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.
14
,
No
.
3
,
J
u
n
e
2
0
2
5
,
p
p
.
1683
~
1
6
9
5
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
14
.i
3
.
p
p
1
6
8
3
-
1
6
9
5
1683
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Techniqu
es o
f
Q
u
ra
n reciters
re
co
g
nition: a r
ev
iew
I
bra
him
Ahm
ed
Al
-
O
m
a
ri
,
Asm
a
Al
-
Sh
a
rg
a
bi
,
M
o
ha
mm
ed
H
a
dwa
n
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
m
a
t
i
o
n
Te
c
h
n
o
l
o
g
y
,
C
o
l
l
e
g
e
o
f
C
o
m
p
u
t
e
r
,
Q
a
ss
i
m U
n
i
v
e
r
si
t
y
,
B
u
r
a
y
d
a
h
,
S
a
u
d
i
A
r
a
b
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Sep
1
,
2
0
2
4
R
ev
is
ed
No
v
22
,
2
0
2
4
Acc
ep
ted
J
an
27
,
2
0
2
5
Th
e
Qu
ra
n
is
t
h
e
h
o
l
y
b
o
o
k
o
f
t
h
e
Isla
m
.
Re
a
d
in
g
a
n
d
li
ste
n
in
g
to
th
e
Qu
ra
n
is
a
n
imp
o
rtan
t
p
a
rt
o
f
t
h
e
d
a
il
y
l
ife
o
f
M
u
slim
s
.
M
u
slim
s
a
re
k
e
e
n
to
li
ste
n
to
re
c
it
a
ti
o
n
s
o
f
Q
u
ra
n
b
y
s
k
il
le
d
re
c
it
e
rs
to
lea
rn
t
h
e
c
o
rre
c
t
re
c
it
a
ti
o
n
fo
r
th
e
p
u
rp
o
se
o
f
u
n
d
e
rsta
n
d
in
g
a
n
d
c
o
n
tem
p
latin
g
.
Th
e
re
fo
re
,
th
e
r
e
a
re
larg
e
v
a
riety
o
f
a
u
d
io
re
c
it
a
ti
o
n
s
fo
r
m
a
n
y
sk
il
le
d
re
c
it
e
rs.
Wi
th
t
h
e
a
v
a
il
a
b
il
it
y
o
f
th
is
h
u
g
e
a
m
o
u
n
t
o
f
re
c
it
a
ti
o
n
s
a
n
d
a
lso
wit
h
th
e
g
re
a
t
p
ro
g
re
ss
in
v
o
ice
re
c
o
g
n
it
i
o
n
tec
h
n
o
l
o
g
ies
,
m
a
n
y
re
se
a
rc
h
e
ffo
rts
h
a
v
e
b
e
e
n
d
e
v
o
ted
to
c
o
n
tri
b
u
te
m
a
k
in
g
re
c
it
a
ti
o
n
b
e
tt
e
r
u
sin
g
a
rti
ficia
l
i
n
telli
g
e
n
c
e
.
On
e
u
se
fu
l
a
p
p
li
c
a
ti
o
n
in
t
h
is
a
re
a
is
id
e
n
ti
fy
in
g
th
e
re
c
it
e
rs
o
f
th
e
Qu
ra
n
.
Th
e
re
a
re
v
a
rio
u
s
so
l
u
ti
o
n
s
i
n
tr
o
d
u
c
e
d
b
y
re
se
a
rc
h
e
rs
;
h
o
we
v
e
r,
t
h
e
se
so
lu
ti
o
n
s
v
a
ry
sig
n
ifi
c
a
n
t
ly
i
n
term
s
o
f
a
c
c
u
ra
c
y
,
a
n
d
e
fficie
n
c
y
.
T
h
is
re
se
a
rc
h
se
e
k
s
to
p
ro
v
id
e
a
re
v
iew
o
f
th
e
se
so
l
u
ti
o
n
s.
It
a
lso
re
v
iew
s
a
v
a
il
a
b
le
d
a
ta
se
ts
u
sin
g
d
iffere
n
t
c
rit
e
ria.
F
in
a
ll
y
,
so
m
e
o
p
e
n
issu
e
s a
n
d
c
h
a
ll
e
n
g
e
s we
re
a
d
d
re
ss
e
d
.
K
ey
w
o
r
d
s
:
Ar
ab
ic
lan
g
u
a
g
e
p
r
o
ce
s
s
in
g
Dee
p
lear
n
in
g
Ma
ch
in
e
lear
n
in
g
Natu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
Qu
r
an
r
ec
iter
s
r
ec
o
g
n
itio
n
Sp
ea
k
er
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
:
Asma
Al
-
Sh
ar
g
ab
i
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atio
n
T
e
ch
n
o
lo
g
y
,
C
o
lleg
e
o
f
C
o
m
p
u
t
er
,
Qass
im
Un
iv
er
s
ity
B
u
r
ay
d
ah
5
1
4
5
2
,
Sau
d
i A
r
ab
i
a
E
m
ail:
as.a
ls
h
ar
g
ab
i@
q
u
.
ed
u
.
s
a
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
Qu
r
an
,
th
e
h
o
ly
b
o
o
k
o
f
I
s
lam
,
h
o
ld
s
u
n
p
ar
alleled
s
ig
n
if
ican
ce
in
b
o
th
lin
g
u
is
tic
an
d
r
elig
io
u
s
co
n
tex
ts
.
R
en
o
wn
e
d
f
o
r
its
el
o
q
u
en
ce
,
clar
ity
,
an
d
r
h
eto
r
ic
al
m
aster
y
,
th
e
Qu
r
a
n
is
r
e
v
e
r
ed
as
th
e
h
ig
h
est
Ar
ab
ic
liter
ar
y
wo
r
k
.
I
t
is
m
eticu
lo
u
s
ly
d
iv
id
ed
in
to
3
0
p
ar
t
s
,
ea
ch
n
am
ed
af
ter
th
e
ch
ap
t
er
(
s
u
r
ah
)
it
b
eg
in
s
with
,
en
co
m
p
ass
in
g
a
to
tal
o
f
1
1
4
c
h
ap
ter
s
.
T
h
ese
c
h
ap
ter
s
co
llectiv
ely
co
n
tain
6
,
2
3
6
v
er
s
es
an
d
3
2
3
,
6
7
0
letter
s
.
No
tab
ly
,
th
e
lo
n
g
est
c
h
ap
ter
is
Al
-
B
aq
ar
ah
,
with
2
8
6
v
er
s
es,
wh
ile
th
e
s
h
o
r
test
is
Al
-
Kaw
th
ar
,
with
ju
s
t
3
v
er
s
es
[
1
]
.
T
h
e
Qu
r
an
is
r
ec
ited
u
s
in
g
a
s
p
ec
ialized
m
eth
o
d
k
n
o
wn
as
“
T
ajwe
ed
,
”
wh
ich
en
co
m
p
ass
es
a
s
et
o
f
r
u
les
f
o
r
co
r
r
ec
t
p
r
o
n
u
n
ciatio
n
d
u
r
in
g
r
ec
itatio
n
[
2
]
.
