I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l En
g
ineering
a
nd
Co
m
pu
t
er
Science
Vo
l.
3
7
,
No
.
3
,
Ma
r
ch
20
2
5
,
p
p
.
1
85
6
~
1
86
4
I
SS
N:
2
502
-
4
7
52
,
DOI
: 1
0
.
1
1
5
9
1
/ijee
cs
.v
3
7
.
i
3
.
pp
1
85
6
-
1
86
4
1856
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs
.
ia
esco
r
e.
co
m
Co
nception o
f
spe
ech emotio
n
reco
g
nition metho
ds:
a
review
Abdelk
a
der
B
enzira
r
1
,
M
o
h
a
m
ed
H
a
m
idi
2
,
M
o
un
ce
f
F
ila
li B
o
ua
m
i
1
1
La
b
o
r
a
t
o
r
y
o
f
A
p
p
l
i
e
d
M
a
t
h
e
m
a
t
i
c
s
a
n
d
I
n
f
o
r
ma
t
i
o
n
S
y
st
e
ms,
M
u
l
t
i
d
i
s
c
i
p
l
i
n
a
r
y
F
a
c
u
l
t
y
o
f
N
a
d
o
r
,
M
o
h
a
mm
e
d
P
r
e
mi
e
r
U
n
i
v
e
r
s
i
t
y
,
O
u
j
d
a
,
M
o
r
o
c
c
o
2
Te
a
m
o
f
M
o
d
e
l
i
n
g
a
n
d
S
c
i
e
n
t
i
f
i
c
C
o
mp
u
t
i
n
g
,
M
u
l
t
i
d
i
s
c
i
p
l
i
n
a
r
y
F
a
c
u
l
t
y
o
f
N
a
d
o
r
,
M
o
h
a
mm
e
d
P
r
e
mi
e
r
U
n
i
v
e
r
si
t
y
,
O
u
j
d
a
,
M
o
r
o
c
c
o
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ap
r
8
,
2
0
2
4
R
ev
is
ed
Oct
2
,
202
4
Acc
ep
ted
Oct
7
,
2
0
2
4
In
re
c
e
n
t
y
e
a
rs,
sp
e
e
c
h
e
m
o
ti
o
n
r
e
c
o
g
n
it
i
o
n
(
S
ER)
h
a
s
e
m
e
rg
e
d
a
s
a
p
iv
o
tal
to
o
l
f
o
r
u
n
d
e
rsta
n
d
i
n
g
a
n
d
e
n
h
a
n
c
in
g
h
u
m
a
n
-
c
o
m
p
u
ter
in
tera
c
ti
o
n
(HCI),
th
u
s
g
a
r
n
e
rin
g
si
g
n
if
ica
n
t
a
tt
e
n
ti
o
n
fro
m
re
se
a
rc
h
e
rs
d
u
e
to
it
s
d
iv
e
rse
ra
n
g
e
o
f
a
p
p
li
c
a
ti
o
n
s.
Ho
we
v
e
r,
S
ER
sy
ste
m
s
e
n
c
o
u
n
ter
n
u
m
e
ro
u
s
c
h
a
ll
e
n
g
e
s
,
p
a
rti
c
u
larly
c
o
n
c
e
rn
i
n
g
th
e
se
lec
ti
o
n
o
f
a
p
p
r
o
p
riate
fe
a
tu
re
s
a
n
d
c
las
sifiers
fo
r
e
m
o
ti
o
n
re
c
o
g
n
it
i
o
n
.
T
h
is
p
a
p
e
r
p
ro
v
i
d
e
s
a
c
o
n
c
ise
su
rv
e
y
o
f
th
e
field
o
f
sp
e
e
c
h
e
m
o
ti
o
n
re
c
o
g
n
it
i
o
n
,
e
lu
c
id
a
ti
n
g
it
s
c
las
sifica
ti
o
n
a
lg
o
rit
h
m
s
a
n
d
v
a
rio
u
s
fe
a
tu
re
e
x
trac
ti
o
n
t
e
c
h
n
iq
u
e
s
a
c
ro
ss
m
u
lt
ip
le
lan
g
u
a
g
e
s
.
Ad
d
it
i
o
n
a
ll
y
,
it
e
x
p
l
o
re
s
th
e
li
m
it
a
ti
o
n
s
a
n
d
we
a
k
n
e
ss
e
s
in
h
e
re
n
t
in
sp
e
e
c
h
e
m
o
ti
o
n
re
c
o
g
n
i
ti
o
n
sy
ste
m
s.
F
u
r
th
e
rm
o
re
,
th
e
p
a
p
e
r
e
n
d
e
a
v
o
rs t
o
c
a
teg
o
rize
re
c
e
n
t
re
se
a
r
c
h
e
n
d
e
a
v
o
rs
in
Ara
b
ic
sp
e
e
c
h
e
m
o
ti
o
n
re
c
o
g
n
it
i
o
n
,
e
m
p
lo
y
i
n
g
d
iv
e
rse
m
o
d
e
li
n
g
a
p
p
r
o
a
c
h
e
s a
n
d
e
x
trac
ti
o
n
m
e
th
o
d
s.
K
ey
w
o
r
d
s
:
Ar
ab
ic
s
p
ee
ch
em
o
tio
n
r
ec
o
g
n
itio
n
C
las
s
if
icatio
n
alg
o
r
ith
m
s
Featu
r
e
ex
tr
ac
tio
n
tech
n
iq
u
es
Hu
m
an
-
co
m
p
u
ter
i
n
ter
ac
tio
n
Sp
ee
ch
em
o
tio
n
r
ec
o
g
n
itio
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
:
Mo
h
am
ed
Ham
id
i
T
ea
m
o
f
Mo
d
elin
g
a
n
d
Scien
ti
f
ic
C
o
m
p
u
tin
g
,
Mu
ltid
is
cip
lin
ar
y
Facu
lty
o
f
Nad
or
Mo
h
am
m
ed
Pre
m
ier
Un
iv
er
s
it
y
Ou
jd
a,
Mo
r
o
cc
o
E
m
ail: m
.
h
am
id
i
@u
m
p
.
ac
.
m
a
1.
I
NT
RO
D
UCT
I
O
N
R
ec
o
g
n
izin
g
em
o
tio
n
s
ca
n
b
e
ac
co
m
p
lis
h
ed
th
r
o
u
g
h
v
ar
i
o
u
s
m
o
d
alities
,
en
co
m
p
ass
in
g
tex
t
an
d
s
p
ee
ch
.
T
ex
t
e
m
o
tio
n
r
ec
o
g
n
itio
n
(
T
E
R
)
is
a
s
p
ec
ialized
r
esear
ch
d
o
m
ai
n
ce
n
ter
e
d
o
n
id
e
n
tify
in
g
an
d
ca
teg
o
r
izin
g
em
o
tio
n
s
c
o
n
v
e
y
ed
th
r
o
u
g
h
wr
itten
c
o
m
m
u
n
icatio
n
,
in
clu
d
in
g
s
o
cial
m
ed
ia
p
o
s
ts
[
1
]
.
Sp
ee
ch
em
o
tio
n
r
e
co
g
n
itio
n
(
SER)
i
s
an
in
ter
d
is
cip
lin
ar
y
f
ield
a
im
ed
at
au
to
m
atica
lly
d
etec
ti
n
g
an
d
class
if
y
in
g
em
o
tio
n
s
f
r
o
m
s
p
ee
ch
,
with
ap
p
licatio
n
s
in
ar
ea
s
lik
e
h
u
m
an
-
c
o
m
p
u
ter
in
ter
ac
tio
n
(
HC
I
)
,
h
ea
lth
ca
r
e,
ed
u
ca
tio
n
,
an
d
en
ter
tain
m
e
n
t.
SER
s
y
s
tem
s
g
en
er
ally
co
n
s
is
t
o
f
th
r
ee
m
ain
s
tep
s
:
p
r
e
p
r
o
ce
s
s
in
g
,
f
ea
t
u
r
e
ex
tr
ac
tio
n
,
a
n
d
class
if
icatio
n
.
Pre
p
r
o
ce
s
s
in
g
en
h
a
n
ce
s
s
p
ee
c
h
q
u
ality
a
n
d
s
eg
m
e
n
ts
it
in
to
m
an
ag
ea
b
le
u
n
its
.
Featu
r
e
ex
tr
ac
tio
n
id
en
tifie
s
k
ey
ac
o
u
s
tic,
p
r
o
s
o
d
ic,
a
n
d
lin
g
u
is
tic
elem
en
ts
th
at
co
n
v
ey
em
o
tio
n
.
C
las
s
if
icatio
n
u
s
es
alg
o
r
ith
m
s
to
lab
el
em
o
tio
n
s
b
ased
o
n
th
ese
f
ea
tu
r
es.
SER
i
s
ch
a
llen
g
in
g
d
u
e
to
th
e
co
m
p
lex
ity
o
f
h
u
m
an
em
o
tio
n
s
an
d
v
ar
iab
ilit
y
in
s
p
ee
c
h
ac
r
o
s
s
s
p
ea
k
er
s
,
lan
g
u
ag
es,
an
d
co
n
tex
ts
,
r
eq
u
i
r
in
g
ca
r
ef
u
l
d
esig
n
an
d
o
p
tim
izatio
n
f
o
r
ac
c
u
r
ate
r
esu
lts
[
2
]
.
I
n
1
9
9
7
,
r
esear
c
h
er
s
cr
ea
ted
th
e
d
an
is
h
em
o
tio
n
al
s
p
ee
c
h
(
DE
S)
d
atab
ase,
co
n
ta
in
in
g
r
ec
o
r
d
in
g
s
o
f
two
m
en
an
d
two
wo
m
en
ex
p
r
ess
in
g
f
iv
e
em
o
tio
n
s
,
wh
ich
wer
e
ev
alu
ated
b
y
twen
ty
lis
ten
er
s
to
id
en
tify
th
e
em
o
tio
n
s
[
3
]
.
A
2
0
0
1
s
tu
d
y
o
n
em
o
tio
n
r
ec
o
g
n
itio
n
u
s
ed
th
e
Sp
an
is
h
I
E
SS
DB
co
r
p
u
s
an
d
th
e
R
AM
SES
en
g
in
e,
lev
er
ag
in
g
s
em
i
-
co
n
tin
u
o
u
s
h
id
d
en
m
ar
k
o
v
m
o
d
els
(
HM
Ms)
an
d
lo
w
-
lev
el
f
ea
t
u
r
es
f
o
r
an
aly
s
is
[
4
]
.
I
n
2
0
0
5
,
a
s
tu
d
y
ap
p
lie
d
two
class
if
icatio
n
m
eth
o
d
s
,
HM
M
an
d
s
u
p
p
o
r
t
v
ec
t
o
r
m
ac
h
in
es
(
SVM)
,
to
class
if
y
f
iv
e
em
o
tio
n
s
f
r
o
m
t
h
e
DE
S
d
atab
ase.
Featu
r
es
in
clu
d
ed
f
u
n
d
am
e
n
tal
f
r
e
q
u
en
c
y
,
e
n
er
g
y
,
f
o
r
m
an
t
f
r
eq
u
en
cies,
m
el
-
f
r
eq
u
en
cy
ce
p
s
tr
al
co
e
f
f
icien
ts
(
MFC
C
s
)
,
an
d
m
el
f
r
eq
u
e
n
cy
s
u
b
-
b
an
d
e
n
er
g
ies,
ac
h
i
ev
in
g
n
o
tab
le
r
ec
o
g
n
itio
n
r
ates
[
5
]
.
I
n
2
0
0
7
,
r
esear
ch
er
s
d
elv
ed
in
to
th
e
d
o
m
ain
o
f
h
u
m
an
-
r
o
b
o
t
in
ter
a
ctio
n
(
HR
I
)
b
y
d
ev
el
o
p
in
g
an
in
tellig
en
t
r
o
b
o
t
ca
p
a
b
le
o
f
c
o
m
p
r
eh
e
n
d
in
g
a
nd
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
C
o
n
ce
p
tio
n
o
f sp
ee
c
h
emo
tio
n
r
ec
o
g
n
itio
n
meth
o
d
s
:
a
r
ev
iew
(
A
b
d
elka
d
er B
en
z
ir
a
r
)
1857
r
esp
o
n
d
in
g
to
h
u
m
an
em
o
tio
n
s
[
6
]
.
A
s
ig
n
if
ican
t
b
r
ea
k
t
h
r
o
u
g
h
in
em
o
tio
n
r
ec
o
g
n
itio
n
ca
m
e
in
2
0
1
3
with
th
e
u
s
e
o
f
d
ee
p
b
elief
n
etwo
r
k
s
(
DB
N)
to
ex
tr
ac
t
u
n
s
u
p
er
v
is
ed
au
d
io
-
v
is
u
al
f
ea
tu
r
es
f
o
r
e
m
o
tio
n
class
if
icatio
n
.
T
h
e
f
ield
a
d
v
an
ce
d
f
u
r
th
er
in
2
0
1
7
,
wh
e
n
d
ee
p
n
eu
r
al
n
et
wo
r
k
s
(
DNN)
wer
e
c
o
m
b
i
n
ed
with
v
o
ice
ac
tiv
ity
d
etec
tio
n
(
VAD)
to
ef
f
icien
tl
y
r
em
o
v
e
s
ilen
t
s
eg
m
en
ts
f
r
o
m
s
p
ee
ch
s
ig
n
als,
en
h
an
cin
g
r
ec
o
g
n
itio
n
ac
cu
r
ac
y
[
7
]
,
[
8
].
I
n
2
0
1
9
,
r
esear
ch
er
s
co
m
b
in
ed
b
id
ir
ec
tio
n
al
lo
n
g
-
s
h
o
r
t
ter
m
m
em
o
r
y
(
B
i
L
STM
)
with
VAD
an
d
a
n
atten
tio
n
m
o
d
el
to
f
ilter
o
u
t
s
il
en
ce
an
d
n
o
n
-
em
o
tio
n
al
p
ar
ts
o
f
s
p
ee
ch
,
im
p
r
o
v
in
g
f
o
cu
s
o
n
em
o
tio
n
al
co
n
te
n
t
[
9
]
.
I
n
2
0
2
2
,
Atm
aja
an
d
Sas
o
u
[
1
0
]
ap
p
lied
f
o
u
r
d
ata
au
g
m
en
tatio
n
tech
n
iq
u
es
g
lo
ttal
s
o
u
r
ce
ex
tr
ac
tio
n
,
s
ilen
ce
r
em
o
v
al,
im
p
u
ls
e
r
es
p
o
n
s
e
c
o
n
v
o
lu
tio
n
,
an
d
n
o
is
e
ad
d
itio
n
o
n
J
T
E
S
an
d
I
M
OC
AP
d
atab
ases
,
s
h
o
win
g
th
at
co
m
b
in
in
g
th
ese
m
eth
o
d
s
im
p
r
o
v
ed
s
p
ee
c
h
em
o
tio
n
r
ec
o
g
n
itio
n
p
er
f
o
r
m
an
ce
.
T
h
ey
also
ex
p
lo
r
ed
s
elf
-
s
u
p
e
r
v
is
ed
lear
n
in
g
(
SS
L
)
f
o
r
tr
ain
i
n
g
m
o
d
els
with
o
u
t
ex
ter
n
al
lab
els
[
1
1
]
.
I
n
2
0
2
3
,
th
e
wav
2
v
ec
2
.
0
m
o
d
el
was
im
p
lem
en
ted
o
n
th
e
I
talian
"E
m
o
zio
n
alm
en
te"
d
ata
b
ase,
o
u
tp
er
f
o
r
m
i
n
g
h
u
m
an
a
c
c
u
r
a
c
y
i
n
v
o
c
a
l
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
a
n
d
d
e
m
o
n
s
t
r
a
t
i
n
g
p
o
t
e
n
t
i
a
l
f
o
r
i
n
t
e
g
r
a
t
i
o
n
i
n
t
o
c
o
n
v
e
r
s
a
t
i
o
n
a
l
a
g
e
n
t
s
[
1
2
]
.
I
n
ad
d
itio
n
to
wav
2
v
ec
2
.
0
,
m
o
d
els
lik
e
YAM
n
et
an
d
VGGish
h
av
e
b
ee
n
u
s
ed
f
o
r
s
p
ee
ch
em
o
tio
n
r
ec
o
g
n
itio
n
,
b
u
t
s
p
ee
ch
SS
L
P
T
M
em
b
ed
d
in
g
s
s
h
o
wed
s
u
p
e
r
io
r
p
e
r
f
o
r
m
an
ce
.
