
.
3
1
3
3
8
4
.
2
1
0
.
6
l
o
g
-
m
e
l
e
n
e
r
g
i
e
s
M
L
P
C
N
N
t
h
e
sh
o
l
d
[
2
2
]
0
.
3
1
7
3
82
0
.
9
l
o
g
G
a
mm
a
t
o
n
e
c
e
p
s
t
r
a
l
c
o
e
f
f
i
c
i
e
n
t
s
t
a
i
l
o
r
e
d
-
l
o
ss
D
N
N
+
C
N
N
me
d
i
a
n
f
i
l
t
e
r
i
n
g
[
2
1
]
0
.
3
2
6
7
8
3
.
9
1
3
.
0
l
o
g
-
m
e
l
e
n
e
r
g
i
e
s
C
R
N
N
maj
o
r
i
t
y
v
o
t
e
[
1
1
]
0
.
4
1
0
7
7
9
.
1
1
0
.
2
l
o
g
-
m
e
l
S
p
e
c
t
r
o
g
r
a
msM
F
C
C
M
L
P
C
N
N
R
N
N
me
d
i
a
n
f
i
l
t
e
r
i
n
g
e
n
se
mb
l
i
n
g
h
a
r
d
T
h
r
e
sh
o
l
d
i
n
g
[
2
5
]
0
.
4
2
6
7
7
8
.
6
8
.
3
M
F
C
C
Z
C
R
e
n
e
r
g
y
sp
e
c
t
r
a
l
c
e
n
t
r
o
i
d
p
i
t
c
h
e
n
se
mb
l
e
t
h
r
e
sh
o
l
d
i
n
g
[
1
2
]
0
.
4
3
2
7
3
.
4
2
0
.
0
l
o
g
-
m
e
l
e
n
e
r
g
i
e
s
DNN
me
d
i
a
n
f
i
l
t
e
r
i
n
g
[
5
]
0
.
5
7
4
.
2
1
7
.
9
sp
e
c
t
r
o
g
r
a
m
N
M
F
mo
v
i
n
g
a
v
e
r
a
g
e
f
i
l
t
e
r
[
7
]
0
.
6
6
9
.
8
1
1
.
6
D
N
N
(
M
F
C
C
)
Bi
-
L
S
T
M
t
o
p
o
u
t
p
u
t
p
r
o
b
a
b
i
l
i
t
y
[
6
]
0
.
6
3
7
3
6
4
.
1
1
4
.
2
l
o
g
-
m
e
l
e
n
e
r
g
i
e
s
M
L
P
me
d
i
a
n
f
i
l
t
e
r
i
n
g
[
2
3
]
0
.
6
7
7
3
6
5
.
8
1
6
.
5
l
o
g
-
m
e
l
e
n
e
r
g
i
e
s fr
o
m N
M
F
so
u
r
c
e
se
p
a
r
a
t
i
o
n
M
L
P
me
d
i
a
n
f
i
l
t
e
r
i
n
g
Fu
r
t
h
er
m
o
r
e
w
e
ca
n
o
b
s
er
v
e
t
h
e
s
i
n
g
le
cla
s
s
ac
c
u
r
ac
y
d
ev
i
atio
n
b
elo
w
i
n
T
ab
le
3
.
A
s
y
o
u
ca
n
s
ee
th
e
d
i
f
f
er
en
ce
s
b
et
w
ee
n
ea
ch
class
p
er
f
o
r
m
an
ce
s
h
o
w
s
i
m
b
alan
ce
t
h
at
co
u
ld
b
e
r
elate
d
to
th
e
ap
p
r
o
ac
h
as
a
n
ad
v
an
ta
g
e.
T
h
e
d
ev
iatio
n
o
f
an
ap
p
r
o
ac
h
w
ill
s
h
o
w
it
s
s
in
g
le
cla
s
s
p
er
f
o
r
m
an
ce
co
u
ld
b
e
em
p
lo
y
ed
to
s
p
ec
if
icall
y
tar
g
eted
e
v
en
t.
T
ab
le
3
s
h
o
w
s
t
h
e
m
u
lti
-
clas
s
i
n
co
n
s
is
ten
c
y
i
n
w
h
ic
h
i
s
t
h
e
d
e
v
iatio
n
o
f
th
e
t
h
r
ee
tar
g
et
clas
s
es
d
o
n
e
in
p
r
ev
io
u
s
ex
p
er
i
m
e
n
ts
.
T
ab
le
3
.