T
ajwe
ed
en
s
u
r
es
th
at
ea
ch
wo
r
d
is
ar
ticu
lated
ac
cu
r
ately
,
as m
is
p
r
o
n
u
n
ciatio
n
ca
n
s
ig
n
i
f
ican
tly
alter
th
e
m
ea
n
in
g
o
f
t
h
e
v
er
s
es
[
3
]
.
Sp
ea
k
er
r
ec
o
g
n
itio
n
tech
n
o
lo
g
y
h
as
r
ev
o
lu
tio
n
ized
th
e
wa
y
Qu
r
an
ic
r
ec
itatio
n
s
ar
e
ca
talo
g
ed
an
d
ac
ce
s
s
ed
.
B
y
lev
er
ag
in
g
ad
v
an
ce
d
alg
o
r
ith
m
s
to
an
aly
ze
th
e
u
n
iq
u
e
v
o
ca
l
ch
ar
ac
ter
i
s
tics
o
f
in
d
iv
id
u
al
r
ec
iter
s
,
th
is
tech
n
o
lo
g
y
is
in
s
tr
u
m
en
tal
in
ca
teg
o
r
izin
g
an
d
a
r
r
an
g
in
g
Qu
r
an
ic
r
ec
itatio
n
s
o
r
en
ab
lin
g
u
s
er
s
to
s
ea
r
ch
f
o
r
r
ec
itatio
n
s
b
ased
o
n
th
eir
p
r
ef
er
r
ed
r
ec
iter
s
[
4
]
.
E
d
u
ca
tio
n
al
in
s
titu
tio
n
s
an
d
ac
ad
em
ics
d
er
iv
e
v
alu
ab
le
b
en
ef
its
f
r
o
m
th
is
tec
h
n
o
lo
g
y
,
as
it
f
ac
ilit
ates
th
e
s
tu
d
y
an
d
teac
h
in
g
o
f
T
ajwe
ed
b
y
o
f
f
er
i
n
g
p
r
ec
is
e
id
en
tific
atio
n
o
f
r
ec
iter
s
an
d
cr
ed
ib
le
ex
am
p
les
f
o
r
ed
u
c
atio
n
al
u
s
e.
T
h
e
p
r
ec
is
e
id
en
tific
atio
n
h
elp
s
in
en
s
u
r
in
g
th
at
s
tu
d
en
ts
lear
n
t
h
e
co
r
r
ec
t
p
r
o
n
u
n
ciatio
n
an
d
in
to
n
atio
n
,
p
r
eser
v
in
g
th
e
in
teg
r
ity
o
f
Qu
r
a
n
ic
r
ec
itatio
n
.
Fu
r
th
er
m
o
r
e
,
th
e
p
r
eser
v
atio
n
o
f
d
iv
er
s
e
r
ec
itatio
n
s
ty
les
i
s
s
u
p
p
o
r
ted
,
en
s
u
r
in
g
th
at
th
e
r
ich
o
r
al
tr
ad
itio
n
o
f
Qu
r
a
n
ic
r
ec
itatio
n
is
m
ain
tain
ed
f
o
r
f
u
t
u
r
e
g
e
n
er
atio
n
s
.
B
y
ca
talo
g
in
g
v
ar
i
o
u
s
r
ec
itatio
n
s
ty
les,
f
r
o
m
d
if
f
er
e
n
t
r
eg
i
o
n
s
an
d
d
i
alec
ts
,
th
e
tech
n
o
lo
g
y
h
elp
s
i
n
p
r
eser
v
i
n
g
th
e
cu
ltu
r
al
an
d
lin
g
u
is
tic
d
iv
er
s
ity
with
in
th
e
Mu
s
lim
c
o
m
m
u
n
ity
.
T
h
is
p
r
eser
v
atio
n
is
v
ital
f
o
r
m
ain
tain
in
g
th
e
au
th
e
n
ticity
an
d
r
ic
h
n
ess
o
f
t
h
e
Qu
r
an
ic
r
ec
itatio
n
tr
a
d
itio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
14
,
No
.
3
,
J
u
n
e
20
25
:
1
6
8
3
-
1
6
9
5
1684
I
n
ad
d
itio
n
,
au
to
m
atic
s
p
ee
c
h
r
ec
o
g
n
itio
n
(
ASR
)
s
y
s
tem
s
co
n
tr
ib
u
te
to
th
e
ef
f
icien
t
tr
an
s
cr
ip
tio
n
an
d
in
d
ex
in
g
o
f
Qu
r
an
ic
v
e
r
s
es
[
5
]
,
s
tr
ea
m
lin
in
g
th
e
s
ea
r
ch
f
o
r
s
p
ec
if
ic
v
er
s
es
o
r
ch
ap
ter
s
.
T
h
ese
tech
n
o
lo
g
ical
ad
v
an
ce
m
e
n
ts
h
av
e
m
ad
e
Qu
r
an
ic
r
ec
itatio
n
s
m
o
r
e
ac
ce
s
s
ib
le
to
a
wid
er
au
d
ien
ce
,
i
n
clu
d
i
n
g
in
d
iv
i
d
u
als
with
d
is
ab
ilit
ies
o
r
th
o
s
e
wh
o
p
r
ef
er
d
ig
ital
f
o
r
m
ats.
ASR
tech
n
o
lo
g
y
e
n
s
u
r
es
th
at
v
is
u
ally
i
m
p
air
ed
in
d
iv
id
u
als
ca
n
ac
ce
s
s
ac
cu
r
ate
au
d
io
r
ec
itatio
n
s
,
wh
ile
d
ig
ital
f
o
r
m
ats
allo
w
f
o
r
co
n
v
en
ie
n
t
ac
ce
s
s
o
n
v
ar
i
o
u
s
d
ev
ices.
T
h
is
en
h
an
ce
d
ac
ce
s
s
ib
ilit
y
n
o
t
o
n
l
y
im
p
r
o
v
es
th
e
o
v
e
r
all
u
s
er
ex
p
er
ie
n
ce
b
u
t
also
f
o
s
ter
s
a
d
ee
p
e
r
co
n
n
ec
tio
n
with
th
e
s
ac
r
ed
tex
t,
allo
win
g
m
o
r
e
in
d
iv
id
u
als to
en
g
a
g
e
with
th
e
Qu
r
a
n
in
a
m
ea
n
in
g
f
u
l w
ay
.
T
h
e
tech
n
i
q
u
es
u
s
ed
to
r
ec
o
g
n
ize
Qu
r
an
r
ec
iter
s
en
c
o
m
p
as
s
a
v
ar
iety
o
f
m
eth
o
d
s
a
n
d
alg
o
r
ith
m
s
,
p
ar
ticu
lar
ly
i
n
th
e
f
ea
tu
r
e
e
x
tr
ac
tio
n
an
d
class
if
icatio
n
s
tag
e
s
[
6
]
.
T
ec
h
n
i
q
u
es
s
u
ch
as
Mel
-
f
r
eq
u
e
n
cy
ce
p
s
tr
al
co
ef
f
icien
ts
(
MFC
C
s
)
an
d
d
e
ep
lear
n
in
g
m
o
d
els
lik
e
co
n
v
o
lu
tio
n
al
n
e
u
r
al
n
etw
o
r
k
s
(
C
NNs)
an
d
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
s
(
R
NNs)
ar
e
co
m
m
o
n
ly
em
p
lo
y
e
d
to
ac
h
iev
e
h
ig
h
ac
cu
r
ac
y
i
n
r
ec
iter
r
ec
o
g
n
itio
n
.