N
o
tab
ly
,
x
-
v
ec
to
r
em
b
ed
d
in
g
s
co
m
b
in
ed
with
ex
tr
e
m
e
g
r
ad
i
en
t
b
o
o
s
tin
g
(
XGBo
o
s
t
)
o
u
tp
er
f
o
r
m
e
d
o
th
er
m
o
d
els,
in
clu
d
in
g
wav
2
v
ec
2
.
0
,
u
n
is
p
ee
ch
-
SAT,
wav
L
M,
a
n
d
E
C
APA
[
1
3
]
,
[
1
4
]
.
T
h
is
s
tu
d
y
in
v
esti
g
ates
th
e
f
ield
o
f
S
E
R
ac
r
o
s
s
m
u
ltip
le
lan
g
u
ag
es,
f
o
c
u
s
in
g
o
n
th
e
u
s
e
o
f
v
ar
io
u
s
class
if
icatio
n
alg
o
r
ith
m
s
.
W
h
ile
ea
r
lier
s
tu
d
ies
h
av
e
ex
p
lo
r
e
d
th
e
im
p
ac
t
o
f
tr
ad
itio
n
al
m
ac
h
in
e
lear
n
in
g
m
o
d
els
lik
e
SVM
an
d
HM
M
o
n
SER,
th
ey
h
av
e
n
o
t
ex
p
licitly
ad
d
r
ess
ed
th
e
ch
allen
g
es
o
f
ap
p
ly
in
g
th
ese
m
o
d
els
ac
r
o
s
s
d
iv
er
s
e
lin
g
u
is
tic
co
n
tex
ts
,
p
ar
ticu
lar
ly
i
n
less
-
s
tu
d
ied
lan
g
u
ag
es
s
u
ch
as
Am
az
ig
h
an
d
Ar
a
b
ic.
Fu
r
th
er
m
o
r
e,
m
an
y
ex
is
tin
g
s
tu
d
ies
f
o
c
u
s
p
r
ed
o
m
i
n
an
tly
o
n
well
-
r
ep
r
esen
ted
lan
g
u
ag
es,
o
f
ten
n
e
g
lectin
g
t
h
e
p
e
r
f
o
r
m
an
ce
an
d
ad
ap
ta
b
ilit
y
o
f
SER
s
y
s
tem
s
in
th
ese
u
n
d
er
r
ep
r
esen
ted
lan
g
u
ag
es.
B
y
r
ev
iewin
g
th
e
ev
o
lu
tio
n
o
f
f
ea
tu
r
e
ex
tr
ac
tio
n
tech
n
iq
u
e
s
an
d
th
e
s
h
if
t
f
r
o
m
b
asic
to
o
ls
to
ad
v
an
ce
d
d
ee
p
lear
n
in
g
class
if
ier
s
,
th
is
p
ap
er
s
ee
k
s
to
f
ill
th
ese
g
ap
s
,
o
f
f
er
in
g
n
ew
in
s
ig
h
ts
in
to
th
e
p
er
f
o
r
m
an
ce
a
n
d
c
r
o
s
s
-
lin
g
u
is
tic
ad
ap
tab
ilit
y
o
f
SER s
y
s
tem
s
.
T
h
e
r
em
ain
d
er
o
f
th
is
p
ap
er
i
s
s
tr
u
ctu
r
ed
as
f
o
llo
ws:
s
ec
tio
n
2
d
elv
es
in
to
r
elate
d
wo
r
k
s
,
p
r
o
v
id
in
g
an
o
v
er
v
iew
o
f
ex
is
tin
g
r
esea
r
ch
in
th
e
f
ield
.
Sectio
n
3
o
f
f
er
s
a
co
m
p
r
eh
e
n
s
iv
e
o
v
er
v
ie
w
o
f
s
p
ee
ch
em
o
tio
n
r
ec
o
g
n
itio
n
.
I
n
s
ec
tio
n
4
,
v
ar
i
o
u
s
ap
p
r
o
ac
h
es
to
s
p
ee
ch
em
o
tio
n
r
ec
o
g
n
itio
n
ar
e
d
is
cu
s
s
ed
in
d
etail.
Sectio
n
5
h
ig
h
lig
h
ts
th
e
lim
itatio
n
s
an
d
wea
k
n
ess
es
in
h
er
e
n
t
in
s
p
ee
ch
em
o
tio
n
r
ec
o
g
n
itio
n
.
Sectio
n
6
f
o
cu
s
es
s
p
ec
if
ically
o
n
s
tu
d
ies
co
n
d
u
cted
in
Ar
ab
ic
s
p
ee
ch
em
o
tio
n
r
ec
o
g
n
itio
n
.
Sectio
n
7
p
r
esen
ts
th
e
r
esu
lts
an
d
d
is
cu
s
s
io
n
o
f
th
e
d
if
f
er
en
t stu
d
ies th
at
we
m
en
tio
n
ed
b
ef
o
r
e
.
Fin
ally
,
th
e
p
a
p
er
en
d
s
with
a
co
n
clu
s
io
n
.
2.
RE
L
AT
E
D
WO
RK
S
A
s
tu
d
y
co
n
d
u
cted
b
y
No
g
u
ei
r
as
et
a
l.
[
4
]
u
tili
ze
d
th
e
HM
M
in
co
n
ju
n
ctio
n
with
p
itch
a
n
d
en
er
g
y
f
ea
tu
r
es
to
class
if
y
s
ev
en
em
o
tio
n
al
s
tates:
an
g
er
,
d
is
g
u
s
t,
f
ea
r
,
j
o
y
,
s
ad
n
ess
,
s
u
r
p
r
is
e,
an
d
n
e
u
tr
ality
.
T
h
e
f
in
d
in
g
s
r
ev
ea
led
th
at
u
s
in
g
i
n
s
tan
tan
eo
u
s
p
itch
led
to
o
v
er
8
0
%
ac
c
u
r
ac
y
in
s
p
ee
c
h
em
o
tio
n
r
ec
o
g
n
itio
n
.
I
n
an
o
th
er
s
tu
d
y
,
L
in
et
a
l.
[
5
]
ap
p
lied
HM
M
a
n
d
SVM
class
if
ier
s
with
f
ea
tu
r
es
lik
e
f
u
n
d
am
en
tal
f
r
e
q
u
en
c
y
,
f
o
r
m
an
t
f
r
eq
u
e
n
cies,
MFC
C
s
,
an
d
m
el
s
u
b
-
b
an
d
en
er
g
ies,
alo
n
g
s
id
e
s
eq
u
en
tial
f
o
r
war
d
s
elec
tio
n
(
SF
S)
f
o
r
f
ea
tu
r
e
o
p
tim
izatio
n
.
T
h
e
HM
M
class
if
ier
ac
h
iev
ed
im
p
r
ess
iv
e
r
ec
o
g
n
itio
n
r
ates,
r
ea
ch
in
g
9
8
.
9
%
f
o
r
f
em
ale
s
u
b
jects,
1
0
0
% f
o
r
m
ales,
an
d
9
9
.
5
% f
o
r
g
en
d
er
-
in
d
ep
en
d
en
t
ca
s
es.
Har
ár
et
a
l.
[
8
]
u
s
ed
a
DNN
with
VAD
to
r
ec
o
g
n
ize
th
r
ee
em
o
tio
n
al
s
t
ates
an
g
r
y
,
s
ad
,
an
d
n
e
u
tr
al
in
t
h
e
E
m
o
-
DB
d
ataset,
ac
h
iev
in
g
a
9
6
.
9
7
%
r
ec
o
g
n
itio
n
r
ate.
C
atan
ia
[
1
2
]
ap
p
lied
th
e
wav
2
v
ec
2
.
0
m
o
d
el
to
th
e
I
talian
E
m
o
zio
n
alm
en
te
d
atab
ase,
ac
h
iev
in
g
8
3
%
ac
cu
r
ac
y
i
n
s
p
ea
k
er
-
d
ep
en
d
en
t
ca
s
es
an
d
8
1
%
in
s
p
ea
k
er
-
in
d
e
p
en
d
e
n
t
c
ases
,
o
u
tp
er
f
o
r
m
i
n
g
r
esu
lts
f
r
o
m
th
e
E
m
o
v
o
d
atas
et.
Ph
u
k
an
et
a
l.
[
1
4
]
c
o
m
p
a
r
ed
eig
h
t
p
r
e
-
tr
ain
ed
m
o
d
els,
i
n
clu
d
in
g
wa
v
2
v
ec
2
.
0
,
x
-
v
ec
to
r
,
an
d
E
C
APA,
u
s
in
g
XGBo
o
s
t,
r
an
d
o
m
f
o
r
est
(
R
F)
,
an
d
f
u
lly
co
n
v
o
lu
tio
n
al
n
etwo
r
k
(
FC
N)
ac
r
o
s
s
d
atab
ases
lik
e
C
R
E
MA
-
D,
T
E
SS
,
SAVE
E
,
a
n
d
E
m
o
-
DB
,
with
s
p
ea
k
er
r
ec
o
g
n
i
tio
n
m
o
d
els
s
h
o
win
g
th
e
b
est
p
er
f
o
r
m
an
ce
.
Hu
a
n
g
et
a
l.
[
1
5
]
ex
p
l
o
r
ed
s
em
i
-
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NNs
)
in
u
n
s
u
p
er
v
is
ed
a
n
d
s
em
i
-
s
u
p
er
v
is
ed
s
ettin
g
s
,
u
s
in
g
an
o
b
je
ctiv
e
f
u
n
ctio
n
to
lear
n
af
f
ec
t
-
r
elev
an
t
f
ea
tu
r
es,
wh
ich
s
ig
n
if
ican
tly
o
u
tp
er
f
o
r
m
ed
n
o
n
-
d
is
cr
im
in
ativ
e
o
n
es
ac
r
o
s
s
f
o
u
r
d
ata
b
ases
.
Asg
h
ar
et
a
l.
[
1
6
]
d
ev
el
o
p
ed
an
Ur
d
u
em
o
tio
n
d
atab
ase,
ap
p
ly
in
g
K
-
n
ea
r
est
n
eig
h
b
o
r
(
K
NN
)
,
SVM,
an
d
R
F
cla
s
s
if
ier
s
with
f
ea
tu
r
es
l
ik
e
MFC
C
,
lin
e
ar
p
r
ed
icti
o
n
c
o
e
f
f
ic
ie
n
ts
(
L
PC
)
,
an
d
e
n
er
g
y
,
i
m
p
r
o
v
i
n
g
ac
cu
r
ac
y
f
r
o
m
6
6
.
5
%
to
7
6
.
5
%
af
ter
ex
clu
d
in
g
d
is
g
u
s
t.
Xia
an
d
Z
h
ao
[
1
7
]
u
s
ed
a
C
NN
-
B
i
L
STM
with
an
atten
tio
n
m
o
d
el
an
d
3
D
MFC
C
f
ea
tu
r
es
,
ac
h
iev
in
g
a
6
%
ac
cu
r
ac
y
b
o
o
s
t.
Atm
aja
an
d
Sas
o
u
[
1
8
]
em
p
lo
y
ed
late
f
u
s
io
n
o
f
n
in
e
p
r
e
-
tr
ain
ed
m
o
d
els
to
r
ec
o
g
n
ize
s
h
ar
ed
em
o
tio
n
s
f
r
o
m
m
u
ltil
in
g
u
al
s
p
ee
ch
(
E
n
g
lis
h
an
d
Sp
an
is
h
)
,
ac
h
ie
v
in
g
a
t
o
p
Sp
ea
r
m
an
s
co
r
e
o
f
0
.
5
3
7
with
SVM
class
if
icat
io
n
.
Ph
am
et
a
l.
[
1
9
]
p
r
o
p
o
s
ed
a
h
y
b
r
id
d
ata
a
u
g
m
e
n
tatio
n
ap
p
r
o
ac
h
co
m
b
in
e
d
with
g
en
er
ativ
e
ad
v
er
s
ar
ial
n
e
two
r
k
s
(
GANs)
f
o
r
em
o
tio
n
r
ec
o
g
n
itio
n
o
n
th
e
E
m
o
-
DB
d
a
taset.
T
h
ey
u
tili
ze
d
3
D
lo
g
m
el
-
s
p
ec
tr
o
g
r
am
f
ea
t
u
r
es
with
an
ADCR
NN
m
o
d
el,
ac
h
iev
in
g
8
7
.
1
2
%
ac
cu
r
a
cy
with
tr
a
d
itio
n
al
m
eth
o
d
s
an
d
8
8
.
4
7
%
with
G
ANs.
Z
h
an
g
et
a
l.
[
2
0
]
in
tr
o
d
u
ce
d
a
ca
p
s
u
le
n
etwo
r
k
(
C
a
p
s
Net)
ap
p
r
o
ac
h
f
o
r
SER,
en
h
an
ce
d
b
y
d
ata
au
g
m
en
tatio
n
,
ac
h
iev
in
g
9
1
.
6
7
%
ac
cu
r
ac
y
,
wh
ich
in
cr
ea
s
ed
to
9
3
.
3
3
%
wh
en
d
ata
au
g
m
en
tatio
n
was a
p
p
lied
,
ef
f
ec
tiv
ely
ad
d
r
ess
in
g
d
ee
p
lear
n
in
g
ch
allen
g
es a
n
d
d
ata
s
ca
r
cit
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
7
,
No
.
3
,
Ma
r
ch
20
2
5
:
1
85
6
-
1
86
4
1858
3.
SPEE
CH
E
M
O
T
I
O
N
R
E
C
O
G
NIT
I
O
N
3
.
1
.
Sp
ee
ch
em
o
t
io
n r
ec
o
g
ni
t
io
n
s
y
s
t
em
a
rc
hite
ct
ure
T
h
e
SER
task
in
v
o
lv
es
s
p
e
ec
h
p
r
o
ce
s
s
in
g
an
d
c
o
m
p
u
tatio
n
al
p
ar
alin
g
u
is
tic
an
aly
s
is
with
th
e
o
b
jectiv
e
o
f
id
en
tif
y
in
g
an
d
ca
teg
o
r
izin
g
em
o
tio
n
s
co
n
v
e
y
ed
in
s
p
o
k
en
la
n
g
u
a
g
e.
Fig
u
r
e
1
illu
s
tr
ates
th
e
v
ar
io
u
s
p
h
ases
in
v
o
lv
e
d
in
s
p
ee
ch
em
o
tio
n
r
ec
o
g
n
itio
n
.
W
h
er
e,
p
r
e
-
p
r
o
ce
s
s
in
g
:
in
v
o
l
v
es
s
tan
d
ar
d
izin
g
th
e
v
o
lu
m
e
an
d
in
ten
s
ity
o
f
v
o
c
al
s
ig
n
als.
Featu
r
e
ex
tr
ac
tio
n
:
co
n
s
is
ts
o
f
th
e
ex
tr
ac
tio
n
o
f
f
ea
tu
r
es
s
u
ch
as
MFC
C
,
L
PC
an
d
lin
ea
r
p
r
ed
i
ctio
n
ce
p
s
tr
al
co
ef
f
icien
ts
(
L
PC
C
)
.
T
r
ain
in
g
:
th
e
m
o
d
el
is
tr
ain
ed
o
n
a
lar
g
e
d
ataset
o
f
lab
eled
em
o
tio
n
al
s
tates
lik
e
h
ap
p
y
,
an
g
r
y
,
an
d
s
u
r
p
r
is
ed
.
I
n
th
is
s
tep
,
th
e
m
o
d
el
lear
n
s
to
ass
o
ciate
th
e
ex
tr
ac
ted
f
ea
tu
r
es
with
th
e
co
r
r
esp
o
n
d
in
g
em
o
tio
n
s
in
t
h
e
tr
ain
in
g
d
ata.
E
m
o
tio
n
r
ec
o
g
n
itio
n
:
t
h
e
m
o
d
el
u
s
es
th
e
ex
tr
ac
ted
f
ea
tu
r
es
f
r
o
m
th
e
test
au
d
io
to
p
r
ed
ict
th
e
m
o
s
t
p
r
o
b
ab
le
em
o
tio
n
.
T
o
d
o
th
is
,
it
co
m
p
ar
es
th
e
f
ea
tu
r
es to
th
e
lea
r
n
ed
m
o
d
els d
u
r
in
g
t
r
ain
in
g
a
n
d
ass
ig
n
s
th
e
em
o
tio
n
lab
el
with
th
e
h
i
g
h
est p
r
o
b
a
b
ilit
y
.
Fig
u
r
e
1
.
Sp
ee
c
h
em
o
tio
n
r
ec
o
g
n
itio
n
s
y
s
tem
ar
ch
itectu
r
e
3
.
2
.