Mu
lti
-
clas
s
i
n
co
n
s
i
s
te
n
c
y
R
e
se
a
r
c
h
e
r
s
B
a
b
y
c
r
y
G
l
a
ss b
r
e
a
k
G
u
n
sh
o
t
F
1
M
u
l
t
i
-
C
l
a
ss I
n
c
o
n
si
s
t
e
n
c
y
ER
F1
ER
F1
ER
F1
[
1
9
]
0
.
1
5
2
9
2
.
2
0
.
0
4
8
9
7
.
6
0
.
1
9
2
8
9
.
6
4
.
1
[
7
]
0
.
1
8
4
9
0
.
8
0
.
1
0
4
9
4
.
7
0
.
2
3
2
8
7
.
4
3
.
7
[
3
]
0
.
2
8
4
8
5
.
7
0
.
2
2
8
8
.
8
0
.
3
2
8
8
1
.
6
3
.
6
[
2
2
]
0
.
1
7
2
9
1
.
4
0
.
2
2
8
9
.
1
0
.
5
4
8
72
1
0
.
6
[
1
1
]
0
.
3
5
6
83
0
.
3
1
2
8
4
.
7
0
.
3
1
2
84
0
.
9
[
2
1
]
0
.
2
6
4
8
7
.
3
0
.
1
6
9
1
.
5
0
.
5
2
8
6
7
.
2
1
3
.
0
[
2
5
]
0
.
4
4
8
0
.
6
0
.
2
2
8
8
8
.
5
0
.
5
6
4
6
8
.
2
1
0
.
2
[
2
4
]
0
.
5
7
5
.
9
0
.
2
3
6
8
7
.
8
0
.
5
4
4
7
1
.
9
8
.
3
[
5
]
0
.
4
0
8
7
8
.
8
0
.
1
6
4
9
1
.
5
0
.
9
2
8
5
2
.
3
2
0
.
0
[
1
2
]
0
.
4
4
7
7
.
3
0
.
2
1
2
8
9
.
1
0
.
6
4
4
5
3
.
9
1
7
.
9
[
2
]
0
.
7
8
6
7
.
4
0
.
3
2
4
8
2
.
4
0
.
6
9
6
5
9
.
5
1
1
.
6
[
2
3
]
0
.
8
8
4
6
5
.
3
0
.
3
9
6
8
0
.
2
0
.
7
5
2
5
1
.
8
1
4
.
2
[
6
]
0
.
8
0
4
6
6
.
8
0
.
3
8
7
9
.
1
0
.
7
2
8
4
6
.
5
1
6
.
5
C
o
m
p
ar
is
o
n
o
f
th
e
e
x
p
er
i
m
e
n
t
r
esu
lts
t
h
at
w
as
d
o
n
e
o
n
a
lev
eled
d
ataset.
T
h
e
d
ev
iatio
n
o
f
s
o
m
e
ap
p
r
o
ac
h
es
s
u
g
g
est
t
h
e
m
et
h
o
d
s
co
u
ld
b
e
e
m
p
lo
y
ed
as
a
s
in
g
le
clas
s
e
f
f
icien
t.
T
h
e
co
m
b
i
n
atio
n
o
f
s
p
ec
tr
o
g
r
a
m
,
NM
F
an
d
m
o
v
in
g
a
v
er
ag
e
f
ilter
[
5
]
s
h
o
w
s
t
h
e
h
ig
h
e
s
t
d
ev
ia
tio
n
a
m
o
n
g
ea
c
h
class
p
er
f
o
r
m
an
ce
.
Hig
h
d
ev
iatio
n
m
a
y
i
n
d
icate
a
s
p
ec
if
ic
class
o
r
p
atter
n
o
v
er
f
itti
n
g
.
T
h
e
r
ev
ie
w
b
eliev
es
t
h
at
lo
w
er
d
ev
iatio
n
s
h
o
u
ld
co
n
clu
d
e
a
r
o
b
u
s
t
ap
p
r
o
ac
h
to
w
ar
d
s
lar
g
e
class
v
ar
ia
tio
n
s
d
etec
tio
n
in
t
h
e
f
u
tu
r
e.
L
ar
g
e
v
ar
ie
t
y
is
t
h
e
p
r
o
b
lem
h
u
m
a
n
s
ar
e
ab
le
to
c
o
m
p
u
te.