R
ec
en
t
ad
v
an
ce
m
en
ts
in
ar
tifi
cial
in
tellig
en
ce
h
av
e
f
u
r
th
er
r
ef
in
ed
th
ese
s
y
s
tem
s
,
m
ak
in
g
th
em
m
o
r
e
r
o
b
u
s
t
an
d
r
eliab
le.
Fo
r
in
s
tan
ce
,
T
an
tawi
et
a
l
.
[
7
]
d
ev
elo
p
ed
a
lar
g
e
-
v
o
c
ab
u
la
r
y
s
p
ea
k
er
-
i
n
d
ep
e
n
d
en
t
ASR
s
y
s
tem
f
o
r
Qu
r
an
ic
r
ec
itatio
n
s
,
ac
h
iev
in
g
p
r
o
m
is
in
g
r
esu
lts
with
a
wo
r
d
er
r
o
r
r
ate
(
W
E
R
)
r
an
g
in
g
f
r
o
m
0
.
2
7
%
to
6
.
3
1
%.
Ad
d
itio
n
ally
,
Gh
o
r
i
et
a
l
.
[
8
]
ex
p
lo
r
ed
ac
o
u
s
tic
m
o
d
elin
g
u
s
in
g
d
ee
p
lear
n
in
g
f
o
r
Qu
r
an
r
e
citatio
n
ass
is
tan
ce
,
d
em
o
n
s
tr
atin
g
th
e
p
o
ten
tial
o
f
th
ese
tech
n
iq
u
es
in
im
p
r
o
v
in
g
r
ec
o
g
n
itio
n
ac
cu
r
ac
y
.
T
h
ese
d
ev
elo
p
m
en
ts
n
o
t
o
n
ly
f
ac
ilit
ate
m
o
r
e
ef
f
icien
t
ca
teg
o
r
izatio
n
b
u
t
also
p
r
o
v
id
e
u
s
er
s
with
a
p
er
s
o
n
al
ized
an
d
e
n
r
ich
ed
ex
p
er
ien
ce
wh
en
en
g
ag
in
g
with
th
e
Qu
r
an
.
Mo
r
e
o
v
er
,
t
h
e
in
teg
r
atio
n
o
f
s
p
ea
k
er
r
e
co
g
n
itio
n
an
d
ASR
tech
n
o
lo
g
ies
in
Qu
r
an
ic
r
ec
itatio
n
h
as
h
a
d
a
p
r
o
f
o
u
n
d
im
p
ac
t
o
n
ac
ce
s
s
ib
ilit
y
,
ed
u
ca
tio
n
,
an
d
p
r
eser
v
atio
n
[
9
]
.
B
y
h
ar
n
ess
in
g
th
e
p
o
wer
o
f
ad
v
an
ce
d
alg
o
r
ith
m
s
an
d
m
ac
h
in
e
lear
n
in
g
,
th
ese
tech
n
o
lo
g
ies
en
s
u
r
e
th
at
th
e
Qu
r
an
ic
r
ec
itatio
n
tr
ad
itio
n
r
em
ain
s
v
ib
r
an
t
an
d
ac
ce
s
s
ib
le
to
all,
wh
ile
also
p
u
s
h
in
g
th
e
b
o
u
n
d
ar
ies
o
f
wh
at
is
p
o
s
s
ib
le
in
th
e
f
ield
o
f
s
p
ee
ch
r
ec
o
g
n
itio
n
.
T
h
is
p
ap
er
r
ev
iews
th
e
m
eth
o
d
s
u
s
ed
f
o
r
Qu
r
an
r
ec
iter
r
ec
o
g
n
itio
n
an
d
th
e
d
atasets
av
ailab
le
in
th
is
f
ield
.
Ad
d
itio
n
ally
,
it
s
u
m
m
ar
i
ze
s
k
ey
in
s
ig
h
ts
an
d
s
tr
ateg
ies
th
at
ca
n
en
h
a
n
ce
th
e
ac
c
u
r
ac
y
an
d
ef
f
ec
tiv
en
ess
o
f
th
ese
r
ec
o
g
n
itio
n
s
y
s
tem
s
.
T
h
e
s
u
b
s
eq
u
en
t
s
ec
tio
n
s
o
f
th
is
p
ap
er
ar
e
m
eticu
lo
u
s
ly
d
elin
ea
ted
as
f
o
llo
ws:
s
ec
tio
n
2
d
e
v
o
ted
to
in
t
r
o
d
u
c
e
a
g
e
n
er
al
v
iew
o
f
s
p
ea
k
er
r
ec
o
g
n
itio
n
s
y
s
tem
s
an
d
h
o
w
i
t
wo
r
k
s
.
Sectio
n
3
o
u
tlin
es
th
e
r
esear
ch
m
eth
o
d
o
lo
g
y
em
p
l
o
y
ed
in
th
is
r
ev
i
ew,
d
etailin
g
th
e
ap
p
r
o
ac
h
t
ak
en
.
I
n
s
ec
tio
n
4
,
r
ec
iter
s
r
ec
o
g
n
itio
n
tech
n
iq
u
e
s
ar
e
ex
am
in
ed
,
p
r
esen
tin
g
a
r
ev
iew
o
f
k
ey
s
o
lu
tio
n
s
p
r
o
p
o
s
ed
b
y
th
e
r
esear
c
h
co
m
m
u
n
ity
f
o
r
r
ec
o
g
n
izin
g
Qu
r
an
r
ec
iter
s
.
Sectio
n
5
o
f
f
er
s
a
b
r
ief
s
u
m
m
ar
y
o
f
av
ailab
le
Qu
r
an
ic
r
ec
itatio
n
s
d
atasets
.
Sectio
n
6
p
r
esen
ts
a
d
is
cu
s
s
io
n
an
d
o
p
en
is
s
u
es
,
an
aly
zin
g
th
e
liter
atu
r
e
an
d
ad
d
r
ess
in
g
cu
r
r
en
t
ch
allen
g
es
an
d
u
n
r
eso
lv
ed
q
u
esti
o
n
s
.
Fin
ally
,
s
ec
tio
n
7
co
n
clu
d
es
with
a
s
u
m
m
ar
y
o
f
th
e
wo
r
k
in
tr
o
d
u
ce
d
,
h
ig
h
lig
h
tin
g
k
ey
co
n
clu
s
io
n
s
a
n
d
s
u
g
g
esti
n
g
d
ir
ec
tio
n
s
f
o
r
f
u
r
th
er
r
esear
ch
.
2.
O
VE
RVI
E
W
O
F
SP
E
AK
E
R
RE
CO
G
NI
T
I
O
N
SY
ST
E
M
S
Sp
ea
k
er
r
ec
o
g
n
itio
n
is
o
n
e
o
f
th
e
m
o
s
t
im
p
o
r
tan
t
u
s
es
o
f
v
o
ice
r
ec
o
g
n
itio
n
s
y
s
tem
s
.