F
e
a
t
ure
ex
t
r
a
ct
io
n t
ec
h
niq
ue
s
Feat
u
r
e
e
x
t
r
a
cti
o
n
is
t
h
e
p
i
v
o
t
al
p
r
o
ce
s
s
o
f
c
o
n
v
e
r
t
in
g
t
h
e
s
p
e
ec
h
s
i
g
n
al
i
n
t
o
a
s
et
o
f
p
ar
a
m
et
er
s
t
h
at
f
a
cili
tat
e
th
e
i
d
en
tif
ic
ati
o
n
a
n
d
c
lass
i
f
i
ca
t
io
n
o
f
v
a
r
i
o
u
s
s
p
e
ec
h
s
o
u
n
d
s
a
n
d
e
m
o
ti
o
n
s
.
B
el
o
w
,
we
will
e
v
o
k
e
s
o
m
e
f
e
at
u
r
e
e
x
t
r
ac
ti
o
n
m
et
h
o
d
s
.
3
.
2
.
1
.
M
el
f
re
qu
ency
ce
ps
t
ra
l c
o
ef
f
icient
s
MFC
C
ar
e
a
wid
ely
u
s
ed
m
eth
o
d
f
o
r
f
ea
tu
r
e
e
x
tr
ac
tio
n
in
s
p
ee
ch
p
r
o
ce
s
s
i
n
g
,
ef
f
ec
tiv
ely
ca
p
tu
r
in
g
k
ey
s
p
ec
tr
al
p
r
o
p
er
ties
th
r
o
u
g
h
1
0
to
1
2
c
o
ef
f
icien
ts
.
W
h
ile
MFC
C
is
p
o
p
u
lar
,
it
is
h
i
g
h
l
y
s
en
s
itiv
e
to
n
o
is
e,
wh
ich
ca
n
im
p
air
th
e
ac
cu
r
a
cy
o
f
s
p
ee
ch
r
ec
o
g
n
itio
n
s
y
s
tem
s
.
T
h
is
s
en
s
i
tiv
ity
s
tem
s
f
r
o
m
its
r
elian
ce
o
n
s
p
ec
tr
al
f
ea
tu
r
es,
m
ak
in
g
it
v
u
ln
er
ab
le
to
d
is
to
r
tio
n
b
y
b
ac
k
g
r
o
u
n
d
n
o
is
e.
T
h
er
e
f
o
r
e,
im
p
r
o
v
in
g
th
e
r
o
b
u
s
tn
ess
o
f
MFC
C
in
n
o
is
y
en
v
ir
o
n
m
en
ts
r
em
ain
s
a
cr
u
cial
f
o
cu
s
in
s
p
ee
ch
p
r
o
ce
s
s
in
g
r
e
s
ea
r
ch
[
2
1
]
-
[
2
6
]
.
3
.
2
.
2
.
L
inea
r
predict
io
n c
o
ef
f
icient
s
L
PC
m
o
d
els
a
s
p
e
ec
h
s
i
g
n
a
l
b
y
p
r
e
d
ict
in
g
ea
ch
s
a
m
p
le
as
a
wei
g
h
te
d
s
u
m
o
f
p
r
ev
io
u
s
s
a
m
p
les
.
T
h
is
m
et
h
o
d
e
f
f
ec
t
iv
el
y
c
ap
tu
r
es
t
h
e
v
o
c
al
t
r
a
ct'
s
s
h
a
p
e
a
n
d
is
v
it
al
f
o
r
i
d
e
n
t
if
y
i
n
g
f
o
r
m
a
n
t
f
r
e
q
u
en
ci
es,
w
h
i
c
h
co
n
t
r
i
b
u
te
t
o
a
v
o
i
ce
'
s
d
is
t
in
c
t
t
im
b
r
e
.
L
PC
c
o
e
f
f
ic
ie
n
ts
a
r
e
t
h
e
r
e
f
o
r
e
c
r
it
ica
l
i
n
s
p
ee
c
h
a
n
al
y
s
is
f
o
r
p
r
ec
is
e
s
ig
n
al
m
o
d
e
li
n
g
a
n
d
s
y
n
th
es
is
.
A
d
d
it
io
n
all
y
,
L
PC
C
s
e
r
v
e
as
a
f
u
n
d
am
en
tal
f
ea
tu
r
e
in
v
ar
i
o
u
s
s
p
e
ec
h
p
r
o
ce
s
s
i
n
g
ap
p
lic
ati
o
n
s
[
2
2
]
,
[
2
7
]
-
[
2
9
]
.
3
.
2
.
3
.
L
inea
r
predict
io
n c
eps
t
ra
l c
o
ef
f
icient
s
Per
ce
p
t
u
al
li
n
ea
r
p
r
ed
ict
io
n
(
P
L
P)
e
n
h
a
n
ce
s
t
h
e
s
h
o
r
t
-
t
er
m
s
p
e
ec
h
s
p
ec
t
r
u
m
an
al
y
s
is
b
y
i
n
co
r
p
o
r
a
ti
n
g
p
s
y
c
h
o
p
h
y
s
ic
al
a
d
j
u
s
t
m
en
ts
t
h
at
ali
g
n
m
o
r
e
cl
o
s
e
ly
wit
h
h
u
m
an
au
d
i
t
o
r
y
p
e
r
c
ep
ti
o
n
.
I
t
u
ti
liz
es
p
a
r
a
m
et
er
s
f
r
o
m
a
f
ilt
er
b
a
n
k
o
f
1
8
f
ilt
e
r
s
,
d
is
tr
ib
u
t
ed
ac
co
r
d
i
n
g
to
t
h
e
B
ar
k
s
ca
le
,
wh
ic
h
r
ef
lec
ts
t
h
e
n
o
n
li
n
ea
r
f
r
eq
u
en
cy
p
e
r
c
ep
ti
o
n
o
f
t
h
e
h
u
m
a
n
e
ar
.
C
o
v
er
in
g
a
r
a
n
g
e
f
r
o
m
0
t
o
5
,
0
0
0
Hz
,
t
h
ese
f
ilt
er
s
e
f
f
e
cti
v
e
l
y
ca
p
t
u
r
e
t
h
e
c
r
iti
ca
l
asp
ec
ts
o
f
s
p
ee
c
h
.
T
h
is
a
p
p
r
o
ac
h
e
n
s
u
r
es
th
at
PL
P
p
r
o
v
id
e
s
a
m
o
r
e
ac
cu
r
ate
r
e
p
r
ese
n
t
ati
o
n
o
f
h
o
w
h
u
m
a
n
s
p
e
r
c
ei
v
e
s
o
u
n
d
[
2
2
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
C
o
n
ce
p
tio
n
o
f sp
ee
c
h
emo
tio
n
r
ec
o
g
n
itio
n
meth
o
d
s
:
a
r
ev
iew
(
A
b
d
elka
d
er B
en
z
ir
a
r
)
1859
4.
SPEE
CH
E
M
O
T
I
O
N
R
E
C
O
G
NIT
I
O
N
AP
P
RO
A
CH
E
S
4
.
1
.
Na
i
v
e
B
a
y
es c
la
s
s
if
ier
T
h
e
Naiv
e
B
ay
es
cla
s
s
if
ier
(
NB
)
is
a
p
r
o
b
ab
ilis
tic
clas
s
if
i
er
r
o
o
ted
in
B
ay
es'
th
eo
r
em
.
R
esear
ch
er
s
h
av
e
em
p
lo
y
e
d
t
h
is
class
if
ier
i
n
n
u
m
er
o
u
s
s
tu
d
ies
f
o
c
u
s
ed
o
n
tr
ad
itio
n
al
s
en
tim
en
t
a
n
aly
s
is
task
s
.
I
t
r
elies
o
n
ca
lcu
latin
g
co
n
d
itio
n
al
p
r
o
b
ab
ilit
ies u
s
in
g
(
1
)
[
30
]
:
(
|
)
=
(
)
.
(
)
/
(
)
(
1
)
wh
er
e:
A
is
a
class
,
an
d
B
is
an
in
d
ep
en
d
en
t v
a
r
iab
le
o
r
ev
en
t.
P(A
|
B
)
:
r
ep
r
esen
ts
th
e
p
o
s
ter
i
o
r
p
r
o
b
ab
ilit
y
o
f
B
d
ep
e
n
d
in
g
to
class
A
.
P(B|
A)
:
r
ep
r
esen
ts
th
e
lik
elih
o
o
d
o
f
B
wh
en
class
is
B
.
P(A
)
:
i
s
th
e
p
r
io
r
in
f
o
r
m
atio
n
o
f
th
e
class
A.
P(B):
i
s
th
e
ev
id
en
ce
o
f
th
e
in
d
ep
en
d
e
n
t v
ar
iab
le
B
.
4
.
2
.
H
idd
en
m
a
r
k
o
v
mo
del
T
h
e
HM
M
is
a
wid
ely
-
u
s
ed
class
if
ier
in
s
p
ee
ch
em
o
tio
n
r
ec
o
g
n
itio
n
,
ef
f
ec
tiv
ely
m
o
d
elin
g
th
e
d
y
n
am
ic
n
at
u
r
e
o
f
s
p
ee
ch
.
Stu
d
ies
s
h
o
w
th
at
HM
M
ac
h
iev
es
im
p
r
o
v
ed
p
er
f
o
r
m
an
ce
wh
e
n
u
s
i
n
g
lo
g
a
r
ith
m
ic
f
r
eq
u
e
n
cy
p
o
wer
c
o
ef
f
icien
ts
as
f
ea
tu
r
es.
T
h
is
m
eth
o
d
h
as
b
ee
n
f
o
u
n
d
to
o
u
tp
er
f
o
r
m
tr
a
d
itio
n
al
tech
n
iq
u
es
lik
e
L
PC
C
an
d
MFC
C
in
em
o
tio
n
r
ec
o
g
n
itio
n
ac
c
u
r
ac
y
[
3
1
]
-
[
33
]
.
4
.
3
.
Su
pp
o
rt
v
ec
t
o
r
ma
chine
SVM
is
k
n
o
wn
f
o
r
its
s
im
p
licity
an
d
co
m
p
u
tatio
n
al
e
f
f
i
cien
cy
,
m
ak
i
n
g
it
a
p
o
p
u
lar
ch
o
i
ce
i
n
m
ac
h
in
e
lear
n
in
g
.
Desp
ite
its
s
tr
aig
h
tf
o
r
war
d
s
tr
u
ctu
r
e,
it
ex
ce
ls
in
class
if
icat
io
n
task
s
with
h
ig
h
p
r
ec
is
io
n
.
R
esear
ch
in
d
icate
s
th
at
SVM
o
f
ten
s
u
r
p
ass
es
o
th
er
m
o
d
els
i
n
class
if
icatio
n
ac
cu
r
ac
y
,
m
a
k
in
g
it
a
r
eliab
le
t
o
o
l
in
v
ar
io
u
s
a
p
p
licatio
n
s
[
34
].
4
.
4
.
K
-
nea
re
s
t
neig
hb
o
r
KNN
is
a
p
o
p
u
lar
s
u
p
er
v
is
ed
alg
o
r
ith
m
u
s
ed
f
o
r
b
o
th
clas
s
if
icatio
n
an
d
r
eg
r
ess
io
n
task
s
.
I
t
g
r
o
u
p
s
d
ata
p
o
i
n
ts
b
ased
o
n
f
ea
tu
r
e
s
im
ilar
ity
,
ass
u
m
in
g
n
ea
r
b
y
p
o
in
ts
in
th
e
f
ea
tu
r
e
s
p
ac
e
s
h
a
r
e
th
e
s
am
e
lab
el
o
r
v
alu
e.
KNN
ty
p
ically
u
s
es
E
u
clid
ea
n
d
is
tan
ce
to
m
ea
s
u
r
e
cl
o
s
en
ess
,
m
ak
in
g
p
r
ed
ictio
n
s
b
ased
o
n
th
e
n
ea
r
est
n
eig
h
b
o
r
s
[
2
]
,
[
3
5
]
.
(
,
)
=
√
∑
(
−
)
²
=
1
(
2
)
wh
er
e
a
an
d
b
ar
e
two
p
o
i
n
ts
in
E
u
clid
ea
n
s
p
ac
e
,
wh
ile
a
k
an
d
b
k
ar
e
E
u
clid
ea
n
v
ec
to
r
s
an
d
n
is
th
e
n
-
th
s
p
ac
e.
4
.
5
.
Art
if
ici
a
l
neura
l net
wo
rk
T
h
e
ar
tific
ial
n
eu
r
al
n
etwo
r
k
(
ANN)
is
m
o
d
eled
a
f
ter
b
io
lo
g
ical
n
eu
r
al
s
y
s
tem
s
,
with
f
ee
d
-
f
o
r
war
d
n
etwo
r
k
s
b
ei
n
g
wid
ely
u
s
ed
i
n
class
if
icatio
n
task
s
.
T
h
ese
n
etwo
r
k
s
co
n
s
is
t
o
f
in
te
r
co
n
n
e
cted
n
eu
r
o
n
s
ac
r
o
s
s
lay
er
s
,
wh
er
e
ea
ch
n
eu
r
o
n
c
o
n
n
ec
ts
to
th
o
s
e
in
th
e
p
r
e
v
io
u
s
lay
er
.
T
h
is
s
tr
u
ctu
r
e
e
n
a
b
les
th
e
n
etwo
r
k
t
o
p
r
o
ce
s
s
in
p
u
t d
ata
a
n
d
lear
n
p
atter
n
s
f
o
r
d
ec
is
io
n
-
m
ak
in
g
[
3
6
]
.
4
.
6
.
Rec
urre
nt
neura
l net
wo
rk
R
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
s
(
R
NNs)
ar
e
p
o
wer
f
u
l
d
ee
p
lear
n
in
g
class
if
ier
s
,
p
ar
ticu
lar
ly
ef
f
ec
tiv
e
f
o
r
task
s
in
v
o
lv
in
g
s
eq
u
en
tial
d
ata.
T
h
ey
ex
ce
l
in
a
p
p
licatio
n
s
lik
e
s
p
ee
ch
em
o
tio
n
r
e
co
g
n
itio
n
,
s
p
ee
c
h
r
ec
o
g
n
itio
n
,
an
d
lan
g
u
ag
e
tr
a
n
s
latio
n
b
y
u
tili
zin
g
in
f
o
r
m
at
io
n
f
r
o
m
p
r
ev
io
u
s
in
p
u
ts
.
R
NNs
ar
e
k
n
o
wn
f
o
r
th
eir
im
p
r
ess
iv
e
r
esu
lts
,
h
an
d
lin
g
co
m
p
lex
d
ata
p
atter
n
s
with
ea
s
e.
T
h
eir
v
e
r
s
atility
an
d
s
tr
en
g
t
h
h
av
e
estab
lis
h
ed
R
NNs
as
ess
en
tia
l
to
o
ls
in
ad
v
a
n
ce
d
m
ac
h
in
e
le
ar
n
in
g
an
d
ar
tific
ial
in
tellig
en
ce
(
AI
)
ap
p
licatio
n
s
[
30
]
,
[
37
]
.
4
.
7
.
L
o
ng
s
ho
rt
-
t
er
m
m
emo
ry
net
wo
rk
s
L
STM
n
etw
o
r
k
s
i
m
p
r
o
v
e
u
p
o
n
R
NNs
b
y
ad
d
r
ess
i
n
g
g
r
a
d
ie
n
t
e
x
p
l
o
s
i
o
n
a
n
d
v
a
n
is
h
i
n
g
g
r
a
d
ie
n
t
is
s
u
es
,
r
esu
lti
n
g
in
h
i
g
h
e
r
ac
cu
r
a
cy
.
L
S
T
Ms
u
s
e
s
p
ec
i
ali
z
ed
g
at
es
in
p
u
t
,
o
u
t
p
u
t
a
n
d
f
o
r
g
et
t
o
ef
f
ec
tiv
ely
m
a
n
a
g
e
i
n
f
o
r
m
ati
o
n
f
l
o
w
o
v
e
r
lo
n
g
s
e
q
u
e
n
c
es.
T
h
is
m
a
k
es
t
h
e
m
es
p
ec
i
all
y
p
o
we
r
f
u
l
f
o
r
t
a
s
k
s
r
e
q
u
ir
in
g
lo
n
g
-
ter
m
d
e
p
e
n
d
e
n
c
y
u
n
d
e
r
s
t
an
d
i
n
g
.
C
o
n
s
e
q
u
en
tl
y
,
L
STM
s
ar
e
f
av
o
r
e
d
f
o
r
ap
p
l
ic
ati
o
n
s
n
e
e
d
i
n
g
r
o
b
u
s
t
s
eq
u
en
ce
m
o
d
eli
n
g
[
1
7
]
,
[
30
]
,
[
37
]
.
4
.
8
.