T
h
is
d
ev
iatio
n
co
u
ld
b
e
f
u
r
th
er
r
esear
ch
ed
to
f
u
l
l
y
u
n
d
er
s
ta
n
d
th
e
tr
en
d
an
d
r
ea
s
o
n
s
b
eh
in
d
s
u
ch
a
p
h
e
n
o
m
e
n
o
n
.
C
o
m
p
ar
i
s
o
n
b
et
w
ee
n
SE
D
co
m
p
o
n
e
n
t
s
co
m
b
i
n
ati
o
n
s
u
g
g
e
s
t p
o
p
u
la
r
s
p
ec
if
icatio
n
s
f
o
r
f
u
tu
r
e
r
esear
ch
an
d
d
ev
elo
p
m
e
n
t.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8938
IJ
-
A
I
Vo
l.
8
,
No
.
3
,
Sep
tem
b
er
201
9
:
2
6
4
–
2
6
9
268
4.
CO
NCLU
SI
O
N
T
h
e
s
tu
d
y
co
m
p
ar
es
p
r
ev
io
u
s
w
o
r
k
o
n
a
n
o
m
alo
u
s
SE
D
s
o
l
u
tio
n
s
o
b
s
er
v
ed
m
ai
n
l
y
o
n
t
h
e
co
m
p
o
s
i
tio
n
o
f
t
h
e
t
y
p
ica
l
S
E
D
p
ip
elin
e
co
m
p
o
n
en
ts
.
T
h
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
s
y
s
te
m
w
er
e
ev
a
lu
ated
o
n
a
s
in
g
le
a
n
d
m
u
lti
-
clas
s
b
asi
s
.
I
t
is
f
o
u
n
d
t
h
at
th
e
p
er
f
o
r
m
a
n
c
e
o
n
s
in
g
le
clas
s
te
n
d
to
b
e
i
m
b
a
lan
ce
d
a
n
d
v
ar
y
b
et
w
ee
n
ap
p
r
o
ac
h
es.
T
h
e
lo
w
d
ev
iatio
n
i
n
s
in
g
le
clas
s
p
er
f
o
r
m
a
n
ce
s
h
o
u
ld
i
n
d
icate
a
v
er
s
atile
ap
p
r
o
ac
h
t
h
at
is
n
o
t b
iased
o
n
a
ce
r
tain
t
y
p
e
o
f
ev
e
n
t.
T
h
u
s
t
h
e
ap
p
r
o
ac
h
is
b
eliev
ed
to
b
e
w
e
ll o
p
ti
m
ized
f
o
r
a
w
id
er
v
ar
iet
y
o
f
SED
p
r
o
b
lem
s
.
I
n
s
ea
r
ch
f
o
r
a
r
o
b
u
s
t
SED
p
ip
elin
e
co
m
p
o
n
e
n
t
s
co
m
p
o
s
itio
n
o
p
tim
ized
f
o
r
th
e
b
est.
I
t
is
s
u
g
g
ested
f
o
r
f
u
r
th
er
s
t
u
d
ies
to
e
m
p
lo
y
t
h
e
co
m
b
in
atio
n
o
f
s
y
s
te
m
c
h
ar
ac
ter
is
ti
cs
o
f
s
u
c
h
f
ea
t
u
r
e
ex
tr
ac
tio
n
,
cla
s
s
i
f
ier
s
an
d
d
ec
is
io
n
m
a
k
i
n
g
s
tep
s
in
th
e
p
ip
elin
e.
T
h
e
m
o
s
t
v
iab
le
ap
p
r
o
ac
h
i
n
t
h
is
r
ev
ie
w
is
th
at
a
co
m
b
in
atio
n
o
f
u
s
i
n
g
f
ea
tu
r
es
as
lo
g
-
m
el
en
er
g
ie
s
in
co
n
v
o
lu
tio
n
al
r
ec
u
r
r
en
t
n
e
u
r
al
n
et
w
o
r
k
(
C
R
NN)
w
it
h
lo
n
g
s
h
o
r
t
ter
m
m
e
m
o
r
y
(
L
ST
M)
w
i
th
a
v
er
i
f
icatio
n
s
t
ep
o
f
th
r
e
s
h
o
ld
i
n
g
h
a
s
o
b
tain
ed
9
3
.
1
%
F1
s
co
r
e
an
d
0
.
1
3
0
7
E
R
.