Sp
ea
k
er
r
ec
o
g
n
itio
n
aim
s
at
d
eter
m
i
n
in
g
wh
o
is
s
p
ea
k
in
g
i
n
a
g
i
v
e
n
s
p
ee
ch
b
y
m
atch
i
n
g
th
e
v
o
i
ce
b
io
m
etr
ics
o
f
th
e
s
p
ea
k
er
’
s
v
o
ice,
th
e
v
o
ice
b
io
m
etr
ics
p
atter
n
is
lear
n
ed
b
y
t
r
ain
in
g
[
2
]
.
As
a
r
esu
lt
o
f
v
a
r
i
atio
n
s
in
v
o
ca
l
tr
ac
t
f
o
r
m
a
n
d
lar
y
n
x
v
o
l
u
m
e,
ea
c
h
p
er
s
o
n
’
s
v
o
ice
is
d
is
tin
ctiv
e.
E
v
er
y
s
p
ea
k
er
h
as
a
u
n
i
q
u
e
f
e
atu
r
e
o
f
s
p
ea
k
in
g
,
wh
ich
m
ay
in
cl
u
d
e
r
h
y
th
m
,
t
o
n
e,
a
p
atter
n
o
f
p
r
o
n
u
n
ciatio
n
,
o
r
o
th
e
r
s
[
1
0
]
.
Sp
ea
k
e
r
r
ec
o
g
n
itio
n
m
eth
o
d
s
h
av
e
u
tili
ze
d
th
ese
u
n
iq
u
e
f
ea
tu
r
es
to
p
r
o
d
u
ce
ac
cu
r
ate
s
o
lu
tio
n
s
.
As
s
h
o
wn
in
Fig
u
r
e
1
,
s
p
ea
k
er
r
ec
o
g
n
itio
n
s
y
s
tem
s
ty
p
ically
in
clu
d
e
m
an
y
p
h
ases
;
th
e
m
ain
o
n
es
ar
e
p
r
e
-
p
r
o
ce
s
s
in
g
,
f
ea
tu
r
e
e
x
tr
ac
tio
n
,
m
o
d
el
tr
ain
in
g
,
an
d
r
ec
o
g
n
itio
n
[
1
1
]
.
Ma
n
y
r
esear
ch
er
s
h
av
e
u
s
ed
v
ar
i
o
u
s
ex
tr
ac
tio
n
m
et
h
o
d
o
lo
g
ie
s
an
d
class
if
icatio
n
tech
n
iq
u
es
to
r
ec
o
g
n
ize
th
e
s
p
ea
k
er
s
with
r
em
ar
k
ab
le
r
esu
lts
,
f
o
r
in
s
tan
ce
,
th
e
h
id
d
e
n
Ma
r
k
o
v
m
o
d
el
(H
M
M)
[
1
2
]
–
[
1
5
]
,
Gau
s
s
ian
m
ix
tu
r
e
m
o
d
el
(
GM
M)
[
1
6
]
,
[
1
7
]
,
a
r
tific
ial
n
eu
r
al
n
etwo
r
k
(
A
NN)
[
1
8
]
,
K
-
n
ea
r
est
n
eig
h
b
o
r
(
KNN)
an
d
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
es
(
SVM)
[
1
9
]
.
Fu
r
th
er
m
o
r
e
,
f
ea
tu
r
e
e
x
tr
ac
tio
n
is
ess
en
tial
s
in
ce
it
ca
n
s
ig
n
if
ican
tly
b
o
o
s
t
th
e
ef
f
icien
cy
o
f
th
e
m
o
d
el
b
y
m
ak
in
g
th
e
d
ata
m
o
r
e
u
s
ef
u
l
an
d
lo
wer
in
g
its
d
im
en
s
io
n
ality
.
MFC
C
[
2
0
]
an
d
lin
ea
r
p
r
ed
ictiv
e
co
d
in
g
(
L
PC
)
[
1
9
]
ar
e
th
e
m
o
s
t
p
o
p
u
lar
ac
o
u
s
tic
f
ea
tu
r
e
ex
tr
ac
tio
n
m
eth
o
d
s
.
Desp
ite
r
ec
en
t
im
p
r
o
v
em
en
ts
in
s
p
ea
k
er
r
ec
o
g
n
itio
n
,
th
er
e
ar
e
s
till
m
an
y
ch
allen
g
es
s
u
ch
as
v
ar
iab
ilit
y
an
d
a
lack
o
f
d
ata
[
2
1
]
,
s
ec
tio
n
6
in
tr
o
d
u
c
es
a
d
etailed
d
is
cu
s
s
io
n
an
d
h
ig
h
lig
h
ts
th
e
m
ai
n
ch
allen
g
es a
n
d
o
p
en
is
s
u
es.
T
h
e
f
o
llo
win
g
s
ec
tio
n
s
d
is
cu
s
s
t
h
e
m
ain
p
h
ases
.
2
.
1
.
P
re
-
pro
ce
s
s
ing
Pre
-
p
r
o
ce
s
s
in
g
is
an
ess
en
tial
s
tag
e
in
p
r
o
d
u
cin
g
an
ac
cu
r
ate
s
p
ea
k
er
r
ec
o
g
n
itio
n
s
y
s
te
m
d
u
e
to
a
n
u
m
b
er
o
f
f
ac
to
r
s
.
No
is
e
is
o
n
e
o
f
th
e
m
o
s
t
im
p
o
r
ta
n
t
f
ac
t
o
r
s
d
u
e
to
th
e
p
o
o
r
q
u
ality
o
f
s
p
ee
ch
an
d
au
d
io
r
ec
o
r
d
in
g
s
y
s
tem
s
an
d
d
ev
ice
s
[
2
2
]
.
I
n
ad
d
itio
n
,
th
e
s
p
ee
c
h
s
ig
n
als
n
o
r
m
ally
co
n
tain
n
u
m
er
o
u
s
p
er
i
o
d
s
o
f
s
ilen
ce
.
T
h
e
s
ilen
ce
s
ig
n
al
is
m
ea
n
in
g
less
s
in
ce
it
co
n
tain
s
n
o
in
f
o
r
m
atio
n
.
Z
er
o
cr
o
s
s
in
g
r
ate
(
Z
C
R
)
an
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Tech
n
iq
u
es o
f Q
u
r
a
n
r
ec
iter
s
r
ec
o
g
n
itio
n
:
a
r
ev
iew
(
I
b
r
a
h
i
m
A
h
med
A
l
-
Oma
r
i
)
1685
s
h
o
r
t
-
tim
e
en
er
g
y
(
STE
)
ar
e
two
way
s
to
elim
in
ate
th
ese
p
er
io
d
s
[
2
3
]
.
Fu
r
th
er
m
o
r
e,
th
e
p
r
e
-
em
p
h
asis
o
f
th
e
s
p
ee
ch
s
ig
n
al
is
ess
en
tial f
o
r
h
ig
h
-
f
r
e
q
u
en
c
y
p
r
e
p
r
o
ce
s
s
in
g
.
I
t
is
o
f
ten
u
s
ed
t
o
en
h
an
ce
h
ig
h
-
f
r
eq
u
en
cy
p
ar
ts
o
f
t
h
e
s
ig
n
a
l.