Co
nv
o
lutio
na
l
neura
l net
wo
rk
C
NNs
ar
e
h
i
g
h
ly
ef
f
ec
ti
v
e
i
n
d
ee
p
le
ar
n
i
n
g
,
p
ar
tic
u
l
ar
ly
in
im
ag
e
a
n
d
s
p
ee
ch
r
e
c
o
g
n
i
ti
o
n
,
d
u
e
t
o
t
h
ei
r
ab
i
lit
y
t
o
le
ar
n
co
m
p
l
ex
d
ata
p
a
tte
r
n
s
.
T
h
e
y
u
t
ili
ze
co
n
v
o
l
u
ti
o
n
al
la
y
er
s
f
o
r
f
ea
tu
r
e
d
et
ec
ti
o
n
an
d
p
o
o
li
n
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
7
,
No
.
3
,
Ma
r
ch
20
2
5
:
1
85
6
-
1
86
4
1860
lay
e
r
s
f
o
r
d
im
e
n
s
i
o
n
ali
ty
r
e
d
u
cti
o
n
w
h
i
le
m
ai
n
ta
in
in
g
c
r
u
ci
al
in
f
o
r
m
a
ti
o
n
.
B
at
ch
n
o
r
m
al
i
za
ti
o
n
a
n
d
d
r
o
p
o
u
t
lay
e
r
s
a
r
e
o
f
te
n
em
p
l
o
y
e
d
to
s
t
ab
i
liz
e
tr
ai
n
i
n
g
a
n
d
p
r
e
v
e
n
t
o
v
er
f
it
ti
n
g
,
r
es
p
ec
t
iv
el
y
.
T
h
is
c
o
m
b
in
ati
o
n
o
f
l
a
y
e
r
s
en
s
u
r
es C
N
Ns
ca
n
ef
f
i
cie
n
t
ly
p
r
o
ce
s
s
a
n
d
le
a
r
n
f
r
o
m
la
r
g
e
d
atas
ets.
A
d
d
iti
o
n
al
ly
,
DB
Ns
ar
e
v
al
u
a
b
l
e
in
s
p
ee
c
h
em
o
ti
o
n
r
e
c
o
g
n
i
ti
o
n
,
o
f
f
e
r
i
n
g
s
tr
o
n
g
c
ap
a
b
ili
ties
i
n
th
is
d
o
m
ain
[
1
7
]
,
[
2
9
]
,
[
38
]
-
[
40
]
.
5.
L
I
M
I
T
S
AND
WE
AK
N
E
S
S
E
S O
F
SPE
E
CH
E
M
O
T
I
O
N
RE
CO
G
NI
T
I
O
N
T
h
e
s
p
ee
ch
em
o
tio
n
r
ec
o
g
n
it
io
n
h
as
s
ev
er
al
lim
its
an
d
w
ea
k
n
ess
es
th
at
wil
l
b
e
p
r
esen
ted
at
th
e
f
o
llo
win
g
:
5
.
1
.
E
mo
t
io
n v
a
ria
bil
it
y
E
m
o
tio
n
s
ar
e
co
m
p
le
x
an
d
c
an
v
ar
y
s
ig
n
if
ican
tly
b
etwe
en
in
d
iv
id
u
als,
r
ef
lectin
g
d
if
f
er
en
ce
s
in
p
er
s
o
n
al
ex
p
er
ien
ce
s
an
d
p
s
y
ch
o
lo
g
ical
s
tates.
C
u
ltu
r
al
n
o
r
m
s
an
d
v
alu
es
also
p
lay
a
cr
u
cial
r
o
le
in
s
h
ap
in
g
h
o
w
em
o
tio
n
s
ar
e
ex
p
r
ess
ed
an
d
p
e
r
ce
iv
ed
,
lead
in
g
to
v
ar
ia
tio
n
s
ac
r
o
s
s
d
if
f
e
r
en
t
s
o
cieties.
Ad
d
itio
n
ally
,
th
e
co
n
tex
t
in
wh
ich
an
em
o
tio
n
is
ex
p
er
ien
ce
d
ca
n
in
f
lu
e
n
ce
its
ex
p
r
ess
io
n
,
m
ak
in
g
it
d
if
f
i
cu
lt
to
estab
lis
h
a
u
n
iv
er
s
al
f
r
am
ewo
r
k
f
o
r
em
o
tio
n
id
en
tific
atio
n
.
As
a
r
e
s
u
lt,
th
e
in
ter
p
r
etatio
n
o
f
e
m
o
tio
n
s
is
h
ig
h
ly
s
u
b
jectiv
e
an
d
ca
n
d
if
f
e
r
n
o
t
o
n
ly
ac
r
o
s
s
lan
g
u
a
g
es b
u
t a
ls
o
with
in
d
iv
er
s
e
cu
ltu
r
al
a
n
d
s
o
c
ial
co
n
tex
ts
[
41
]
.
5
.
2
.
Da
t
a
s
ca
rc
it
y
a
nd
qu
a
lity
SER
s
y
s
tem
s
n
ec
es
s
itate
ex
ten
s
iv
e
an
d
v
ar
ied
d
atasets
o
f
s
p
ee
ch
s
ig
n
als
an
n
o
tated
with
r
eliab
le
em
o
tio
n
lab
els.
No
n
eth
eless
,
ass
em
b
lin
g
s
u
ch
d
atasets
is
ch
allen
g
in
g
d
u
e
to
th
eir
r
a
r
ity
,
ex
p
en
s
e,
a
n
d
th
e
tim
e
r
eq
u
ir
ed
f
o
r
c
o
llectio
n
an
d
lab
elin
g
.
Fu
r
t
h
er
m
o
r
e,
th
e
q
u
ality
o
f
s
p
ee
ch
s
ig
n
als
m
a
y
b
e
co
m
p
r
o
m
is
ed
b
y
f
ac
to
r
s
lik
e
n
o
is
e,
d
is
to
r
tio
n
,
s
p
ea
k
er
v
ar
iab
ilit
y
,
an
d
ch
a
n
n
el
v
ar
iatio
n
s
,
all
o
f
wh
ich
ca
n
d
im
in
is
h
th
e
p
er
f
o
r
m
an
ce
o
f
SER s
y
s
tem
s
[
42
]
.
5
.
3
.
F
e
a
t
ure
ex
t
r
a
ct
io
n a
nd
s
elec
t
io
n
T
h
e
o
p
tim
al
f
ea
t
u
r
e
s
et
f
o
r
S
E
R
is
s
till
d
eb
ated
,
as
n
o
s
in
g
le
s
et
h
as
b
ee
n
u
n
i
v
er
s
ally
a
cc
ep
ted
as
b
est.
T
h
e
ef
f
ec
tiv
e
n
ess
o
f
f
ea
tu
r
es
v
ar
ies
b
ased
o
n
th
e
e
m
o
tio
n
m
o
d
el,
d
ataset,
an
d
class
if
icatio
n
alg
o
r
ith
m
u
s
ed
.
So
m
e
f
ea
tu
r
es
m
ay
in
t
r
o
d
u
ce
r
ed
u
n
d
an
c
y
o
r
n
o
is
e,
c
o
m
p
licatin
g
th
e
an
aly
s
is
an
d
p
o
ten
tially
r
ed
u
cin
g
SER
p
er
f
o
r
m
a
n
ce
.
T
h
er
ef
o
r
e
,
ca
r
ef
u
l
s
elec
tio
n
a
n
d
e
v
al
u
atio
n
o
f
f
ea
tu
r
es
ar
e
ess
en
tial
f
o
r
d
ev
elo
p
in
g
ef
f
icien
t a
n
d
ac
c
u
r
ate
SER s
y
s
tem
s
[
43
]
.
5
.
4
.
Cla
s
s
if
ica
t
io
n a
lg
o
rit
hm
s
Alg
o
r
ith
m
p
er
f
o
r
m
an
ce
in
SE
R
v
ar
ies
b
ased
o
n
th
e
em
o
tio
n
m
o
d
el,
d
ataset,
an
d
f
ea
tu
r
e
s
u
tili
ze
d
.
W
h
ile
s
o
m
e
alg
o
r
ith
m
s
p
er
f
o
r
m
well
u
n
d
er
s
p
ec
if
ic
co
n
d
itio
n
s
,
th
ey
m
a
y
s
tr
u
g
g
le
i
n
o
th
e
r
s
,
m
a
k
in
g
a
u
n
iv
er
s
al
s
o
lu
tio
n
ch
allen
g
i
n
g
.
Ov
er
f
itti
n
g
an
d
u
n
d
er
f
itti
n
g
ar
e
co
m
m
o
n
is
s
u
es
th
at
ca
n
h
i
n
d
er
SER
ef
f
ec
tiv
en
ess
.
T
h
er
e
f
o
r
e,
ca
r
e
f
u
l
s
elec
tio
n
a
n
d
f
in
e
-
tu
n
in
g
o
f
alg
o
r
ith
m
s
ar
e
cr
u
cial
f
o
r
o
p
tim
izin
g
SER
s
y
s
tem
s
ac
co
r
d
in
g
to
th
e
task
an
d
d
ata
c
h
ar
ac
ter
is
tics
[
44
]
.
6.
ARAB
I
C
L
A
NG
UAG
E
-
B
AS
E
D
S
E
R
AP
P
RO
ACH
I
n
th
is
s
ec
tio
n
,
we
d
elv
e
in
t
o
s
ev
er
al
in
v
esti
g
atio
n
s
with
i
n
th
e
r
ea
lm
o
f
Ar
ab
ic
s
p
ee
ch
em
o
tio
n
r
ec
o
g
n
itio
n
.
Kh
alil
et
a
l.
[
4
5
]
f
o
cu
s
ed
o
n
d
etec
tin
g
a
n
g
er
in
h
u
m
an
-
h
u
m
an
d
ialo
g
u
es,
p
ar
ticu
lar
ly
in
ca
ll
ce
n
ter
s
,
u
s
in
g
class
if
ier
s
lik
e
SVM,
NB
,
KNN,
an
d
d
ec
is
i
o
n
tr
ee
(
DT
)
,
an
d
f
ea
tu
r
es
s
u
ch
as
f
u
n
d
a
m
en
tal
f
r
eq
u
e
n
cy
,
f
o
r
m
an
ts
,
en
e
r
g
y
,
an
d
MFC
C
,
ac
h
iev
in
g
7
7
%
ac
cu
r
ac
y
with
SVM.
Me
f
tah
et
a
l.
[
4
6
]
s
tu
d
ied
Ar
ab
ic
s
p
ee
ch
em
o
tio
n
r
ec
o
g
n
itio
n
,
an
aly
zin
g
em
o
tio
n
s
lik
e
s
ad
n
ess
,
h
ap
p
in
ess
,
an
d
an
g
er
with
r
h
y
t
h
m
m
etr
ics
an
d
th
e
KSUE
m
o
tio
n
s
co
r
p
u
s
,
f
i
n
d
in
g
th
at
s
ad
n
e
s
s
h
ad
th
e
h
ig
h
est
class
if
icatio
n
ac
cu
r
ac
y
u
s
in
g
m
u
ltil
ay
er
p
e
r
ce
p
tr
o
n
(
MLP
)
an
d
SVM.
Hif
n
y
a
n
d
Ali
[
4
7
]
e
n
h
an
ce
d
Ar
a
b
ic
s
p
ee
ch
e
m
o
tio
n
r
ec
o
g
n
itio
n
with
an
a
tten
tio
n
-
b
ased
C
NN
-
L
STM
-
DNN
m
o
d
el,
s
h
o
win
g
a
2
.
2
%
im
p
r
o
v
em
en
t
o
v
er
th
e
d
ee
p
C
NN
b
aselin
e.
C
h
er
if
et
a
l.
[
4
8
]
i
n
v
esti
g
ated
em
o
tio
n
d
etec
tio
n
i
n
th
e
Alg
er
ian
d
ialec
t,
f
o
c
u
s
in
g
o
n
em
o
tio
n
s
lik
e
h
ap
p
y
,
an
g
r
y
,
n
eu
tr
al,
an
d
s
ad
.
T
h
ey
em
p
lo
y
ed
an
L
STM
-
C
NN
class
if
ier
with
MFC
C
as
th
e
f
ea
tu
r
e
ex
tr
ac
tio
n
tech
n
i
q
u
e,
u
s
in
g
a
m
an
u
ally
an
n
o
tated
c
o
r
p
u
s
o
f
Alg
er
ian
telev
is
io
n
b
r
o
a
d
ca
s
ts
.
T
h
e
m
o
d
el
ac
h
iev
ed
a
n
ac
cu
r
ac
y
o
f
9
3
.
3
4
%.
Alju
h
an
i
et
a
l.
[
4
9
]
d
ev
elo
p
ed
a
s
p
ee
c
h
em
o
tio
n
r
e
co
g
n
i
tio
n
s
y
s
tem
f
o
r
th
e
Sau
d
i
d
ialec
t,
u
s
in
g
SVM,
ML
P,
an
d
KNN
m
o
d
els
with
v
ar
io
u
s
f
ea
tu
r
e
ex
tr
ac
tio
n
m
eth
o
d
s
,
ac
h
ie
v
in
g
7
7
.
1
4
%
ac
c
u
r
ac
y
with
SVM.
Mo
h
am
ed
a
n
d
Aly
[
5
0
]
s
h
o
w
ed
im
p
r
o
v
ed
r
ec
o
g
n
itio
n
ac
cu
r
ac
y
with
ML
P
an
d
Bi
-
L
STM
m
o
d
els
o
n
th
e
B
A
VE
D
d
ataset.
T
ajals
ir
et
a
l.
[
5
1
]
u
s
ed
L
STM
an
d
C
NN
to
en
h
an
ce
em
o
tio
n
r
ec
o
g
n
itio
n
in
h
u
m
an
-
co
m
p
u
t
er
in
ter
ac
tio
n
.
Alam
r
i
an
d
Als
h
an
b
ar
i
[
5
2
]
r
ep
o
r
ted
9
5
%
ac
cu
r
ac
y
u
s
in
g
C
NN
with
MFC
C
o
n
an
Ar
ab
ic
Yo
u
T
u
b
e
d
ataset.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
C
o
n
ce
p
tio
n
o
f sp
ee
c
h
emo
tio
n
r
ec
o
g
n
itio
n
meth
o
d
s
:
a
r
ev
iew
(
A
b
d
elka
d
er B
en
z
ir
a
r
)
1861
7.
RE
SU
L
T
S
AND
D
I
SCU
SS
I
O
N
T
ab
le
1
illu
s
tr
ates
th
e
ev
o
lu
ti
o
n
o
f
f
ea
tu
r
e
ex
tr
ac
tio
n
tech
n
i
q
u
es
an
d
m
eth
o
d
s
u
s
ed
in
SE
R
s
y
s
tem
s
.
T
h
e
ea
r
lies
t
s
tu
d
y
in
2
0
0
1
u
ti
lized
b
asic
f
ea
tu
r
es
s
u
ch
as
p
i
tch
an
d
e
n
er
g
y
co
m
b
in
ed
wit
h
HM
M,
ac
h
iev
in
g
an
ac
cu
r
ac
y
o
f
8
0
%.
As
r
esea
r
ch
,
p
r
o
g
r
ess
ed
,
m
o
r
e
s
o
p
h
is
ticated
f
ea
tu
r
es
a
n
d
co
m
b
i
n
atio
n
s
wer
e
ex
p
lo
r
ed
.
Fo
r
ex
am
p
le,
a
2
0
0
5
s
tu
d
y
in
t
r
o
d
u
ce
d
a
r
an
g
e
o
f
f
ea
tu
r
es
in
clu
d
in
g
en
er
g
y
,
f
u
n
d
am
en
tal
f
r
eq
u
en
c
y
,
MFC
C
s
,
an
d
f
o
r
m
an
t
f
r
eq
u
en
cies,
co
u
p
led
with
HM
M
an
d
SVM,
l
ea
d
in
g
to
r
em
ar
k
ab
le
ac
c
u
r
ac
y
r
ates,
with
u
p
to
1
0
0
%
f
o
r
m
ale
s
p
ea
k
er
s
.
Fr
o
m
2
0
1
7
o
n
war
d
s
,
d
ee
p
lear
n
in
g
m
eth
o
d
s
b
ec
am
e
in
c
r
e
asin
g
ly
p
r
o
m
in
en
t.
A
2
0
1
7
s
tu
d
y
e
m
p
lo
y
i
n
g
D
NN
ac
h
iev
ed
9
6
.
9
7
%
ac
cu
r
a
cy
,
in
d
icatin
g
th
e
p
o
ten
tial
o
f
d
ee
p
lear
n
in
g
in
ca
p
tu
r
in
g
co
m
p
lex
e
m
o
tio
n
a
l
cu
es
in
s
p
ee
ch
.