Fu
t
u
r
e
w
o
r
k
is
s
u
g
g
es
ted
to
b
e
r
ev
o
lv
in
g
ar
o
u
n
d
th
e
s
e
s
p
ec
i
f
icatio
n
s
m
ak
in
g
u
p
th
e
e
n
tire
p
ip
elin
e
to
p
r
o
d
u
ce
a
r
eliab
le
p
r
o
d
u
ct
f
o
r
r
ea
l
-
w
o
r
ld
ap
p
licatio
n
.
ACK
NO
WL
E
D
G
E
M
E
NT
S
Un
i
v
er
s
iti
T
ek
n
o
lo
g
i
M
AR
A
a
f
o
r
th
e
g
r
a
n
t
o
f
6
0
0
-
I
R
MI
/P
E
R
D
A
N
A
5
/3
B
E
ST
A
R
I
(
0
9
6
/2
0
1
8
)
as
w
ell
as
Fac
u
lt
y
o
f
C
o
m
p
u
ter
a
n
d
Ma
th
e
m
atica
l
Scie
n
ce
s
,
Un
iv
er
s
iti
T
ek
n
o
lo
g
i
MA
R
A
,
S
h
ah
Ala
m
,
Ma
la
y
s
ia
f
o
r
p
r
o
v
id
in
g
es
s
en
tial s
u
p
p
o
r
t a
n
d
k
n
o
w
led
g
e
f
o
r
th
e
w
o
r
k
.
RE
F
E
R
E
NC
E
S
[1
]
Bo
d
d
a
p
a
ti
,
V
.
,
P
e
tef
,
A
.
,
Ra
s
m
u
ss
o
n
,
J.,
&
L
u
n
d
b
e
rg
,
L
.
(2
0
1
7
).
Clas
sify
in
g
e
n
v
iro
n
m
e
n
tal
so
u
n
d
s
u
sin
g
im
a
g
e
re
c
o
g
n
it
io
n
n
e
tw
o
rk
s.
Pro
c
e
d
ia
C
o
mp
u
ter
S
c
ien
c
e
,
1
1
2
,
2
0
4
8
–
2
0
5
6
.
h
tt
p
s://
d
o
i
.
o
rg
/1
0
.
1
0
1
6
/j
.
p
r
o
c
s.2
0
1
7
.
0
8
.
2
5
0
.
[2
]
L
i,
Y.,
&
L
i,
X
.
(2
0
1
7
).
T
h
e
{S
EIE
-
S
CUT
}
S
y
ste
m
s
f
o
r
{IEE
E}
{AA
S
P
}
Ch
a
ll
e
n
g
e
o
n
{DCA
S
E}
2
0
1
7
:
De
e
p
L
e
a
rn
in
g
Tec
h
n
iq
u
e
s f
o
r
A
u
d
io
R
e
p
re
se
n
tatio
n
a
n
d
Clas
sif
ica
ti
o
n
.
[3
]
P
h
a
n
,
H.,
K
o
c
h
,
P
.
,
Ka
tzb
e
rg
,
F
.
,
M
a
a
ss
,
M
.
,
M
a
z
u
r,
R.
,
M
c
L
o
u
g
h
l
in
,
I.
,
&
M
e
rti
n
s,
A
.
(2
0
1
7
).
W
h
a
t
m
a
k
e
s
a
u
d
io
e
v
e
n
t
d
e
tec
ti
o
n
h
a
rd
e
r
t
h
a
n
c
las
s
if
ica
ti
o
n
?
2
5
t
h
E
u
ro
p
e
a
n
S
ig
n
a
l
P
r
o
c
e
ss
in
g
Co
n
f
e
re
n
c
e
,
EUS
I
P
C
O
2
0
1
7
,
2
0
1
7
-
Ja
n
u
a
ry
,
2
7
3
9
–
2
7
4
3
.
h
tt
p
s://
d
o
i.
o
rg
/1
0
.
2
3
9
1
9
/E
USI
P
CO.
2
0
1
7
.
8
0
8
1
7
0
9
[4
]
Yu
so
f
f
,
M
&
M
.
Afe
n
d
i
A
.
S
.
(2
0
1
9
).
A
c
o
u
stic
S
u
rv
e
il
lan
c
e
In
tru
sio
n
De
tec
ti
o
n
w
it
h
L
in
e
a
r
P
re
d
ictiv
e
Co
d
in
g
a
n
d
Ra
n
d
o
m
F
o
re
st:
4
th
I
n
tern
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
,
S
CDS
2
0
1
8
,
Ba
n
g
k
o
k
,
T
h
a
il
a
n
d
,
A
u
g
u
st
15
-
1
6
,
2
0
1
8
,
P
r
o
c
e
e
d
in
g
s.