T
h
is
en
tails
ap
p
l
y
in
g
a
h
i
g
h
-
p
ass
f
ilter
(
FIR)
to
th
e
s
ig
n
al
to
e
n
h
an
ce
th
e
h
ig
h
e
r
f
r
eq
u
e
n
cies
an
d
r
e
d
u
ce
th
e
lo
wer
f
r
eq
u
e
n
cies
[
2
4
]
.
A
d
d
itio
n
ally
,
i
n
s
ig
n
al
p
r
o
ce
s
s
in
g
,
th
e
n
o
r
m
a
lizatio
n
tech
n
iq
u
e
is
u
s
ed
to
en
s
u
r
e
th
at
th
e
s
ig
n
al
h
as
an
ac
co
r
d
a
n
t
lev
el.
No
r
m
aliza
tio
n
d
ec
r
ea
s
es
th
e
im
p
ac
t
o
f
s
ig
n
al
g
ap
s
as
well.
W
in
d
o
win
g
ar
e
also
s
ig
n
al
-
p
r
o
ce
s
s
in
g
tech
n
iq
u
es
th
at
ar
e
em
p
lo
y
e
d
to
m
in
im
iz
e
th
e
im
p
ac
t
o
f
s
ig
n
al
g
a
p
s
[
2
5
]
.
I
t
in
clu
d
es
n
o
r
m
alizin
g
th
e
s
ig
n
al
to
a
s
p
ec
if
ic
lev
el,
s
p
litt
in
g
it
i
n
to
in
te
r
f
er
in
g
s
eg
m
en
ts
,
an
d
th
en
p
er
f
o
r
m
in
g
a
win
d
o
w
f
u
n
ctio
n
o
n
ea
ch
s
eg
m
en
t
to
d
ec
r
ea
s
e
g
ap
s
at
th
e
ed
g
es.
Fig
u
r
e
1
.
Sp
ea
k
er
r
ec
o
g
n
itio
n
s
y
s
tem
f
r
am
ewo
r
k
[
2
6
]
2
.
2
.
F
e
a
t
ure
ex
t
r
a
ct
io
n
I
n
s
p
ea
k
e
r
r
e
c
o
g
n
i
ti
o
n
,
f
e
at
u
r
e
ex
tr
ac
t
i
o
n
is
t
h
e
te
c
h
n
iq
u
e
o
f
ex
t
r
ac
ti
n
g
s
i
g
n
i
f
i
ca
n
t
f
e
at
u
r
es
f
r
o
m
a
u
d
i
o
s
ig
n
als
t
h
at
m
ig
h
t
b
e
u
t
ili
ze
d
to
r
e
c
o
g
n
i
ze
a
s
p
ea
k
e
r
.
T
o
b
e
tte
r
u
n
d
er
s
ta
n
d
a
s
p
ee
ch
s
i
g
n
a
l,
i
t
c
a
n
b
e
b
r
o
k
e
n
d
o
wn
i
n
t
o
i
ts
co
n
s
t
it
u
e
n
t p
ar
ts
,
wh
ic
h
is
w
h
at
f
ea
tu
r
e
e
x
t
r
a
cti
o
n
is
in
te
n
d
e
d
t
o
d
o
.
P
r
i
m
a
r
il
y
d
u
e
t
o
t
h
e
f
ac
t
t
h
a
t
ac
o
u
s
tic
s
i
g
n
als
i
n
c
lu
d
e
an
o
v
er
w
h
e
lm
in
g
a
m
o
u
n
t
o
f
d
at
a,
m
u
c
h
o
f
w
h
i
c
h
is
u
n
n
ec
ess
a
r
y
f
o
r
th
e
r
e
c
o
g
n
i
ti
o
n
p
r
o
ce
s
s
[
2
7
]
.
T
h
e
ex
tr
ac
t
io
n
o
f
a
u
d
io
f
ea
t
u
r
es
i
n
v
o
l
v
es
t
h
e
alt
er
ati
o
n
o
r
p
r
o
ce
s
s
i
n
g
o
f
a
u
d
i
o
s
i
g
n
als
b
y
tr
a
n
s
f
o
r
m
i
n
g
d
i
g
it
al
a
n
d
a
n
al
o
g
s
ig
n
a
ls
,
r
e
d
u
ci
n
g
u
n
d
esi
r
a
b
l
e
n
o
is
e
,
an
d
o
p
t
im
izi
n
g
t
h
e
t
i
m
e
-
f
r
e
q
u
e
n
cy
r
a
n
g
es
[
2
8
]
.
T
h
e
r
e
ar
e
m
u
lti
p
le
wa
y
s
t
o
p
a
r
a
m
e
te
r
iz
e
a
n
d
e
x
p
r
ess
th
e
v
o
i
ce
s
i
g
n
a
ls
f
o
r
t
h
e
d
ete
cti
o
n
p
r
o
c
ess
.
T
h
e
m
ai
n
u
s
e
d
m
et
h
o
d
s
ar
e
MFC
C
a
n
d
L
PC
.
T
ab
le
1
s
h
o
ws
a
co
m
p
ar
is
o
n
o
f
MFC
C
a
n
d
L
PC
.
T
ab
le
1
.
C
o
m
p
a
r
is
o
n
o
f
MFC
C
an
d
L
PC
F
e
a
t
u
r
e
/
t
e
c
h
n
i
q
u
e
M
F
C
C
LPC
B
a
si
s
F
i
l
t
e
r
b
a
n
k
c
o
e
f
f
i
c
i
e
n
t
s
M
o
d
e
l
e
d
b
y
a
l
l
P
o
l
e
mo
d
e
l
F
i
l
t
e
r
i
n
g
f
e
a
t
u
r
e
s
M
e
l
-
sca
l
e
S
o
u
r
c
e
-
f
i
l
t
e
r
mo
d
e
l
A
d
v
a
n
t
a
g
e
s
H
i
g
h
a
c
c
u
r
a
c
y
H
i
g
h
s
p
e
e
d
D
i
sad
v
a
n
t
a
g
e
s
S
e
n
s
i
t
i
v
e
t
o
b
a
c
k
g
r
o
u
n
d
n
o
i
se
H
i
g
h
l
y
s
e
n
s
i
t
i
v
e
t
o
q
u
a
n
t
i
z
i
n
g
n
o
i
s
e
MFC
C
is
th
e
m
o
s
t
co
m
m
o
n
ly
u
tili
ze
d
f
ea
tu
r
e
ex
tr
ac
tio
n
tech
n
iq
u
e
f
o
r
s
p
ee
ch
an
d
au
d
io
p
r
o
ce
s
s
in
g
[
2
9
]
.
I
t
is
wid
ely
em
p
lo
y
ed
in
s
p
ee
ch
r
ec
o
g
n
itio
n
s
y
s
tem
s
as
a
way
to
r
e
p
r
esen
t
th
e
s
ig
n
al
’
s
s
p
ec
tr
al
co
n
ten
t.