T
h
e
s
h
if
t
to
war
d
s
d
ee
p
lear
n
in
g
co
n
tin
u
ed
,
with
th
e
in
tr
o
d
u
ctio
n
o
f
C
NN
in
2
0
2
0
,
wh
ich
d
em
o
n
s
tr
ated
a
n
im
p
r
e
s
s
iv
e
r
an
g
e
o
f
ac
cu
r
ac
ies
b
etw
ee
n
9
2
%
a
n
d
9
8
%.
T
h
e
ad
o
p
tio
n
o
f
a
d
v
an
ce
d
tech
n
iq
u
es
s
u
ch
as
3
D
lo
g
m
el
-
s
p
ec
tr
o
g
r
a
m
s
in
2
0
2
1
,
co
m
b
in
e
d
with
s
o
p
h
is
ticated
m
o
d
els
lik
e
d
ee
p
atten
tio
n
-
b
ased
d
ilated
co
n
v
o
l
u
tio
n
al
-
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
s
(
ADC
R
NN)
,
f
u
r
th
er
en
h
an
ce
d
p
er
f
o
r
m
a
n
ce
,
ac
h
ie
v
in
g
an
ac
c
u
r
ac
y
o
f
u
p
t
o
8
8
.
4
7
%
with
GAN
-
b
ased
m
eth
o
d
s
.
T
h
e
tr
en
d
to
war
d
s
lev
er
ag
in
g
d
ee
p
lear
n
in
g
m
o
d
els is
ev
id
en
t in
2
0
2
2
an
d
2
0
2
3
,
wh
er
e
m
eth
o
d
s
lik
e
w
av
2
v
e
c
2
.
0
,
C
ap
s
Net,
an
d
h
y
b
r
id
ap
p
r
o
ac
h
es
ac
h
iev
ed
ac
cu
r
ac
ies
ex
ce
ed
in
g
9
3
%.
T
h
e
tab
le
h
ig
h
lig
h
ts
a
clea
r
tr
en
d
t
o
war
d
s
th
e
in
cr
ea
s
in
g
ad
o
p
tio
n
o
f
d
ee
p
le
ar
n
in
g
m
o
d
els an
d
m
o
r
e
c
o
m
p
lex
f
ea
tu
r
e
s
ets in
r
ec
en
t
y
ea
r
s
.
T
ab
le
1
.
Su
m
m
a
r
y
o
f
s
p
ee
ch
e
m
o
tio
n
r
ec
o
g
n
itio
n
r
esear
ch
es
Ref
Y
e
a
r
F
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
t
e
c
h
n
i
q
u
e
s
M
e
t
h
o
d
s
R
e
s
u
l
t
s
[
4
]
2
0
0
1
P
i
t
c
h
a
n
d
e
n
e
r
g
y
H
M
M
8
0
%
[
5
]
2
0
0
5
En
e
r
g
y
,
f
u
n
d
a
m
e
n
t
a
l
f
r
e
q
u
e
n
c
y
(
F
0
)
,
M
F
C
C
1
a
n
d
M
F
C
C
2
,
t
h
e
f
i
r
st
f
o
u
r
f
o
r
man
t
f
r
e
q
u
e
n
c
i
e
s (F
1
t
o
F
4
)
,
a
n
d
f
i
v
e
m
e
l
f
r
e
q
u
e
n
c
y
s
u
b
-
b
a
n
d
e
n
e
r
g
i
e
s
(
M
B
E
1
t
o
M
B
E
5
)
H
M
M
a
n
d
S
V
M
9
8
.
9
%
(
f
e
m
a
l
e
)
1
0
0
%
(
ma
l
e
)
9
9
.
5
%
(
g
e
n
d
e
r
-
i
n
d
e
p
e
n
d
e
n
t
)
[
8
]
2
0
1
7
-
DNN
9
6
.
9
7
%
[
1
5
]
2
0
1
4
A
f
f
e
c
t
-
sal
i
e
n
t
f
e
a
t
u
r
e
s
C
N
N
9
2
%
-
9
8
%
[
1
9
]
2
0
2
1
3
D
l
o
g
m
el
-
s
p
e
c
t
r
o
g
r
a
m
(
M
e
l
S
p
e
c
)
A
D
C
R
N
N
a
n
d
H
A
D
(
h
y
b
r
i
d
d
a
t
a
a
u
g
me
n
t
a
t
i
o
n
)
8
7
.
1
2
%
(
t
r
a
d
i
t
i
o
n
a
l
met
h
o
d
)
8
8
.
4
7
%
(
G
A
N
-
b
a
se
d
met
h
o
d
)
[
1
6
]
2
0
2
2
M
F
C
C
,
LPC,
e
n
e
r
g
y
,
s
p
e
c
t
r
a
l
f
l
o
w
,
sp
e
c
t
r
a
l
c
e
n
t
r
o
i
d
,
s
p
e
c
t
r
a
l
a
t
t
e
n
u
a
t
i
o
n
a
n
d
z
e
r
o
c
r
o
ssi
n
g
K
N
N
,
S
V
M
a
n
d
R
F
7
6
.
5
%
[
1
2
]
2
0
2
3
-
W
a
v
2
v
e
c
2
.
0
8
1
%
-
8
3
%
[
1
4
]
2
0
2
3
-
X
G
B
o
o
st
,
R
F
,
F
C
N
,
w
a
v
2
v
e
c
2
.
0
,
U
n
i
S
p
e
e
c
h
-
S
A
T
,
d
a
t
a
2
v
e
c
,
w
a
v
L
M
,
w
a
v
2
c
l
i
p
,
Y
A
M
N
e
t
,
E
C
A
P
A
a
n
d
x
-
v
e
c
t
o
r
-
[
1
7
]
2
0
2
3
M
F
C
C
C
N
N
a
n
d
B
LS
TM
6
7
.
4
4
%
[
1
8
]
2
0
2
3
-
X
LS
-
R
5
3
,
X
LS
-
R
5
3
S
P
(
S
p
a
n
i
s
h
)
,
X
LS
-
R
5
3
EN
(
E
n
g
l
i
s
h
)
,
X
LS
R
-
3
0
0
M
,
X
LSR
-
1
B
,
X
LSR
-
1
B
EN
(
En
g
l
i
s
h
)
X
LS
R
-
1
B
S
P
(
S
p
a
n
i
s
h
)
,
X
LSR
-
2
B
a
n
d
w
a
v
2
v
e
c
2
.
0
S
p
e
a
r
m
a
n
r
a
n
k
c
o
r
r
e
l
a
t
i
o
n
c
o
e
f
f
i
c
i
e
n
t
:
0
.
5
3
7
(
t
e
s
t
se
t
)
0
.
5
2
4
(
v
a
l
i
d
a
t
i
o
n
se
t
)
[
2
0
]
2
0
2
3
M
F
C
C
a
n
d
Z
C
R
C
a
p
sN
e
t
a
n
d
d
a
t
a
a
u
g
me
n
t
a
t
i
o
n
9
3
.
3
3
%
T
ab
le
2
p
r
esen
ts
a
clea
r
p
r
o
g
r
ess
io
n
in
th
e
tech
n
iq
u
es
an
d
m
et
h
o
d
s
u
s
ed
f
o
r
s
p
ee
ch
em
o
tio
n
r
ec
o
g
n
itio
n
in
Ar
ab
ic
lan
g
u
a
g
e
f
r
o
m
2
0
1
8
to
2
0
2
3
.
E
ar
l
y
s
tu
d
ies,
s
u
ch
as
th
e
o
n
e
i
n
2
0
1
8
,
r
elied
o
n
tr
ad
itio
n
al
m
ac
h
in
e
lear
n
in
g
m
o
d
els
lik
e
SVM,
DT
,
KNN,
an
d
NB
,
with
MF
C
C
an
d
b
asic
ac
o
u
s
tic
f
ea
tu
r
es,
ac
h
iev
in
g
a
m
ax
im
u
m
ac
cu
r
ac
y
o
f
7
7
.
2
%.
As
tim
e
p
r
o
g
r
ess
ed
,
m
o
r
e
s
o
p
h
is
ticated
f
ea
tu
r
es
an
d
d
ee
p
lear
n
i
n
g
m
o
d
els
wer
e
in
tr
o
d
u
ce
d
,
s
ig
n
if
ican
tly
im
p
r
o
v
in
g
p
e
r
f
o
r
m
an
ce
.
B
y
2
0
2
1
,
s
tu
d
ies
u
s
in
g
L
STM
-
C
NN
ar
ch
itectu
r
es
with
MFC
C
ac
h
iev
ed
m
u
c
h
h
i
g
h
er
ac
cu
r
ac
y
(
9
3
.
3
4
%),
in
d
icatin
g
th
e
s
u
p
er
io
r
ab
ilit
y
o
f
d
ee
p
lear
n
in
g
m
o
d
els
to
ca
p
tu
r
e
te
m
p
o
r
al
d
y
n
a
m
ics
in
s
p
ee
ch
.
T
h
e
tr
en
d
co
n
tin
u
e
d
in
to
2
0
2
2
an
d
2
0
2
3
,
wh
er
e
m
o
r
e
c
o
m
p
lex
f
ea
t
u
r
e
s
ets
li
k
e
ch
r
o
m
o
g
r
am
,
m
el
-
s
ca
led
s
p
ec
tr
o
g
r
a
m
,
a
n
d
ad
v
a
n
ce
d
d
e
ep
lear
n
i
n
g
m
o
d
els
lik
e
C
NN,
L
STM
,
an
d
DNN
s
wer
e
em
p
lo
y
ed
.
T
h
ese
ap
p
r
o
ac
h
es
ac
h
iev
ed
ev
en
h
i
g
h
er
a
cc
u
r
ac
ies,
s
u
ch
as
9
6
.
8
1
%
with
L
STM
a
n
d
9
5
%
with
C
NN,
h
ig
h
lig
h
tin
g
th
e
in
c
r
ea
s
in
g
e
f
f
ec
tiv
en
es
s
o
f
d
ee
p
lea
r
n
in
g
ap
p
r
o
ac
h
es
co
m
b
in
e
d
with
c
o
m
p
r
eh
e
n
s
iv
e
f
ea
tu
r
e
ex
tr
ac
ti
o
n
tech
n
i
q
u
es.
T
h
is
p
r
o
g
r
ess
io
n
u
n
d
er
s
co
r
es
th
e
im
p
o
r
tan
ce
o
f
b
o
th
ad
v
an
ce
d
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
d
ee
p
lear
n
in
g
m
eth
o
d
s
in
ac
h
iev
in
g
h
ig
h
ac
cu
r
ac
y
in
s
p
ee
ch
em
o
tio
n
r
ec
o
g
n
itio
n
ta
s
k
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
7
,
No
.
3
,
Ma
r
ch
20
2
5
:
1
85
6
-
1
86
4
1862
T
ab
le
2
.
Su
m
m
a
r
y
o
f
Ar
ab
ic
s
p
ee
ch
em
o
tio
n
r
ec
o
g
n
itio
n
r
esear
ch
es
Ref
Y
e
a
r
F
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
t
e
c
h
n
i
q
u
e
s
M
e
t
h
o
d
s
R
e
s
u
l
t
s
[
4
5
]
2
0
1
8
F
u
n
d
a
m
e
n
t
a
l
f
r
e
q
u
e
n
c
y
,
f
o
r
man
t
s,
e
n
e
r
g
y
a
n
d
M
F
C
C
S
V
M
,
D
T
,
K
N
N
a
n
d
N
B
7
7
.
2
%
(
t
h
e
b
e
st
a
c
c
u
r
a
c
y
w
i
t
h
S
V
M
)
[
4
6
]
2
0
2
0
P
i
t
c
h
,
i
n
t
e
n
s
i
t
y
,
f
o
r
ma
n
t
s,
j
i
t
t
e
r
,
sh
i
mm
e
r
,
h
a
r
m
o
n
i
c
s
-
to
-
n
o
i
se
r
a
t
i
o
a
n
d
r
h
y
t
h
m
me
t
r
i
c
s
M
LP
n
e
u
r
a
l
n
e
t
w
o
r
k
s a
n
d
S
V
M
P
h
a
se
1
:
4
9
%
(
w
i
t
h
M
LP)
5
2
%
(
w
i
t
h
S
V
M
)
P
h
a
se
2
:
8
3
.
6
7
%
(
w
i
t
h
M
LP)
8
3
.
9
5
%
(
w
i
t
h
S
V
M
)
[
4
7
]
2
0
2
0
M
F
C
C
a
n
d
L
o
g
M
e
l
-
f
i
l
t
e
r
b
a
n
k
e
n
e
r
g
i
e
s (LF
B
E)
A
t
t
e
n
t
i
o
n
-
b
a
se
d
C
N
N
-
LST
M
-
D
N
N
mo
d
e
l
a
n
d
d
e
e
p
C
N
N
mo
d
e
l
A
t
t
e
n
t
i
o
n
-
b
a
se
d
C
N
N
-
LST
M
-
DNN
mo
d
e
l
:
8
7
.
2
%
D
e
e
p
C
N
N
:
8
5
%
[
4
8
]
2
0
2
1
M
F
C
C
LSTM
-
C
N
N
9
3
.
3
4
%
[
4
9
]
2
0
2
1
M
F
C
C
,
mel
sp
e
c
t
r
o
g
r
a
m
a
n
d
S
p
e
c
t
r
a
l
c
o
n
t
r
a
s
t
S
V
M
,
M
LP
,
a
n
d
K
N
N
7
7
.
1
4
%
(
b
e
st
a
c
c
u
r
a
c
y
o
b
t
a
i
n
e
d
w
i
t
h
S
V
M
)
[
5
0
]
2
0
2
1
w
a
v
2
v
e
c
2
.
0
a
n
d
H
u
B
E
R
T
M
LP
a
n
d
B
i
-
LS
TM
W
a
v
2
v
e
c
2
.
0
:
8
9
%
H
u
B
E
R
T
:
8
7
%
(
H
u
B
E
R
T
b
a
s
e
)
8
3
%
(
H
u
B
E
R
T
l
a
r
g
e
)
[
5
1
]
2
0
2
2
M
F
C
C
,
c
h
r
o
ma
g
r
a
m
,
m
e
l
-
sca
l
e
d
sp
e
c
t
r
o
g
r
a
m
,
s
p
e
c
t
r
a
l
c
o
n
t
r
a
s
t
a
n
d
t
o
n
a
l
c
e
n
t
r
o
i
d
f
e
a
t
u
r
e
s (
t
o
n
n
e
t
z
)
LSTM
a
n
d
D
N
N
LSTM
:
9
6
.
8
1
%
D
N
N
:
9
3
.
3
4
%
[
5
2
]
2
0
2
3
M
F
C
C
a
n
d
z
e
r
o
-
c
r
o
ss
i
n
g
r
a
t
e
(
ZC
R
)
M
a
c
h
i
n
e
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
ms
(
S
V
M
a
n
d
K
N
N
)
.
D
e
e
p
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
ms
(
C
N
N
a
n
d
LST
M
)
.
9
5
%
(
C
N
N
w
i
t
h
M
F
C
C
)
8.
CO
NCLU
SI
O
N
R
ec
en
t
ad
v
an
ce
m
en
ts
in
SER
u
n
d
er
s
co
r
e
th
e
im
p
o
r
tan
ce
o
f
d
ee
p
lear
n
in
g
tech
n
iq
u
es,
esp
ec
ially
wh
en
p
air
ed
with
ad
v
a
n
ce
d
f
ea
tu
r
e
ex
tr
ac
tio
n
m
et
h
o
d
s
lik
e
MFC
C
,
C
NN
-
L
STM
ar
ch
itectu
r
es,
an
d
s
elf
-
s
u
p
er
v
is
ed
m
o
d
els
s
u
ch
as
wav
2
v
ec
2
.
0
an
d
Hu
B
E
R
T
.
T
h
ese
ap
p
r
o
ac
h
es
s
ig
n
if
ican
tly
o
u
tp
er
f
o
r
m
tr
ad
itio
n
al
m
eth
o
d
s
,
d
eliv
er
i
n
g
h
ig
h
er
ac
cu
r
ac
y
an
d
ca
p
tu
r
i
n
g
th
e
n
u
an
ce
s
o
f
em
o
tio
n
al
s
p
ee
c
h
.
T
h
is
s
h
if
t
f
r
o
m
tr
ad
itio
n
al
m
ac
h
i
n
e
lear
n
i
n
g
to
d
ee
p
lear
n
i
n
g
h
as
led
t
o
m
ajo
r
im
p
r
o
v
em
e
n
ts
in
S
E
R
'
s
ac
cu
r
ac
y
an
d
ad
ap
tab
ilit
y
ac
r
o
s
s
v
ar
io
u
s
la
n
g
u
ag
es
a
n
d
c
o
n
tex
ts
.