1
0
.
1
0
0
7
/
9
7
8
-
9
8
1
-
13
-
3
4
4
1
-
2
_
6
.
[5
]
G
h
a
ff
a
rz
a
d
e
g
a
n
,
S
.
,
S
a
lek
in
,
A
.
,
Da
s,
S
.
,
&
F
e
n
g
,
Z
.
(2
0
1
7
).
B
o
sc
h
Ra
re
S
o
u
n
d
Ev
e
n
ts
De
tec
ti
o
n
S
y
ste
m
s
f
o
r
{DCA
S
E2
0
1
7
}
C
h
a
ll
e
n
g
e
.
[6
]
He
it
to
la,
T
.
,
&
M
e
sa
ro
s,
A
.
(2
0
1
7
).
{DCA
S
E}
2
0
1
7
C
h
a
ll
e
n
g
e
S
e
tu
p
:
T
a
sk
s,
Da
tas
e
ts an
d
Ba
se
li
n
e
S
y
st
e
m
.
[7
]
Ca
k
ir,
E.
,
&
V
irt
a
n
e
n
,
T
.
(2
0
1
7
)
.
Co
n
v
o
l
u
ti
o
n
a
l
Re
c
u
rre
n
t
Ne
u
ra
l
Ne
tw
o
rk
s f
o
r
Ra
re
S
o
u
n
d
Ev
e
n
t
De
tec
ti
o
n
.
[8
]
Cro
c
c
o
,
M
.
,
Cristan
i,
M
.
,
T
ru
c
c
o
,
A
.
,
&
M
u
rin
o
,
V.
(2
0
1
4
).
A
u
d
io
S
u
rv
e
il
lan
c
e
:
a
S
y
ste
m
a
ti
c
Re
v
ie
w
,
(O
c
to
b
e
r).
Re
tri
e
v
e
d
f
ro
m
h
tt
p
:/
/arx
iv
.
o
rg
/ab
s/1
4
0
9
.
7
7
8
7
[9
]
Oz
e
r,
I.
,
Oz
e
r,
Z.
,
&
F
in
d
ik
,
O.
(
2
0
1
8
).
No
ise
ro
b
u
st
so
u
n
d
e
v
e
n
t
c
l
a
ss
if
i
c
a
ti
o
n
w
it
h
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
tw
o
rk
.
Ne
u
ro
c
o
mp
u
ti
n
g
,
2
7
2
,
5
0
5
–
5
1
2
.
h
tt
p
s:/
/d
o
i.
o
rg
/1
0
.
1
0
1
6
/j
.
n
e
u
c
o
m
.
2
0
1
7
.
0
7
.
0
2
1
[1
0
]
Ja
y
a
la
k
sh
m
i,
S
.
L
.
,
Ch
a
n
d
ra
k
a
la,
S
.
,
&
Ne
d
u
n
c
h
e
li
a
n
,
R.
(2
0
1
8
).
G
lo
b
a
l
sta
ti
stica
l
f
e
a
tu
re
s
-
b
a
se
d
a
p
p
ro
a
c
h
f
o
r
A
c
o
u
stic
Ev
e
n
t
D
e
tec
ti
o
n
.
A
p
p
li
e
d
A
c
o
u
stics
,
1
3
9
(A
p
ril
),
1
1
3
–
1
1
8
.
h
tt
p
s:/
/d
o
i.
o
rg
/1
0
.
1
0
1
6
/j
.
a
p
a
c
o
u
st
.
2
0
1
8
.
0
4
.
0
2
6
[1
1
]
V
e
sp
e
rin
i
,
F
.
,
Dro
g
h
i
n
i,
D.,
F
e
rre
tt
i,
D.,
P
rin
c
i
p
i,
E.
,
G
a
b
rielli
,
L
.
,
S
q
u
a
rti
n
i,
S
.
,
&
P
iaz
z
a
,
F
.
(2
0
1
7
)
.
A
Hie
ra
r
c
h
ic
M
u
lt
i
-
S
c
a
led
A
p
p
ro
a
c
h
f
o
r
Ra
re
S
o
u
n
d
Ev
e
n
t
De
tec
ti
o
n
.
[1
2
]
W
a
n
g
,
Ju
n
,
&
L
i,
S
.
(2
0
1
7
).