T
h
e
MFC
C
is
a
co
n
cise
r
ep
r
e
s
en
tatio
n
o
f
a
s
ig
n
al’
s
s
p
ec
tr
al
co
n
te
n
t
ex
p
r
ess
ed
as
a
co
llect
io
n
o
f
c
o
ef
f
icien
ts
.
T
h
e
MFC
C
h
as
its
b
asis
in
th
e
s
en
s
itiv
ity
o
f
th
e
h
u
m
an
h
ea
r
in
g
m
ec
h
an
is
m
to
f
r
eq
u
e
n
cy
ch
an
g
es
in
a
s
ig
n
al
[
3
0
]
.
I
n
g
en
er
al,
th
e
f
u
n
d
am
e
n
tal
s
tep
s
f
o
r
ca
lcu
latin
g
MF
C
C
s
ar
e
f
r
am
e
b
lo
ck
in
g
,
wh
ic
h
in
v
o
lv
es
d
iv
id
in
g
th
e
s
ig
n
al
in
to
s
h
o
r
t
f
r
am
es,
an
d
win
d
o
win
g
,
wh
ich
r
eq
u
ir
es
ap
p
ly
in
g
a
win
d
o
w
f
u
n
ctio
n
to
ea
ch
f
r
am
e
to
m
in
im
ize
s
p
ec
tr
al
lea
k
ag
e
[
3
1
]
.
Fo
llo
win
g
th
at,
ap
p
ly
th
e
f
ast
f
o
u
r
ier
tr
an
s
f
o
r
m
(
FF
T
)
to
ca
lcu
late
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
14
,
No
.
3
,
J
u
n
e
20
25
:
1
6
8
3
-
1
6
9
5
1686
d
is
cr
ete
f
o
u
r
ier
tr
an
s
f
o
r
m
(
DFT)
o
f
ea
ch
win
d
o
wed
f
r
a
m
e
to
ac
q
u
ir
e
r
ep
r
esen
tatio
n
s
in
th
e
f
r
eq
u
en
cy
d
o
m
ain
.
Su
b
s
eq
u
e
n
tly
,
u
tili
ze
a
Me
l
f
ilter
b
an
k
(
Me
l
f
r
eq
u
e
n
cy
wr
a
p
p
in
g
)
to
p
r
o
ce
s
s
th
e
p
o
wer
s
p
ec
tr
u
m
b
y
co
n
v
er
tin
g
th
e
f
r
eq
u
en
cies
to
th
e
Me
l
s
ca
le
an
d
s
u
b
s
eq
u
en
tly
im
p
lem
en
tin
g
tr
ian
g
le
f
ilter
s
.
On
ce
th
e
co
m
p
u
tatio
n
s
f
o
r
th
e
Me
l
f
ilter
b
an
k
o
u
t
p
u
ts
h
av
e
b
ee
n
co
m
p
leted
an
d
th
eir
lo
g
ar
ith
m
h
as
b
ee
n
ca
lcu
lated
,
th
e
r
esu
ltin
g
en
er
g
ies ar
e
d
ec
o
r
r
elate
d
u
s
in
g
th
e
d
is
cr
ete
co
s
in
e
tr
an
s
f
o
r
m
(
DC
T
)
.
T
h
is
p
r
o
ce
s
s
g
en
er
ates a
s
et
o
f
co
ef
f
icien
ts
r
ef
er
r
e
d
to
as c
ep
s
tr
al
co
ef
f
icien
ts
.
Fig
u
r
e
2
s
h
o
ws th
e
MFC
C
b
lo
ck
d
iag
r
a
m
.
Fig
u
r
e
2
.
MFC
C
b
lo
ck
d
iag
r
a
m
[
3
2
]
I
n
ad
d
itio
n
,
L
PC
is
a
co
m
m
o
n
m
eth
o
d
f
o
r
p
r
o
ce
s
s
in
g
au
d
i
o
s
ig
n
als,
n
o
tab
ly
s
p
ee
ch
.
I
n
th
e
f
ield
o
f
v
o
ice
s
ig
n
al
c
o
m
p
r
ess
io
n
,
it
h
as
p
r
o
v
en
v
er
y
u
s
ef
u
l,
all
o
win
g
f
o
r
ex
tr
e
m
ely
h
i
g
h
c
o
m
p
r
ess
io
n
r
ates.
L
PC
s
im
u
lates
th
e
h
u
m
an
v
o
ca
l
t
r
ac
t
b
y
m
o
d
elin
g
th
e
s
p
ec
tr
al
en
v
elo
p
e
o
f
a
d
ig
ital
s
p
e
ec
h
s
ig
n
al
with
a
m
ath
em
atica
l
m
o
d
el
[
3
3
]
,
[
3
4
]
.
T
h
e
in
itial
s
tep
s
o
f
co
m
p
u
tin
g
L
PC
ar
e
s
im
ilar
to
MF
C
C
wh
ich
in
clu
d
e:
f
r
am
e
b
lo
ck
in
g
an
d
win
d
o
win
g
.
Af
ter
th
at,
au
to
co
r
r
elatio
n
an
aly
s
is
in
v
o
lv
es
ca
lcu
latin
g
th
e
au
to
co
r
r
elatio
n
f
u
n
ctio
n
o
f
th
e
win
d
o
wed
f
r
a
m
e
[
2
7
]
.
S
u
b
s
eq
u
en
tly
,
th
e
L
PC
a
n
aly
s
is
p
r
o
ce
s
s
co
n
v
er
ts
ea
ch
au
to
c
o
r
r
elate
d
f
r
am
e
in
to
a
s
et
o
f
L
PC
p
ar
am
eter
s
,
wh
ich
ar
e
L
PC
co
ef
f
icien
ts
.
Fig
u
r
e
3
s
h
o
ws
th
e
L
PC
b
lo
ck
d
iag
r
am
.
Fo
r
au
d
ito
r
y
m
o
d
elin
g
,
L
PC
an
d
MFC
C
s
h
av
e
m
o
s
tly
e
m
p
lo
y
ed
f
ea
t
u
r
es.
B
ased
o
n
th
e
p
r
io
r
s
am
p
lin
g
,
L
PC
m
ak
es p
r
ed
ictio
n
s
ab
o
u
t f
u
tu
r
e
v
alu
es.
W
h
ile
th
e
MF
C
C
tak
es in
to
ac
co
u
n
t th
e
s
p
ee
ch
’
s
n
atu
r
e
d
u
r
in
g
f
ea
tu
r
e
ex
tr
ac
tio
n
[
3
5
]
.
Fig
u
r
e
3
.
L
PC
b
lo
ck
d
iag
r
a
m
[
3
6
]
2
.
3
.
M
o
delin
g
a
nd
re
co
g
niti
o
n
Af
ter
f
ea
tu
r
e
e
x
tr
ac
tio
n
,
th
e
n
ex
t
s
tep
is
to
m
o
d
el
th
e
s
p
ea
k
er
s
.
T
h
e
o
b
jectiv
e
o
f
m
o
d
elin
g
ap
p
r
o
ac
h
es
is
to
g
en
e
r
ate
s
p
ea
k
er
m
o
d
els
th
at
f
ac
ilit
ate
t
h
e
p
r
o
ce
s
s
o
f
f
ea
tu
r
e
m
atch
i
n
g
f
o
r
th
e
g
o
al
o
f
id
en
tify
in
g
th
e
s
p
ea
k
er
’
s
v
o
ic
e
[
3
7
]
.