Ou
r
f
u
tu
r
e
r
esear
ch
will
f
o
cu
s
o
n
d
ev
elo
p
in
g
a
s
p
ee
c
h
em
o
tio
n
r
ec
o
g
n
izer
f
o
r
t
h
e
A
m
az
ig
h
lan
g
u
ag
e
,
ad
d
r
ess
in
g
i
ts
u
n
iq
u
e
p
h
o
n
etic
an
d
p
r
o
s
o
d
ic
f
ea
tu
r
es.
W
e
aim
to
cr
ea
te
a
co
m
p
r
e
h
en
s
iv
e
A
m
az
ig
h
d
ataset
an
d
ex
p
er
im
en
t
with
ad
v
an
ce
d
d
ee
p
lear
n
in
g
an
d
h
y
b
r
i
d
m
o
d
els.
C
r
o
s
s
-
lin
g
u
is
tic
an
aly
s
is
will
ex
p
lo
r
e
m
o
d
el
t
r
an
s
f
er
ab
ilit
y
b
etwe
en
lan
g
u
ag
es
lik
e
Ar
ab
ic
an
d
Am
az
ig
h
,
p
o
ten
tially
co
n
tr
ib
u
tin
g
to
m
o
r
e
u
n
iv
er
s
al
SER
s
y
s
tem
s
an
d
ex
p
an
d
in
g
th
eir
ap
p
licat
io
n
ac
r
o
s
s
d
iv
e
r
s
e
lin
g
u
is
tic
en
v
ir
o
n
m
en
ts
.
RE
F
E
R
E
NC
E
S
[
1
]
P
.
N
a
n
d
w
a
n
i
a
n
d
R
.
V
e
r
ma,
“
A
r
e
v
i
e
w
o
n
se
n
t
i
me
n
t
a
n
a
l
y
s
i
s
a
n
d
e
m
o
t
i
o
n
d
e
t
e
c
t
i
o
n
f
r
o
m
t
e
x
t
,
”
S
o
c
i
a
l
N
e
t
w
o
rk
An
a
l
y
si
s
a
n
d
Mi
n
i
n
g
,
v
o
l
.
1
1
,
n
o
.
1
,
p
.
8
1
,
D
e
c
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
3
2
7
8
-
0
2
1
-
0
0
7
7
6
-
6.
[
2
]
T.
M
.
W
a
n
i
,
T
.
S
.
G
u
n
a
w
a
n
,
S
.
A
.
A
.
Q
a
d
r
i
,
M
.
K
a
r
t
i
w
i
,
a
n
d
E.
A
mb
i
k
a
i
r
a
j
a
h
,
“
A
c
o
m
p
r
e
h
e
n
si
v
e
r
e
v
i
e
w
o
f
s
p
e
e
c
h
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
sy
s
t
e
m
s,”
I
EEE
Ac
c
e
ss
,
v
o
l
.
9
,
p
p
.
4
7
7
9
5
–
4
7
8
1
4
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
1
.
3
0
6
8
0
4
5
.
[
3
]
I
.
S
.
En
g
b
e
r
g
,
A
.
V
.
H
a
n
se
n
,
O
.
A
n
d
e
r
sen
,
a
n
d
P
.
D
a
l
s
g
a
a
r
d
,
“
D
e
s
i
g
n
,
r
e
c
o
r
d
i
n
g
a
n
d
v
e
r
i
f
i
c
a
t
i
o
n
o
f
a
d
a
n
i
s
h
e
mo
t
i
o
n
a
l
sp
e
e
c
h
d
a
t
a
b
a
se
,
”
i
n
5
t
h
E
u
r
o
p
e
a
n
C
o
n
f
e
re
n
c
e
o
n
S
p
e
e
c
h
C
o
m
m
u
n
i
c
a
t
i
o
n
a
n
d
T
e
c
h
n
o
l
o
g
y
(
E
u
ro
s
p
e
e
c
h
1
9
9
7
)
,
S
e
p
.
1
9
9
7
,
p
p
.
1
6
9
5
–
1
6
9
8
,
d
o
i
:
1
0
.
2
1
4
3
7
/
E
u
r
o
s
p
e
e
c
h
.
1
9
9
7
-
4
8
2
.
[
4
]
A
.
N
o
g
u
e
i
r
a
s,
A
.
M
o
r
e
n
o
,
A
.
B
o
n
a
f
o
n
t
e
,
a
n
d
J
.
B
.
M
a
r
i
ñ
o
,
“
S
p
e
e
c
h
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
u
s
i
n
g
h
i
d
d
e
n
m
a
r
k
o
v
mo
d
e
l
s,
”
i
n
7
t
h
Eu
r
o
p
e
a
n
C
o
n
f
e
r
e
n
c
e
o
n
S
p
e
e
c
h
C
o
m
m
u
n
i
c
a
t
i
o
n
a
n
d
T
e
c
h
n
o
l
o
g
y
(
Eu
r
o
sp
e
e
c
h
2
0
0
1
)
,
S
e
p
.
2
0
0
1
,
p
p
.
2
6
7
9
–
2
6
8
2
,
d
o
i
:
1
0
.
2
1
4
3
7
/
Eu
r
o
s
p
e
e
c
h
.
2
0
0
1
-
6
2
7
.
[
5
]
Y
.
Li
n
a
n
d
G
.
W
e
i
,
“
S
p
e
e
c
h
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
b
a
s
e
d
o
n
H
M
M
a
n
d
S
V
M
,
”
i
n
2
0
0
5
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
a
n
d
C
y
b
e
r
n
e
t
i
c
s
,
2
0
0
5
,
p
p
.
4
8
9
8
–
4
9
0
1
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
M
L
C
.
2
0
0
5
.
1
5
2
7
8
0
5
.
[
6
]
K
.
H
.
H
y
u
n
,
E
.
H
.
K
i
m,
a
n
d
Y
.
K
.
K
w
a
k
,
“
Emo
t
i
o
n
a
l
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
b
a
se
d
o
n
p
h
o
n
e
m
e
i
n
f
o
r
ma
t
i
o
n
f
o
r
sp
e
e
c
h
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
,
”
i
n
RO
-
MA
N
2
0
0
7
-
T
h
e
1
6
t
h
I
EE
E
I
n
t
e
r
n
a
t
i
o
n
a
l
S
y
m
p
o
s
i
u
m
o
n
Ro
b
o
t
a
n
d
H
u
m
a
n
I
n
t
e
ra
c
t
i
v
e
C
o
m
m
u
n
i
c
a
t
i
o
n
,
2
0
0
7
,
p
p
.
8
0
2
–
8
0
6
,
d
o
i
:
1
0
.
1
1
0
9
/
R
O
M
A
N
.
2
0
0
7
.
4
4
1
5
1
9
5
.
[
7
]
Y
.
K
i
m
,
H
.
Le
e
,
a
n
d
E.
M
.
P
r
o
v
o
s
t
,
“
D
e
e
p
l
e
a
r
n
i
n
g
f
o
r
r
o
b
u
st
f
e
a
t
u
r
e
g
e
n
e
r
a
t
i
o
n
i
n
a
u
d
i
o
v
i
s
u
a
l
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
,
”
i
n
2
0
1
3
I
EEE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
Ac
o
u
st
i
c
s,
S
p
e
e
c
h
a
n
d
S
i
g
n
a
l
Pro
c
e
ssi
n
g
,
M
a
y
2
0
1
3
,
p
p
.
3
6
8
7
–
3
6
9
1
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
S
S
P
.
2
0
1
3
.
6
6
3
8
3
4
6
.
[
8
]
P
.
H
a
r
a
r
,
R
.
B
u
r
g
e
t
,
a
n
d
M
.
K
.
D
u
t
t
a
,
“
S
p
e
e
c
h
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
w
i
t
h
d
e
e
p
l
e
a
r
n
i
n
g
,
”
i
n
2
0
1
7
4
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
S
i
g
n
a
l
Pr
o
c
e
ssi
n
g
a
n
d
I
n
t
e
g
r
a
t
e
d
N
e
t
w
o
r
k
s (
S
PI
N
)
,
F
e
b
.
2
0
1
7
,
p
p
.
1
3
7
–
1
4
0
,
d
o
i
:
1
0
.
1
1
0
9
/
S
P
I
N
.
2
0
1
7
.
8
0
4
9
9
3
1
.
[
9
]
B
.
T
.
A
t
ma
j
a
a
n
d
M
.
A
k
a
g
i
,
“
S
p
e
e
c
h
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
b
a
se
d
o
n
sp
e
e
c
h
seg
m
e
n
t
u
s
i
n
g
LST
M
w
i
t
h
a
t
t
e
n
t
i
o
n
m
o
d
e
l
,
”
i
n
2
0
1
9
I
EEE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
S
i
g
n
a
l
s
a
n
d
S
y
st
e
m
s (I
C
S
i
g
S
y
s)
,
J
u
l
.
2
0
1
9
,
p
p
.
4
0
–
4
4
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
S
I
G
S
Y
S
.
2
0
1
9
.
8
8
1
1
0
8
0
.
[
1
0
]
B
.
T.
A
t
m
a
j
a
a
n
d
A
.
S
a
s
o
u
,
“
Ef
f
e
c
t
s
o
f
d
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
s
o
n
s
p
e
e
c
h
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
,
”
S
e
n
s
o
rs
,
v
o
l
.
2
2
,
n
o
.
1
6
,
p
.
5
9
4
1
,
A
u
g
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
s2
2
1
6
5
9
4
1
.
[
1
1
]
B
.
T
.
A
t
maj
a
a
n
d
A
.
S
a
s
o
u
,
“
Ev
a
l
u
a
t
i
n
g
se
l
f
-
su
p
e
r
v
i
se
d
s
p
e
e
c
h
r
e
p
r
e
s
e
n
t
a
t
i
o
n
s
f
o
r
sp
e
e
c
h
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
0
,
p
p
.
1
2
4
3
9
6
–
1
2
4
4
0
7
,
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
2
.
3
2
2
5
1
9
8
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
C
o
n
ce
p
tio
n
o
f sp
ee
c
h
emo
tio
n
r
ec
o
g
n
itio
n
meth
o
d
s
:
a
r
ev
iew
(
A
b
d
elka
d
er B
en
z
ir
a
r
)
1863
[
1
2
]
F
.
C
a
t
a
n
i
a
,
“
S
p
e
e
c
h
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
i
n
I
t
a
l
i
a
n
u
s
i
n
g
w
a
v
2
v
e
c
2
.
0
a
n
d
t
h
e
n
o
v
e
l
c
r
o
w
d
s
o
u
r
c
e
d
e
m
o
t
i
o
n
a
l
sp
e
e
c
h
c
o
r
p
u
s
e
mo
z
i
o
n
a
l
m
e
n
t
e
,
”
Au
t
h
o
r
e
a
Pre
p
r
i
n
t
s
,
M
a
y
2
0
2
3
,
d
o
i
:
1
0
.
3
6
2
2
7
/
t
e
c
h
r
x
i
v
.
2
2
8
2
1
9
9
2
.
v
1
.
[
1
3
]
O
.
C
.
P
h
u
k
a
n
,
A
.
B
.
B
u
d
u
r
u
,
a
n
d
R
.
S
h
a
r
ma,
“
Tr
a
n
sf
o
r
mi
n
g
t
h
e
e
m
b
e
d
d
i
n
g
s:
a
l
i
g
h
t
w
e
i
g
h
t
t
e
c
h
n
i
q
u
e
f
o
r
sp
e
e
c
h
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
t
a
sk
s
,
”
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
An
n
u
a
l
C
o
n
f
e
re
n
c
e
o
f
t
h
e
I
n
t
e
r
n
a
t
i
o
n
a
l
S
p
e
e
c
h
C
o
m
m
u
n
i
c
a
t
i
o
n
Ass
o
c
i
a
t
i
o
n
,
I
N
T
ERS
PEE
C
H
,
v
o
l
.
2
0
2
3
-
A
u
g
u
s
t
,
p
p
.
1
9
0
3
–
1
9
0
7
,
2
0
2
3
,
d
o
i
:
1
0
.
2
1
4
3
7
/
I
n
t
e
r
sp
e
e
c
h
.
2
0
2
3
-
2
5
6
1
.
[
1
4
]
O
.
C
.
P
h
u
k
a
n
,
A
.
B
.
B
u
d
u
r
u
,
a
n
d
R
.
S
h
a
r
m
a
,
“
A
c
o
m
p
a
r
a
t
i
v
e
st
u
d
y
o
f
p
r
e
-
t
r
a
i
n
e
d
sp
e
e
c
h
a
n
d
a
u
d
i
o
e
m
b
e
d
d
i
n
g
s
f
o
r
sp
e
e
c
h
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
,
”
a
rX
i
v
p
r
e
p
r
i
n
t
a
rXi
v
:
2
3
0
4
.
1
1
4
7
2
,
2
0
2
3
,
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
:
/
/
a
r
x
i
v
.
o
r
g
/
a
b
s/
2
3
0
4
.
1
1
4
7
2
.
[
1
5
]
Z.
H
u
a
n
g
,
M
.
D
o
n
g
,
Q
.
M
a
o
,
a
n
d
Y
.
Zh
a
n
,
“
S
p
e
e
c
h
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
u
s
i
n
g
C
N
N
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
2
2
n
d
AC
M
i
n
t
e
r
n
a
t
i
o
n
a
l
c
o
n
f
e
re
n
c
e
o
n
M
u
l
t
i
m
e
d
i
a
,
N
o
v
.
2
0
1
4
,
p
p
.
8
0
1
–
8
0
4
,
d
o
i
:
1
0
.
1
1
4
5
/
2
6
4
7
8
6
8
.
2
6
5
4
9
8
4
.
[
1
6
]
A
.
A
sg
h
a
r
,
S
.
S
o
h
a
i
b
,
S
.
I
f
t
i
k
h
a
r
,
M
.
S
h
a
f
i
,
a
n
d
K
.
F
a
t
i
ma
,
“
A
n
U
r
d
u
sp
e
e
c
h
c
o
r
p
u
s
f
o
r
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
,
”
Pe
e
r
J
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
8
,
p
.
e
9
5
4
,
M
a
y
2
0
2
2
,
d
o
i
:
1
0
.
7
7
1
7
/
p
e
e
r
j
-
c
s.
9
5
4
.
[
1
7
]
Y
.
X
i
a
a
n
d
L.
Z
h
a
o
,
“
C
N
N
-
B
L
S
TM
w
i
t
h
a
t
t
e
n
t
i
o
n
m
o
d
e
l
f
o
r
s
p
e
e
c
h
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
.
”
O
c
t
.
0
4
,
2
0
2
3
,
d
o
i
:
1
0
.
2
1
2
0
3
/
r
s
.
3
.
r
s
-
3
3
9
2
0
0
8
/
v
1
.
[
1
8
]
B
.
T
.
A
t
m
a
j
a
a
n
d
A
.
S
a
s
o
u
,
“
En
s
e
mb
l
i
n
g
m
u
l
t
i
l
i
n
g
u
a
l
p
r
e
-
t
r
a
i
n
e
d
m
o
d
e
l
s f
o
r
p
r
e
d
i
c
t
i
n
g
m
u
l
t
i
-
l
a
b
e
l
r
e
g
r
e
ss
i
o
n
e
m
o
t
i
o
n
s
h
a
r
e
f
r
o
m
sp
e
e
c
h
,
”
i
n
2
0
2
3
As
i
a
P
a
c
i
f
i
c
S
i
g
n
a
l
a
n
d
I
n
f
o
rm
a
t
i
o
n
Pro
c
e
ss
i
n
g
Ass
o
c
i
a
t
i
o
n
A
n
n
u
a
l
S
u
m
m
i
t
a
n
d
C
o
n
f
e
re
n
c
e
(
APS
I
P
A
A
S
C
)
,
O
c
t
.
2
0
2
3
,
p
p
.
1
0
2
6
–
1
0
2
9
,
d
o
i
:
1
0
.
1
1
0
9
/
A
P
S
I
P
A
A
S
C
5
8
5
1
7
.
2
0
2
3
.
1
0
3
1
7
1
0
9
.
[
1
9
]
N
.
T.
P
h
a
m
e
t
a
l
.
,
“
H
y
b
r
i
d
d
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
a
n
d
d
e
e
p
a
t
t
e
n
t
i
o
n
-
b
a
s
e
d
d
i
l
a
t
e
d
c
o
n
v
o
l
u
t
i
o
n
a
l
-
r
e
c
u
r
r
e
n
t
n
e
u
r
a
l
n
e
t
w
o
r
k
s fo
r
s
p
e
e
c
h
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
,
”
Ex
p
e
r
t
S
y
st
e
m
s
w
i
t
h
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
2
3
0
,
p
.