M
u
lt
i
-
F
ra
m
e
Co
n
c
a
ten
a
ti
o
n
f
o
r
De
tec
ti
o
n
o
f
Ra
re
S
o
u
n
d
Ev
e
n
ts
B
a
se
d
o
n
De
e
p
Ne
u
ra
l
Ne
tw
o
rk
.
[1
3
]
M
e
sa
ro
s,
A
.
,
He
it
to
la,
T
.
,
&
Virtan
e
n
,
T
.
(2
0
1
6
).
M
e
tri
c
s
f
o
r
P
o
ly
p
h
o
n
ic
S
o
u
n
d
Ev
e
n
t
De
tec
ti
o
n
.
A
p
p
l
ied
S
c
ien
c
e
s,
6
(6
),
1
6
2
.
h
tt
p
s://
d
o
i
.
o
r
g
/1
0
.
3
3
9
0
/ap
p
6
0
6
0
1
6
2
[1
4
]
Be
z
o
u
i,
M
.
(
2
0
1
9
).
S
p
e
e
c
h
Re
c
o
g
n
it
io
n
o
f
M
o
ro
c
c
a
n
Dia
lec
t
Us
i
n
g
Hid
d
e
n
M
a
rk
o
v
M
o
d
e
ls.
IAE
S
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Arti
fi
c
i
a
l
I
n
telli
g
e
n
c
e
(
IJ
-
AI)
,
8
(1
),
7
.
h
tt
p
s:/
/d
o
i.
o
rg
/1
0
.
1
1
5
9
1
/i
jai.
v
8
.
i
1
.
p
p
7
-
13
[1
5
]
W
a
n
g
,
Ji
a
n
f
e
i,
Zh
a
n
g
,
W
.
,
&
L
i
u
,
J.
(2
0
1
7
).
T
ra
n
sf
e
r
L
e
a
rn
in
g
Ba
se
d
{DN
N}
-
{HMM
}
H
y
b
rid
S
y
st
e
m
f
o
r
Ra
r
e
S
o
u
n
d
Ev
e
n
t
De
tec
ti
o
n
.
[1
6
]
P
h
a
n
,
H.,
Kra
w
c
z
y
k
-
Be
c
k
e
r,
M
.
,
G
e
rk
m
a
n
n
,
T
.
,
&
M
e
rti
n
s,
A
.
(2
0
1
7
).
{DN
N}
a
n
d
{CNN
}
w
it
h
W
e
ig
h
ted
a
n
d
M
u
lt
i
-
T
a
sk
L
o
ss
F
u
n
c
ti
o
n
s f
o
r
Au
d
i
o
Ev
e
n
t
De
tec
ti
o
n
.
[1
7
]
Ka
m
a
la,
R.
,
&
T
h
a
n
g
a
iah
,
R.
J.
(2
0
1
9
).
A
n
Im
p
ro
v
e
d
Hy
b
rid
F
e
a
tu
re
S
e
lec
ti
o
n
M
e
th
o
d
f
o
r
Hu
g
e
Dim
e
n
sio
n
a
l
Da
tas
e
ts.
IAE
S
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Ar
ti
fi
c
i
a
l
In
telli
g
e
n
c
e
(
IJ
-
AI)
,
8
(1
),
7
7
.
h
tt
p
s:/
/d
o
i.
o
rg
/1
0
.
1
1
5
9
1
/i
jai
.
v
8
.
i1
.
p
p
7
7
-
86
[1
8
]
Y.
Ho
sh
e
n
,
R
.
J.
W
e
iss
a
n
d
K.
W
.
W
il
so
n
,
"
S
p
e
e
c
h
a
c
o
u
stic
m
o
d
e
li
n
g
f
ro
m
ra
w
m
u
lt
ich
a
n
n
e
l
w
a
v
e
f
o
r
m
s,"
2
0
1
5
IEE
E
In
tern
a
ti
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
A
c
o
u
stics
,
S
p
e
e
c
h
a
n
d
S
ig
n
a
l
P
ro
c
e
ss
in
g
(ICA
S
S
P
),
Brisb
a
n
e
,
Q
L
D,
2
0
1
5
,
p
p
.
4
6
2
4
-
4
6
2
8
.
h
tt
p
s://
d
o
i:
1
0
.
1
1
0
9
/I
CA
S
S
P
.
2
0
1
5
.
7
1
7
8
8
4
7
[1
9
]
L
i
m
,
H.,
P
a
rk
,
J.,
&
Ha
n
,
Y.