T
h
er
e
a
r
e
d
if
f
er
en
t
m
o
d
elin
g
m
eth
o
d
s
.
T
h
e
class
ical
ap
p
r
o
ac
h
es
in
clu
d
e
v
ec
to
r
q
u
an
tizatio
n
(
VQ)
[
3
8
]
,
d
y
n
a
m
ic
tim
e
wr
ap
p
i
n
g
(
D
T
W
)
[
3
9
]
,
GM
M
[
4
0
]
,
an
d
HM
M
[
4
1
]
.
Ma
c
h
in
e
lear
n
in
g
tec
h
n
iq
u
es
ar
e
also
u
t
ilized
s
u
ch
as
SVM
[
4
2
]
an
d
ANN
[
4
3
]
.
Hy
b
r
id
a
p
p
r
o
ac
h
es
in
co
r
p
o
r
ate
two
o
r
m
o
r
e
o
f
th
e
p
r
ev
i
o
u
s
tech
n
iq
u
es
ar
e
also
u
s
ed
in
th
e
liter
atu
r
e
s
u
ch
as
wh
at
in
tr
o
d
u
ce
d
in
[
4
4
]
th
at
em
p
lo
y
s
b
o
th
ANN
an
d
HM
M.
3.
RE
S
E
ARCH
M
E
T
H
O
DO
L
O
G
Y
T
h
is
s
tu
d
y
is
ac
h
iev
ed
b
y
f
o
l
lo
win
g
th
ese
m
ajo
r
s
tep
s
:
i
)
i
d
en
tify
in
g
th
e
p
r
o
b
lem
,
an
d
t
h
e
r
elate
d
r
esear
ch
q
u
esti
o
n
s
,
ii
)
d
e
v
elo
p
in
g
th
e
k
ey
w
o
r
d
s
f
o
r
t
h
e
s
ea
r
ch
,
iii
)
d
ef
in
in
g
th
e
d
o
c
u
m
en
t
d
atab
ases
an
d
t
h
e
in
clu
s
io
n
an
d
e
x
clu
s
io
n
cr
iter
ia,
iv
)
s
ea
r
ch
in
g
th
e
d
o
c
u
m
en
t
d
atab
ases
,
v
)
d
e
f
in
in
g
th
e
d
ata
to
b
e
e
x
tr
ac
ted
f
r
o
m
th
e
d
o
c
u
m
en
ts
id
e
n
tifie
d
,
vi
)
r
ec
o
r
d
in
g
th
e
d
ata,
an
d
v
ii
)
s
y
n
th
esizin
g
an
d
r
ep
o
r
tin
g
th
e
r
esu
lts
.
T
h
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Tech
n
iq
u
es o
f Q
u
r
a
n
r
ec
iter
s
r
ec
o
g
n
itio
n
:
a
r
ev
iew
(
I
b
r
a
h
i
m
A
h
med
A
l
-
Oma
r
i
)
1687
m
ain
s
tep
s
ar
e
d
escr
ib
ed
as
f
o
llo
ws.
T
h
e
p
r
o
b
lem
s
tatem
e
n
t
is
illu
s
tr
ated
in
s
ec
tio
n
I
n
tr
o
d
u
ctio
n
,
an
d
th
e
r
elate
d
q
u
esti
o
n
s
ca
n
b
e
co
v
e
r
ed
with
th
ese
f
o
llo
win
g
q
u
esti
o
n
s
,
as sh
o
wn
b
y
T
ab
le
2
.
T
ab
le
2
.
R
esear
ch
q
u
esti
o
n
s
N
o
.
R
e
se
a
r
c
h
q
u
e
st
i
o
n
Q1
W
h
a
t
a
r
e
t
h
e
p
u
b
l
i
c
l
y
a
v
a
i
l
a
b
l
e
d
a
t
a
b
a
ses/
d
a
t
a
s
e
t
s
o
f
Q
u
r
a
n
i
c
r
e
c
i
t
a
t
i
o
n
s
,
a
n
d
h
o
w
d
o
t
h
e
s
e
d
a
t
a
s
e
t
s
d
i
f
f
e
r
i
n
q
u
a
l
i
t
y
?
Q2
W
h
a
t
a
r
e
t
h
e
f
e
a
t
u
r
e
s
e
x
t
r
a
c
t
i
o
n
t
e
c
h
n
i
q
u
e
s
t
h
a
t
u
se
d
t
o
d
e
f
i
n
e
r
e
c
i
t
e
r
s
f
e
a
t
u
r
e
s,
a
n
d
h
o
w
t
h
e
y
a
f
f
e
c
t
t
h
e
a
c
c
u
r
a
c
y
o
f
r
e
c
i
t
e
r
s re
c
o
g
n
i
t
i
o
n
t
e
c
h
n
i
q
u
e
s?
Q3
W
h
a
t
a
r
e
t
h
e
t
e
c
h
n
i
q
u
e
s t
h
a
t
u
se
d
t
o
r
e
c
o
g
n
i
z
e
/
d
e
t
e
r
mi
n
e
/
c
l
a
ssi
f
y
Q
u
r
a
n
r
e
c
i
t
e
r
s,
a
n
d
h
o
w
t
h
e
y
a
f
f
e
c
t
t
h
e
a
c
c
u
r
a
c
y
?
Q4
W
h
a
t
a
r
e
t
h
e
c
h
a
l
l
e
n
g
e
s
o
f
r
e
c
o
g
n
i
z
i
n
g
Q
u
r
a
n
r
e
c
i
t
e
r
s t
e
c
h
n
i
q
u
e
s?
T
h
ese
q
u
esti
o
n
s
ar
e
id
e
n
tifie
d
d
u
e
to
t
h
e
n
atu
r
e
o
f
th
e
Qu
r
a
n
r
ec
iter
s
r
ec
o
g
n
itio
n
p
r
o
ce
s
s
wh
er
e
th
e
k
ey
f
ac
to
r
s
o
f
r
ec
o
g
n
izi
n
g
p
r
o
ce
s
s
ar
e
two
f
o
ld
:
i)
th
e
f
ea
tu
r
e
ex
tr
ac
tio
n
tech
n
iq
u
e,
a
n
d
ii
)
th
e
r
ec
o
g
n
izin
g
tech
n
iq
u
e.
On
th
e
o
th
er
h
an
d
,
th
e
q
u
ality
o
f
th
e
r
esu
lts
is
a
f
f
ec
ted
b
y
th
e
q
u
ality
o
f
th
e
d
ata
in
ter
m
s
o
f
its
s
ize,
d
iv
er
s
ity
,
an
d
c
o
m
p
r
e
h
e
n
s
iv
en
ess
.