1
2
0
6
0
8
,
N
o
v
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
2
3
.
1
2
0
6
0
8
.
[
2
0
]
H
.
Zh
a
n
g
,
H
.
H
u
a
n
g
,
a
n
d
H
.
H
a
n
,
“
M
A
-
C
a
p
sN
e
t
-
D
A
:
S
p
e
e
c
h
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
b
a
s
e
d
o
n
M
A
-
C
a
p
sN
e
t
u
s
i
n
g
d
a
t
a
a
u
g
me
n
t
a
t
i
o
n
,
”
E
x
p
e
rt
S
y
st
e
m
s
w
i
t
h
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
2
4
4
,
p
.
1
2
2
9
3
9
,
Ju
n
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
2
3
.
1
2
2
9
3
9
.
[
2
1
]
M
.
J.
A
l
a
m
,
Y
.
A
t
t
a
b
i
,
P
.
D
u
m
o
u
c
h
e
l
,
P
.
K
e
n
n
y
,
a
n
d
D
.
O
’
S
h
a
u
g
h
n
e
ss
y
,
“
A
mp
l
i
t
u
d
e
m
o
d
u
l
a
t
i
o
n
f
e
a
t
u
r
e
s
f
o
r
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
f
r
o
m sp
e
e
c
h
,
”
i
n
I
n
t
e
rs
p
e
e
c
h
2
0
1
3
,
A
u
g
.
2
0
1
3
,
p
p
.
2
4
2
0
–
2
4
2
4
,
d
o
i
:
1
0
.
2
1
4
3
7
/
I
n
t
e
r
s
p
e
e
c
h
.
2
0
1
3
-
5
6
3
.
[
2
2
]
U
.
S
h
r
a
w
a
n
k
a
r
a
n
d
V
.
M
.
Th
a
k
a
r
e
,
“
Te
c
h
n
i
q
u
e
s
f
o
r
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
i
n
s
p
e
e
c
h
r
e
c
o
g
n
i
t
i
o
n
s
y
s
t
e
m
:
a
c
o
m
p
a
r
a
t
i
v
e
st
u
d
y
,
”
a
r
Xi
v
p
re
p
ri
n
t
a
rXi
v
:
1
3
0
5
.
1
1
4
5
,
2
0
1
3
,
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
:
/
/
a
r
x
i
v
.
o
r
g
/
a
b
s
/
1
3
0
5
.
1
1
4
5
.
[
2
3
]
H
.
G
u
p
t
a
a
n
d
D
.
G
u
p
t
a
,
“
LP
C
a
n
d
L
P
C
C
me
t
h
o
d
o
f
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
i
n
sp
e
e
c
h
r
e
c
o
g
n
i
t
i
o
n
sy
s
t
e
m
,
”
i
n
2
0
1
6
6
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
-
C
l
o
u
d
S
y
s
t
e
m
a
n
d
B
i
g
D
a
t
a
E
n
g
i
n
e
e
r
i
n
g
(
C
o
n
f
l
u
e
n
c
e
)
,
J
a
n
.
2
0
1
6
,
p
p
.
4
9
8
–
5
0
2
,
d
o
i
:
1
0
.
1
1
0
9
/
C
O
N
F
LU
EN
C
E.
2
0
1
6
.
7
5
0
8
1
7
1
.
[
2
4
]
F
.
R
e
g
g
i
s
w
a
r
a
s
h
a
r
i
a
n
d
S
.
W
.
S
i
h
w
i
,
“
S
p
e
e
c
h
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
u
s
i
n
g
2
D
-
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
E
l
e
c
t
r
i
c
a
l
a
n
d
C
o
m
p
u
t
e
r
E
n
g
i
n
e
e
r
i
n
g
(
I
J
E
C
E
)
,
v
o
l
.
1
2
,
n
o
.
6
,
p
p
.
6
5
9
4
–
6
6
0
1
,
D
e
c
.
2
0
2
2
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
c
e
.
v
1
2
i
6
.
p
p
6
5
9
4
-
6
6
0
1
.
[
2
5
]
H
.
A
o
u
a
n
i
a
n
d
Y
.
B
e
n
A
y
e
d
,
“
S
p
e
e
c
h
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
w
i
t
h
d
e
e
p
l
e
a
r
n
i
n
g
,
”
Pr
o
c
e
d
i
a
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
1
7
6
,
p
p
.
2
5
1
–
2
6
0
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
r
o
c
s.2
0
2
0
.
0
8
.
0
2
7
.
[
2
6
]
R
.
Y
.
R
u
m
a
g
i
t
,
G
.
A
l
e
x
a
n
d
e
r
,
a
n
d
I
.
F
.
S
a
p
u
t
r
a
,
“
M
o
d
e
l
c
o
m
p
a
r
i
s
o
n
i
n
s
p
e
e
c
h
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
f
o
r
I
n
d
o
n
e
si
a
n
l
a
n
g
u
a
g
e
,
”
Pro
c
e
d
i
a
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
1
7
9
,
p
p
.
7
8
9
–
7
9
7
,
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
r
o
c
s.
2
0
2
1
.
0
1
.
0
9
8
.
[
2
7
]
M
.
H
a
m
i
d
i
,
H
.
S
a
t
o
r
i
,
O
.
Z
e
a
l
o
u
k
,
a
n
d
K
.
S
a
t
o
r
i
,
“
A
m
a
z
i
g
h
d
i
g
i
t
s
t
h
r
o
u
g
h
i
n
t
e
r
a
c
t
i
v
e
s
p
e
e
c
h
r
e
c
o
g
n
i
t
i
o
n
s
y
s
t
e
m
i
n
n
o
i
s
y
e
n
v
i
r
o
n
m
e
n
t
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
S
p
e
e
c
h
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
2
3
,
n
o
.
1
,
p
p
.
1
0
1
–
1
0
9
,
M
a
r
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
7
7
2
-
019
-
0
9
6
6
1
-
2.
[
2
8
]
S
.
A
n
a
n
t
h
i
a
n
d
P
.
D
h
a
n
a
l
a
k
s
h
m
i
,
“
S
V
M
a
n
d
H
M
M
m
o
d
e
l
i
n
g
t
e
c
h
n
i
q
u
e
s
f
o
r
s
p
e
e
c
h
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
LP
C
C
a
n
d
M
F
C
C
f
e
a
t
u
r
e
s,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
3
r
d
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
Fr
o
n
t
i
e
rs
o
f
I
n
t
e
l
l
i
g
e
n
t
C
o
m
p
u
t
i
n
g
:
T
h
e
o
r
y
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
(
FI
C
T
A)
2
0
1
4
:
Vo
l
u
m
e
1
,
2
0
1
5
,
p
p
.
5
1
9
–
5
2
6
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
3
1
9
-
1
1
9
3
3
-
5
_
5
8
.
[
2
9
]
B
.
P
.
D
a
s
a
n
d
R
.
P
a
r
e
k
h
,
“
R
e
c
o
g
n
i
t
i
o
n
o
f
i
s
o
l
a
t
e
d
w
o
r
d
s
u
s
i
n
g
f
e
a
t
u
r
e
s
b
a
s
e
d
o
n
LPC
,
M
F
C
C
,
ZC
R
a
n
d
S
TE
,
w
i
t
h
n
e
u
r
a
l
n
e
t
w
o
r
k
c
l
a
ss
i
f
i
e
r
s
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
Mo
d
e
rn
E
n
g
i
n
e
e
r
i
n
g
Re
se
a
r
c
h
(
I
J
M
ER)
,
v
o
l
.
2
,
n
o
.
3
,
p
p
.
8
5
4
–
8
5
8
,
2
0
1
2
.
[
3
0
]
S
.
B
o
d
a
p
a
t
i
,
H
.
B
a
n
d
a
r
u
p
a
l
l
y
,
R
.
N
.
S
h
a
w
,
a
n
d
A
.
G
h
o
s
h
,
“
C
o
mp
a
r
i
s
o
n
a
n
d
a
n
a
l
y
s
i
s
o
f
R
N
N
-
LSTM
s
a
n
d
C
N
N
s
f
o
r
so
c
i
a
l
r
e
v
i
e
w
s c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
i
n
A
d
v
a
n
c
e
s
i
n
A
p
p
l
i
c
a
t
i
o
n
s
o
f
D
a
t
a
-
D
r
i
v
e
n
C
o
m
p
u
t
i
n
g
,
2
0
2
1
,
p
p
.
4
9
–
5
9
.
[
3
1
]
S
.
M
a
o
,
D
.
T
a
o
,
G
.
Z
h
a
n
g
,
P
.
C
.
C
h
i
n
g
,
a
n
d
T.
Le
e
,
“
R
e
v
i
s
i
t
i
n
g
h
i
d
d
e
n
mar
k
o
v
mo
d
e
l
s
f
o
r
s
p
e
e
c
h
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
,
”
i
n
I
C
AS
S
P
2
0
1
9
-
2
0
1
9
I
EEE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Ac
o
u
st
i
c
s,
S
p
e
e
c
h
a
n
d
S
i
g
n
a
l
Pr
o
c
e
ss
i
n
g
(
I
C
A
S
S
P)
,
M
a
y
2
0
1
9
,
p
p
.
6
7
1
5
–
6
7
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
S
S
P
.
2
0
1
9
.
8
6
8
3
1
7
2
.
[
3
2
]
T.
L.
N
w
e
,
S
.
W
.
F
o
o
,
a
n
d
L.
C
.
D
e
S
i
l
v
a
,
“
D
e
t
e
c
t
i
o
n
o
f
s
t
r
e
ss
a
n
d
e
mo
t
i
o
n
i
n
sp
e
e
c
h
u
si
n
g
t
r
a
d
i
t
i
o
n
a
l
a
n
d
F
F
T
b
a
s
e
d
l
o
g
e
n
e
r
gy
f
e
a
t
u
r
e
s,
”
i
n
F
o
u
r
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
I
n
f
o
rm
a
t
i
o
n
,
C
o
m
m
u
n
i
c
a
t
i
o
n
s
a
n
d
S
i
g
n
a
l
Pr
o
c
e
s
si
n
g
,
2
0
0
3
a
n
d
t
h
e
F
o
u
rt
h
Pa
c
i
f
i
c
R
i
m
C
o
n
f
e
re
n
c
e
o
n
M
u
l
t
i
m
e
d
i
a
.
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
2
0
0
3
J
o
i
n
t
,
2
0
0
3
,
v
o
l
.
3
,
p
p
.
1
6
1
9
–
1
6
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
I
C
S
.
2
0
0
3
.
1
2
9
2
7
4
1
.
[
3
3
]
M
.
H
a
mi
d
i
,
O
.
Z
e
a
l
o
u
k
,
H
.
S
a
t
o
r
i
,
N
.
La
a
i
d
i
,
a
n
d
A
.
S
a
l
e
k
,
“
C
O
V
I
D
-
1
9
a
ss
e
ssm
e
n
t
u
si
n
g
H
M
M
c
o
u
g
h
r
e
c
o
g
n
i
t
i
o
n
s
y
st
e
m,”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
I
n
f
o
rm
a
t
i
o
n
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
1
5
,
n
o
.
1
,
p
p
.
1
9
3
–
2
0
1
,
Ja
n
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s4
1
8
7
0
-
0
2
2
-
0
1
1
2
0
-
7.
[
3
4
]
Y
.
P
a
n
,
P
.
S
h
e
n
,
a
n
d
L
.
S
h
e
n
,
“
F
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
a
n
d
se
l
e
c
t
i
o
n
i
n
s
p
e
e
c
h
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
,
”
2
0
0
5
.
[
3
5
]
F
.
M
.
J.
M
e
h
e
d
i
S
h
a
mr
a
t
e
t
a
l
.
,
“
S
e
n
t
i
men
t
a
n
a
l
y
si
s
o
n
t
w
i
t
t
e
r
t
w
e
e
t
s
a
b
o
u
t
C
O
V
I
D
-
1
9
v
a
c
c
i
n
e
s u
si
n
g
N
LP
a
n
d
su
p
e
r
v
i
se
d
K
N
N
c
l
a
ss
i
f
i
c
a
t
i
o
n
a
l
g
o
r
i
t
h
m,
”
I
n
d
o
n
e
si
a
n
J
o
u
r
n
a
l
o
f
El
e
c
t
r
i
c
a
l
En
g
i
n
e
e
r
i
n
g
a
n
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
(
I
J
EEC
S
)
,
v
o
l
.
2
3
,
n
o
.
1
,
p
p
.
4
6
3
–
4
7
0
,
J
u
l
.
2
0
2
1
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
e
c
s.
v
2
3
.
i
1
.
p
p
4
6
3
-
4
7
0
.
[
3
6
]
M
.
I
q
b
a
l
,
S
.
A
l
i
,
M
.
A
b
i
d
,
F
.
M
a
j
e
e
d
,
a
n
d
A
.
A
l
i
,
“
A
r
t
i
f
i
c
i
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
b
a
se
d
e
m
o
t
i
o
n
c
l
a
ssi
f
i
c
a
t
i
o
n
a
n
d
r
e
c
o
g
n
i
t
i
o
n
f
r
o
m
sp
e
e
c
h
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
Ad
v
a
n
c
e
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
1
1
,
n
o
.
1
2
,
2
0
2
0
,
d
o
i
:
1
0
.
1
4
5
6
9
/
I
JA
C
S
A
.
2
0
2
0
.
0
1
1
1
2
5
3
.
[
3
7
]
I
.
Zy
o
u
t
a
n
d
M
.
Zy
o
u
t
,
“
S
e
n
t
i
men
t
a
n
a
l
y
si
s
o
f
st
u
d
e
n
t
f
e
e
d
b
a
c
k
u
s
i
n
g
a
t
t
e
n
t
i
o
n
-
b
a
s
e
d
R
N
N
a
n
d
t
r
a
n
s
f
o
r
m
e
r
e
m
b
e
d
d
i
n
g
,
”
I
A
ES
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
A
r
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
(
I
J
-
A
I
)
,
v
o
l
.
1
3
,
n
o
.
2
,
p
p
.
2
1
7
3
–
2
1
8
4
,
J
u
n
.
2
0
2
4
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
a
i
.
v
1
3
.
i
2
.
p
p
2
1
7
3
-
2
1
8
4
.
[
3
8
]
M
.
O
.
A
d
e
b
i
y
i
,
T.
T.
A
d
e
l
i
y
i
,
D
.
O
l
a
n
i
y
a
n
,
a
n
d
J.
O
l
a
n
i
y
a
n
,
“
A
d
v
a
n
c
e
m
e
n
t
s
i
n
a
c
c
u
r
a
t
e
s
p
e
e
c
h
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
t
h
r
o
u
g
h
t
h
e
i
n
t
e
g
r
a
t
i
o
n
o
f
C
N
N
-
A
M
m
o
d
e
l
,
”
T
ELKO
M
N
I
K
A
(
T
e
l
e
c
o
m
m
u
n
i
c
a
t
i
o
n
C
o
m
p
u
t
i
n
g
El
e
c
t
ro
n
i
c
s
a
n
d
C
o
n
t
r
o
l
)
,
v
o
l
.
2
2
,
n
o
.
3
,
p
p
.
6
0
6
–
6
1
8
,
J
u
n
.
2
0
2
4
,
d
o
i
:
1
0
.
1
2
9
2
8
/
t
e
l
k
o
m
n
i
k
a
.
v
2
2
i
3
.
2
5
7
0
8
.
[
3
9
]
H
.
B
o
u
l
a
l
,
M
.
H
a
mi
d
i
,
M
.
A
b
a
r
k
a
n
,
a
n
d
J.
B
a
r
k
a
n
i
,
“
A
m
a
z
i
g
h
C
N
N
sp
e
e
c
h
r
e
c
o
g
n
i
t
i
o
n
sy
st
e
m
b
a
se
d
o
n
m
e
l
sp
e
c
t
r
o
g
r
a
m
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
m
e
t
h
o
d
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
S
p
e
e
c
h
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
2
7
,
n
o
.
1
,
p
p
.
2
8
7
–
2
9
6
,
M
a
r
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
0
7
7
2
-
024
-
1
0
1
0
0
-
0.
[
4
0
]
H
.
A
si
l
a
n
d
J
.