(2
0
1
7
).
Ra
re
S
o
u
n
d
Ev
e
n
t
De
tec
ti
o
n
Us
in
g
{1
D}
C
o
n
v
o
l
u
ti
o
n
a
l
Re
c
u
rre
n
t
Ne
u
ra
l
Ne
tw
o
rk
s.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
AI
I
SS
N:
2252
-
8938
R
ev
iew
o
f a
n
o
ma
lo
u
s
s
o
u
n
d
e
ve
n
t d
etec
tio
n
a
p
p
r
o
a
c
h
es
(
A
mir
u
l S
a
d
ikin
Md
A
fen
d
i
)
269
[2
0
]
S
a
k
,
H.,
S
e
n
io
r
,
A
.
,
Ra
o
,
K.,
&
Be
a
u
f
a
y
s,
F
.
(2
0
1
5
).
F
a
st
a
n
d
a
c
c
u
ra
te
re
c
u
rre
n
t
n
e
u
ra
l
n
e
tw
o
rk
a
c
o
u
stic
m
o
d
e
ls
f
o
r
sp
e
e
c
h
re
c
o
g
n
it
io
n
.
P
r
o
c
e
e
d
in
g
s
o
f
th
e
A
n
n
u
a
l
C
o
n
f
e
re
n
c
e
o
f
th
e
In
ter
n
a
ti
o
n
a
l
S
p
e
e
c
h
C
o
m
m
u
n
ica
ti
o
n
A
s
so
c
iatio
n
,
INT
ERS
P
EE
CH
,
2
0
1
5
-
Ja
n
u
a
,
1
4
6
8
-
1
4
7
2
.
[2
1
]
Ka
iw
u
,
W
.
,
L
ip
in
g
,
Y.,
&
Bin
,
Y.
(2
0
1
7
).
A
u
d
io
Ev
e
n
ts
De
tec
ti
o
n
a
n
d
Clas
sif
ica
ti
o
n
Us
in
g
Ex
ten
d
e
d
{R
-
F
CN}
A
p
p
ro
a
c
h
.
[2
2
]
Zh
o
u
,
Q.,
&
F
e
n
g
,
Z.
(2
0
1
7
).
R
o
b
u
st
S
o
u
n
d
Ev
e
n
t
De
tec
ti
o
n
t
h
r
o
u
g
h
N
o
ise
Esti
m
a
ti
o
n
a
n
d
S
o
u
r
c
e
S
e
p
a
ra
ti
o
n
Us
in
g
{NMF
}.
[2
3
]
Je
o
n
,
K.
M
.
,
&
Kim
,
H.
K.
(
2
0
1
7
).
No
n
n
e
g
a
ti
v
e
M
a
tri
x
F
a
c
to
riza
ti
o
n
-
Ba
se
d
S
o
u
rc
e
S
e
p
a
ra
ti
o
n
w
it
h
On
li
n
e
N
o
ise
L
e
a
rn
in
g
f
o
r
De
te
c
ti
o
n
o
f
Ra
re
S
o
u
n
d
Ev
e
n
ts.
[2
4
]
Ra
v
ich
a
n
d
ra
n
,
A
.
,
&
Da
s,
S
.
(
2
0
1
7
).
Bo
sc
h
Ra
re
S
o
u
n
d
Ev
e
n
ts De
t
e
c
ti
o
n
S
y
ste
m
s
f
o
r
{DCA
S
E2
0
1
7
}
Ch
a
ll
e
n
g
e
.
[2
5
]
Da
n
g
,
A
.
,
V
u
,
T
.
,
&
W
a
n
g
,
J.
-
C.
(
2
0
1
7
).
De
e
p
L
e
a
rn
in
g
f
o
r
{DC
A
S
E2
0
1
7
}
C
h
a
ll
e
n
g
e
.
B
I
O
G
RAP
H
I
E
S
O
F
AUTH
O
RS
M
u
h
a
m
a
d
Am
iru
l
S
a
d
ik
in
M
D
Af
e
n
d
i
is
a
M
a
ste
r
o
f
S
c
ien
c
e
(In
f
o
r
m
a
ti
o
n
T
e
c
h
n
o
lo
g
y
)
stu
d
e
n
t
i
n
Un
iv
e
rsiti
T
e
c
h
n
o
l
o
g
i
M
A
R
A
(Ui
T
M
)
S
h
a
h
A
la
m
,
M
a
l
a
y
sia
.