B
a
g
h
e
r
z
a
d
e
h
,
“
P
r
o
p
o
si
n
g
a
n
e
w
m
e
t
h
o
d
o
f
i
ma
g
e
c
l
a
ssi
f
i
c
a
t
i
o
n
b
a
s
e
d
o
n
t
h
e
A
d
a
B
o
o
s
t
d
e
e
p
b
e
l
i
e
f
n
e
t
w
o
r
k
h
y
b
r
i
d
met
h
o
d
,
”
T
EL
K
O
MN
I
K
A
(
T
e
l
e
c
o
m
m
u
n
i
c
a
t
i
o
n
C
o
m
p
u
t
i
n
g
E
l
e
c
t
r
o
n
i
c
s
a
n
d
C
o
n
t
r
o
l
)
,
v
o
l
.
1
7
,
n
o
.
5
,
p
p
.
2
6
5
0
–
2
6
5
8
,
O
c
t
.
2
0
1
9
,
d
o
i
:
1
0
.
1
2
9
2
8
/
t
e
l
k
o
m
n
i
k
a
.
v
1
7
i
5
.
1
1
7
9
7
.
[
4
1
]
S
.
M
a
r
i
o
o
r
y
a
d
a
n
d
C
.
B
u
sso
,
“
C
o
m
p
e
n
sa
t
i
n
g
f
o
r
sp
e
a
k
e
r
o
r
l
e
x
i
c
a
l
v
a
r
i
a
b
i
l
i
t
i
e
s
i
n
sp
e
e
c
h
f
o
r
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
,
”
S
p
e
e
c
h
C
o
m
m
u
n
i
c
a
t
i
o
n
,
v
o
l
.
5
7
,
p
p
.
1
–
1
2
,
F
e
b
.
2
0
1
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
s
p
e
c
o
m
.
2
0
1
3
.
0
7
.
0
1
1
.
[
4
2
]
A
.
S
h
i
l
a
n
d
a
r
i
,
H
.
M
a
r
v
i
,
H
.
K
h
o
sr
a
v
i
,
a
n
d
W
.
W
a
n
g
,
“
S
p
e
e
c
h
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
d
a
t
a
a
u
g
me
n
t
a
t
i
o
n
met
h
o
d
b
y
c
y
c
l
e
-
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sar
i
a
l
n
e
t
w
o
r
k
s
,
”
S
i
g
n
a
l
,
I
m
a
g
e
a
n
d
V
i
d
e
o
Pr
o
c
e
ssi
n
g
,
v
o
l
.
1
6
,
n
o
.
7
,
p
p
.
1
9
5
5
–
1
9
6
2
,
O
c
t
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
7
6
0
-
0
2
2
-
0
2
1
5
6
-
9.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
In
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
3
7
,
No
.
3
,
Ma
r
ch
20
2
5
:
1
85
6
-
1
86
4
1864
[
4
3
]
R
.
Ja
h
a
n
g
i
r
,
Y
.
W
.
Te
h
,
F
.
H
a
n
i
f
,
a
n
d
G
.
M
u
j
t
a
b
a
,
“
D
e
e
p
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
e
s
f
o
r
sp
e
e
c
h
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
:
st
a
t
e
o
f
t
h
e
a
r
t
a
n
d
r
e
sea
r
c
h
c
h
a
l
l
e
n
g
e
s
,
”
Mu
l
t
i
m
e
d
i
a
T
o
o
l
s
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
8
0
,
n
o
.
1
6
,
p
p
.
2
3
7
4
5
–
2
3
8
1
2
,
Ju
l
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
1
0
4
2
-
020
-
0
9
8
7
4
-
7.
[
4
4
]
M
.
B
.
A
k
ç
a
y
a
n
d
K
.
O
ğ
u
z
,
“
S
p
e
e
c
h
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
:
e
m
o
t
i
o
n
a
l
mo
d
e
l
s
,
d
a
t
a
b
a
s
e
s,
f
e
a
t
u
r
e
s
,
p
r
e
p
r
o
c
e
ssi
n
g
me
t
h
o
d
s
,
su
p
p
o
r
t
i
n
g
m
o
d
a
l
i
t
i
e
s,
a
n
d
c
l
a
ssi
f
i
e
r
s,
”
S
p
e
e
c
h
C
o
m
m
u
n
i
c
a
t
i
o
n
,
v
o
l
.
1
1
6
,
p
p
.
5
6
–
7
6
,
J
a
n
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
sp
e
c
o
m.
2
0
1
9
.
1
2
.
0
0
1
.
[
4
5
]
A
.
K
h
a
l
i
l
,
W
.
A
l
-
K
h
a
t
i
b
,
E
.
-
S
.
El
-
A
l
f
y
,
a
n
d
L.
C
h
e
d
e
d
,
“
A
n
g
e
r
d
e
t
e
c
t
i
o
n
i
n
A
r
a
b
i
c
s
p
e
e
c
h
d
i
a
l
o
g
s
,
”
i
n
2
0
1
8
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
i
n
g
S
c
i
e
n
c
e
s
a
n
d
E
n
g
i
n
e
e
r
i
n
g
(
I
C
C
S
E)
,
M
a
r
.
2
0
1
8
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
C
S
E
1
.
2
0
1
8
.
8
3
7
4
2
0
3
.
[
4
6
]
A
.
H
.
M
e
f
t
a
h
,
M
.
Q
a
mh
a
n
,
Y
.
A
l
o
t
a
i
b
i
,
a
n
d
S
.
-
A
.
S
e
l
o
u
a
n
i
,
“
Em
o
t
i
o
n
a
l
sp
e
e
c
h
r
e
c
o
g
n
i
t
i
o
n
u
s
i
n
g
r
h
y
t
h
m
m
e
t
r
i
c
s
a
n
d
a
n
e
w
A
r
a
b
i
c
C
o
r
p
u
s
,
”
i
n
2
0
2
0
1
6
t
h
I
EE
E
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
l
l
o
q
u
i
u
m
o
n
S
i
g
n
a
l
Pro
c
e
ss
i
n
g
&
I
t
s
A
p
p
l
i
c
a
t
i
o
n
s
(
C
S
PA)
,
F
e
b
.
2
0
2
0
,
p
p
.
5
7
–
6
2
,
d
o
i
:
1
0
.
1
1
0
9
/
C
S
P
A
4
8
9
9
2
.
2
0
2
0
.
9
0
6
8
7
1
0
.
[
4
7
]
Y
.
H
i
f
n
y
a
n
d
A
.
A
l
i
,
“
Ef
f
i
c
i
e
n
t
A
r
a
b
i
c
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
u
s
i
n
g
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
s
,
”
i
n
I
C
A
S
S
P
2
0
1
9
-
2
0
1
9
I
EEE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Ac
o
u
st
i
c
s
,
S
p
e
e
c
h
a
n
d
S
i
g
n
a
l
Pr
o
c
e
ss
i
n
g
(
I
C
A
S
S
P)
,
M
a
y
2
0
1
9
,
p
p
.
6
7
1
0
–
6
7
1
4
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
S
S
P
.
2
0
1
9
.
8
6
8
3
6
3
2
.
[
4
8
]
R
.
Y
.
C
h
e
r
i
f
,
A
.
M
o
u
ss
a
o
u
i
,
N
.
F
r
a
h
t
a
,
a
n
d
M
.
B
e
r
r
i
mi
,
“
Ef
f
e
c
t
i
v
e
sp
e
e
c
h
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
u
si
n
g
d
e
e
p
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
e
s
f
o
r
A
l
g
e
r
i
a
n
d
i
a
l
e
c
t
,
”
i
n
2
0
2
1
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
f
W
o
m
e
n
i
n
D
a
t
a
S
c
i
e
n
c
e
a
t
T
a
i
f
U
n
i
v
e
rsi
t
y
(
Wi
D
S
T
a
i
f
)
,
M
a
r
.
2
0
2
1
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
W
i
D
S
Ta
i
f
5
2
2
3
5
.
2
0
2
1
.
9
4
3
0
2
2
4
.
[
4
9
]
R
.
H
.
A
l
j
u
h
a
n
i
,
A
.
A
l
s
h
u
t
a
y
r
i
,
a
n
d
S
.
A
l
a
h
d
a
l
,
“
A
r
a
b
i
c
s
p
e
e
c
h
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
f
r
o
m
S
a
u
d
i
D
i
a
l
e
c
t
C
o
r
p
u
s,”
I
E
EE
Ac
c
e
ss
,
v
o
l
.
9
,
p
p
.
1
2
7
0
8
1
–
1
2
7
0
8
5
,
2
0
2
1
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
1
.
3
1
1
0
9
9
2
.
[
5
0
]
O
.
M
o
h
a
m
e
d
a
n
d
S
.
A
.
A
l
y
,
“
A
r
a
b
i
c
sp
e
e
c
h
e
m
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
e
m
p
l
o
y
i
n
g
w
a
v
2
v
e
c
2
.
0
a
n
d
h
u
b
e
r
t
b
a
s
e
d
o
n
b
a
v
e
d
d
a
t
a
se
t
,
”
a
rXi
v
p
re
p
r
i
n
t
a
rXi
v
:
2
1
1
0
.
0
4
4
2
5
,
2
0
2
1
,
d
o
i
:
1
0
.
4
8
5
5
0
/
a
r
X
i
v
.
2
1
1
0
.
0
4
4
2
5
.
[
5
1
]
M
.
T
a
j
a
l
s
i
r
,
S
.
M
.
H
e
r
n
´
a
n
d
e
z
,
a
n
d
F
.
A
.
M
o
h
a
mm
e
d
,
“
A
S
ER
S
-
LS
TM
:
A
r
a
b
i
c
sp
e
e
c
h
e
mo
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
s
y
st
e
m
b
a
se
d
on
LSTM
m
o
d
e
l
,
”
S
i
g
n
a
l
&
I
m
a
g
e
Pro
c
e
ssi
n
g
:
An
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
,
v
o
l
.
1
3
,
n
o
.
1
,
p
p
.
1
9
–
2
7
,
F
e
b
.
2
0
2
2
,
d
o
i
:
1
0
.
5
1
2
1
/
si
p
i
j
.
2
0
2
2
.
1
3
1
0
2
.
[
5
2
]
H
.
A
l
a
mr
i
a
n
d
H
.
A
l
sh
a
n
b
a
r
i
,
“
Em
o
t
i
o
n
r
e
c
o
g
n
i
t
i
o
n
i
n
A
r
a
b
i
c
sp
e
e
c
h
f
r
o
m
S
a
u
d
i
D
i
a
l
e
c
t
C
o
r
p
u
s
u
si
n
g
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
a
n
d
d
e
e
p
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
ms
.
”
J
u
n
.
1
3
,
2
0
2
3
,
d
o
i
:
1
0
.
2
1
2
0
3
/
r
s
.
3
.
r
s
-
3
0
1
9
1
5
9
/
v
1
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Abd
e
lk
a
d
e
r
Be
n
z
ira
r
(Ph
.
D.
st
u
d
e
n
t)
re
c
e
iv
e
d
h
is
m
a
ste
r'
s
d
e
g
re
e
in
in
telli
g
e
n
t
sy
ste
m
s
a
n
d
n
e
two
r
k
s
a
t
th
e
F
a
c
u
lt
y
o
f
S
c
ien
c
e
a
n
d
Tec
h
n
o
l
o
g
y
,
S
i
d
i
M
o
h
a
m
e
d
Be
n
Ab
d
e
ll
a
h
Un
iv
e
rsity
o
f
F
e
z
(M
o
r
o
c
c
o
)
in
2
0
1
5
.
He
is
c
u
rre
n
tl
y
p
u
rsu
i
n
g
h
is
P
h
.
D
.
i
n
sp
e
e
c
h
e
m
o
ti
o
n
re
c
o
g
n
it
i
o
n
.
His
re
se
a
rc
h
in
tere
st
in
c
lu
d
e
s
m
a
c
h
in
e
lea
rn
in
g
a
lg
o
r
it
h
m
s
a
n
d
s
p
e
e
c
h
e
m
o
ti
o
n
re
c
o
g
n
it
i
o
n
m
e
th
o
d
s.
He
c
a
n
b
e
c
o
n
tac
ted
at
e
m
a
il
:
a
b
d
e
lk
a
d
e
r.
b
e
n
z
irar.d
2
3
@
u
m
p
.
a
c
.
m
a
.
Pro
f.
Dr
.
M
o
h
a
m
e
d
H
a
m
i
d
i
is
a
P
ro
fe
ss
o
r
in
t
h
e
De
p
a
rt
m
e
n
t
o
f
Co
m
p
u
ter
S
c
ien
c
e
a
t
th
e
M
u
lt
id
isc
ip
li
n
a
r
y
F
a
c
u
lt
y
o
f
Na
d
o
r,
M
o
h
a
m
m
e
d
I
Un
iv
e
rsity
,
Ou
jd
a
,
M
o
ro
c
c
o
.
He
re
c
e
iv
e
d
h
is
P
h
.
D.
in
C
o
m
p
u
ter
S
c
ien
c
e
in
2
0
2
0
fro
m
t
h
e
F
a
c
u
lt
y
o
f
S
c
ien
c
e
s,
Dh
a
r
El
M
a
h
ra
z
,
S
id
i
M
o
h
a
m
m
e
d
Be
n
A
b
d
e
ll
a
h
U
n
iv
e
rsit
y
,
F
e
z
.
He
o
b
ta
in
e
d
h
is
Ba
c
h
e
lo
r'
s
d
e
g
re
e
fro
m
th
e
M
u
l
ti
d
isc
i
p
li
n
a
r
y
F
a
c
u
l
t
y
o
f
Na
d
o
r
(F
P
N),
M
o
h
a
m
m
e
d
I
Un
iv
e
rsity
,
i
n
2
0
1
1
,
a
n
d
h
is
M
a
ste
r'
s
d
e
g
re
e
fro
m
th
e
F
a
c
u
lt
y
o
f
S
c
ien
c
e
s,
Teto
u
a
n
,
Ab
d
e
l
m
a
lek
Essa
a
d
i
Un
iv
e
rsity
,
M
o
ro
c
c
o
,
i
n
2
0
1
3
.
His
c
u
rre
n
t
re
se
a
rc
h
i
n
tere
sts
i
n
c
lu
d
e
m
a
c
h
in
e
lea
rn
i
n
g
,
fo
rm
a
n
t
fre
q
u
e
n
c
ies
,
p
a
tt
e
rn
re
c
o
g
n
it
i
o
n
,
sp
e
e
c
h
a
n
d
lan
g
u
a
g
e
p
r
o
c
e
ss
in
g
,
sp
e
e
c
h
re
c
o
g
n
it
io
n
,
sp
e
e
c
h
se
c
u
rit
y
,
i
n
tera
c
ti
v
e
v
o
ice
sy
ste
m
s,
a
n
d
v
o
ice
o
v
e
r
I
P.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
m
.
h
a
m
id
i@u
m
p
.
a
c
.
m
a
.
Pro
f.
Dr
.
M
o
u
n
c
e
f
Fi
l
a
li
B
o
u
a
m
i
re
c
e
iv
e
d
a
n
M
.
S
c
.
in
E
lec
tro
n
ics
fro
m
t
h
e
Un
iv
e
rsity
o
f
F
e
z
,
M
o
ro
c
c
o
in
1
9
9
8
a
n
d
a
P
h
.
D.
d
e
g
re
e
fro
m
t
h
e
U
n
iv
e
rsit
y
o
f
G
ra
n
a
d
a
,
S
p
a
i
n
in
2
0
0
5
a
fter
h
a
v
i
n
g
d
e
fe
n
d
e
d
a
d
o
c
to
ra
l
t
h
e
sis
o
n
t
h
e
m
o
d
e
li
n
g
o
f
RBF
n
e
u
ra
l
n
e
two
rk
s u
sin
g
T
-
No
rm
a
n
d
T
-
C
o
n
o
rm
o
p
e
ra
to
r
s
a
n
d
we
ig
h
ts
p
a
ra
m
e
teriz
a
ti
o
n
.
S
in
c
e
2
0
1
0
h
e
h
a
s
b
e
e
n
a
S
e
n
io
r
Lec
t
u
re
r
a
t
th
e
P
o
ly
-
Disc
ip
li
n
a
r
y
F
a
c
u
lt
y
o
f
Na
d
o
r,
M
o
h
a
m
m
e
d
p
re
m
ier
Un
iv
e
rsity
,
M
o
ro
c
c
o
.
His
re
se
a
rc
h
in
tere
st
in
c
lu
d
e
s
m
a
c
h
in
e
lea
rn
in
g
a
lg
o
ri
t
h
m
s,
tex
t
c
las
sifica
ti
o
n
a
n
d
sp
e
e
c
h
re
c
o
g
n
it
io
n
m
e
th
o
d
s.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
m
.
f
il
a
li
b
o
u
a
m
i@u
m
p
.
a
c
.
m
a
.
Evaluation Warning : The document was created with Spire.PDF for Python.