S
in
c
e
2
0
1
8
h
a
s
b
e
g
u
n
re
se
a
rc
h
in
th
e
f
ield
o
f
a
rti
f
icia
l
in
telli
g
e
n
c
e
sp
e
c
if
ica
ll
y
o
n
a
u
d
io
p
r
o
c
e
ss
in
g
.
His
e
d
u
c
a
ti
o
n
b
a
c
k
g
ro
u
n
d
c
o
n
sist
o
f
a
Ba
c
h
e
lo
r
o
f
In
f
o
r
m
a
ti
o
n
tec
h
n
o
lo
g
y
(Ho
n
s)
In
tel
li
g
e
n
t
S
y
ste
m
En
g
in
e
e
rin
g
a
t
UiT
M
S
h
a
h
A
la
m
,
M
a
la
y
sia
b
y
th
e
y
e
a
r
2
0
1
9
.
He
is
c
u
rre
n
tl
y
p
u
rsu
in
g
re
se
a
rc
h
o
n
a
u
d
io
b
a
se
d
f
o
re
st
su
rv
e
il
lan
c
e
c
o
ll
a
b
o
ra
ti
n
g
w
it
h
W
il
d
li
f
e
Co
n
se
rv
a
ti
v
e
S
o
c
iet
y
M
a
la
y
sia
to
c
o
m
b
a
t
w
il
d
li
f
e
a
b
u
se
.
M
a
rin
a
Yu
so
f
f
is cu
rre
n
tl
y
a
De
p
u
ty
De
a
n
(Re
se
a
rc
h
a
n
d
In
d
u
stry
L
in
k
a
g
e
s) an
d
S
e
n
io
r
L
e
c
tu
re
r
o
f
F
a
c
u
lt
y
o
f
Co
m
p
u
ter
a
n
d
M
a
th
e
m
a
ti
c
a
l
S
c
ien
c
e
s,
Un
iv
e
rsiti
Tek
n
o
lo
g
i
M
A
R
A
S
h
a
h
A
la
m
,
M
a
la
y
sia
.
S
h
e
h
o
ld
s
a
P
h
D
in
In
telli
g
e
n
t
S
y
ste
m
f
ro
m
th
e
Un
iv
e
rsiti
Tek
n
o
lo
g
i
M
A
R
A
.
S
h
e
p
re
v
io
u
sly
w
o
rk
e
d
a
s
a
se
n
io
r
e
x
e
c
u
ti
v
e
o
f
In
f
o
r
m
a
ti
o
n
T
e
c
h
n
o
lo
g
y
in
S
IRIM
Be
rh
a
d
,
M
a
lay
sia
.
S
h
e
h
o
l
d
s
a
b
a
c
h
e
lo
r’s
d
e
g
re
e
in
c
o
m
p
u
ter
sc
ien
c
e
f
ro
m
th
e
Un
iv
e
rsity
o
f
S
c
ien
c
e
M
a
l
a
y
sia
,
a
n
d
M
a
ste
r
o
f
S
c
ien
c
e
in
In
f
o
rm
a
ti
o
n
T
e
c
h
n
o
l
o
g
y
f
ro
m
Un
iv
e
rsiti
T
e
k
n
o
lo
g
i
M
A
R
A
.
S
h
e
is
in
tere
ste
d
in
th
e
d
e
v
e
lo
p
m
e
n
t
o
f
in
telli
g
e
n
t
a
p
p
li
c
a
ti
o
n
,
m
o
d
ifi
c
a
ti
o
n
a
n
d
e
n
h
a
n
c
e
m
e
n
t
c
o
m
p
u
tatio
n
a
l
in
tel
li
g
e
n
c
e
tec
h
n
iq
u
e
s
in
c
lu
d
e
p
a
rti
c
le
sw
a
r
m
o
p
ti
m
isa
ti
o
n
,
n
e
u
ra
l
n
e
tw
o
rk
,
g
e
n
e
ti
c
a
lg
o
rit
h
m
,
a
n
d
a
n
t
c
o
lo
n
y
.
S
h
e
h
a
s
m
a
n
y
i
m
p
a
c
t
jo
u
rn
a
ls
p
u
b
li
c
a
ti
o
n
s
a
n
d
p
re
se
n
ted
h
e
r
re
se
a
rc
h
in
m
a
n
y
c
o
n
fe
re
n
c
e
s lo
c
a
ll
y
a
n
d
in
tern
a
ti
o
n
a
ll
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.