T
h
er
ef
o
r
e
,
s
tu
d
y
in
g
th
e
d
atasets
th
at
u
s
ed
in
p
r
ev
io
u
s
s
tu
d
ies
an
d
an
aly
zin
g
th
e
im
p
ac
t
o
f
its
q
u
ality
o
n
th
e
q
u
ality
o
f
th
e
r
esu
lts
an
d
id
en
tify
in
g
t
h
e
r
ea
s
o
n
s
f
o
r
th
at
wo
u
l
d
co
n
tr
ib
u
te
t
o
clar
if
y
i
n
g
o
n
e
o
f
th
e
m
o
s
t
im
p
o
r
tan
t
r
ea
s
o
n
s
af
f
ec
tin
g
t
h
e
q
u
ality
o
f
th
e
s
o
lu
tio
n
in
g
en
e
r
al.
Fin
ally
,
id
en
tify
i
n
g
th
e
c
h
allen
g
es
f
ac
in
g
r
esear
ch
er
s
in
t
h
is
f
ield
will
c
o
n
tr
ib
u
te
to
d
i
r
ec
tin
g
r
esear
ch
er
s
’
ef
f
o
r
ts
an
d
clar
if
y
in
g
p
r
o
p
o
s
e
d
r
esear
ch
p
at
h
s
in
th
e
f
u
tu
r
e.
An
s
wer
in
g
all
th
e
q
u
esti
o
n
s
an
d
s
p
ec
if
y
in
g
th
eir
d
etails
will
co
n
tr
ib
u
te
to
clar
i
f
y
in
g
th
e
f
ac
to
r
s
af
f
ec
tin
g
th
e
q
u
ality
o
f
th
e
m
o
d
els
u
s
ed
to
id
en
tify
Qu
r
a
n
r
ec
iter
s
an
d
b
u
ild
h
ig
h
-
ac
cu
r
ac
y
m
o
d
els,
an
d
th
e
n
g
en
er
alizin
g
t
h
e
r
esu
lts
s
cien
tific
ally
.
I
n
ad
d
itio
n
,
s
p
ec
if
ic
cr
iter
ia
wer
e
u
s
ed
to
ev
alu
a
te
an
d
class
if
y
th
e
co
llected
p
u
b
licatio
n
s
to
b
e
in
clu
d
ed
o
r
ex
cl
u
d
ed
in
th
is
s
tu
d
y
;
T
ab
le
3
illu
s
tr
ates
th
ese
cr
iter
ia.
T
h
e
m
o
s
t
o
f
in
clu
d
ed
p
u
b
licatio
n
s
wer
e
jo
u
r
n
al
p
a
p
er
s
as
s
h
o
wn
in
Fig
u
r
e
4
.
Fig
u
r
e
5
d
ep
icted
th
at
a
lim
ited
n
u
m
b
e
r
o
f
r
esear
ch
p
a
p
er
s
h
av
e
b
ee
n
p
u
b
lis
h
ed
b
etwe
en
2
0
1
2
an
d
p
r
e
s
en
t
in
th
is
a
r
ea
.
T
h
e
lack
o
f
r
esear
ch
in
th
is
ar
ea
o
p
en
s
th
e
d
o
o
r
f
o
r
r
esear
ch
er
s
to
f
o
cu
s
m
o
r
e
o
n
th
e
u
s
e
o
f
em
er
g
in
g
tech
n
o
lo
g
ies,
AI
an
d
d
ee
p
lear
n
in
g
tech
n
iq
u
es in
th
is
v
ital a
r
ea
.
T
h
e
d
if
f
er
e
n
t
r
esear
ch
q
u
esti
o
n
s
ar
e
s
y
s
tem
atica
lly
ad
d
r
ess
ed
in
th
e
f
o
llo
win
g
s
ec
tio
n
s
.
Sectio
n
4
ex
am
in
es
q
u
esti
o
n
s
Q1
an
d
Q2
.
W
h
ile
s
ec
tio
n
5
d
elv
es
i
n
to
Q1
.
Sectio
n
6
f
o
cu
s
es
o
n
Q4
an
d
p
r
o
v
id
es
a
co
m
p
r
eh
e
n
s
iv
e
an
aly
s
is
o
f
all
f
in
d
in
g
s
.
T
ab
le
3
.
I
n
clu
s
io
n
an
d
ex
clu
s
i
o
n
cr
iter
ia
f
o
r
co
llected
p
u
b
lic
atio
n
s
C
r
i
t
e
r
i
a
A
r
e
a
o
f
i
n
t
e
r
e
st
i
n
p
u
b
l
i
c
a
t
i
o
n
I
n
c
l
u
s
i
o
n
‒
U
t
i
l
i
z
e
s
p
e
a
k
e
r
r
e
c
o
g
n
i
t
i
o
n
f
o
r
Q
u
r
a
n
i
c
r
e
c
i
t
a
t
i
o
n
‒
R
e
c
o
g
n
i
z
i
n
g
Q
u
r
a
n
r
e
c
i
t
e
r
‒
R
e
c
o
g
n
i
z
i
n
g
Q
u
r
a
n
r
e
c
i
t
a
t
i
o
n
st
y
l
e
Ex
c
l
u
si
o
n
‒
P
u
b
l
i
c
a
t
i
o
n
s
w
i
t
h
o
u
t
e
x
p
e
r
i
me
n
t
a
l
r
e
s
u
l
t
s.
‒
N
o
t
p
u
b
l
i
s
h
e
d
in
t
h
e
E
n
g
l
i
s
h
l
a
n
g
u
a
g
e
Fig
u
r
e
4
.
T
y
p
e
o
f
p
u
b
licatio
n
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
14
,
No
.
3
,
J
u
n
e
20
25
:
1
6
8
3
-
1
6
9
5
1688
Fig
u
r
e
5
.
Nu
m
b
er
o
f
p
u
b
licatio
n
s
in
y
ea
r
s
4.
RE
CI
T
E
RS R
E
CO
G
NI
T
I
O
N
T
E
CH
N
I
Q
UE
S
Sev
er
al
s
tu
d
ies
h
av
e
b
ee
n
p
u
b
lis
h
ed
to
en
h
an
ce
a
n
d
im
p
r
o
v
e
Qu
r
an
ic
r
ec
iter
r
ec
o
g
n
itio
n
s
y
s
tem
s
.
I
n
ad
d
itio
n
,
a
r
a
n
g
e
o
f
a
p
p
r
o
a
ch
es
an
d
class
if
icatio
n
tech
n
i
q
u
es
wer
e
em
p
lo
y
e
d
an
d
e
v
alu
ated
,
as
s
h
o
w
n
in
Fig
u
r
e
6
.
T
h
is
s
ec
tio
n
ex
p
lo
r
e
s
th
e
ex
is
ten
ce
o
f
liter
atu
r
e
th
at
is
d
iv
id
ed
in
to
t
h
r
ee
s
u
b
s
ec
tio
n
s
ac
co
r
d
in
g
to
ca
teg
o
r
izatio
n
:
t
r
ad
itio
n
al,
m
a
ch
in
e
lear
n
in
g
,
an
d
d
ee
p
lear
n
in
g
m
eth
o
d
s
.
Fig
u
r
e
6
.
A
s
u
m
m
a
r
y
o
f
class
if
icatio
n
m
eth
o
d
s
u
s
ed
f
o
r
r
ec
o
g
n
izin
g
Qu
r
an
r
ec
iter
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Tech
n
iq
u
es o
f Q
u
r
a
n
r
ec
iter
s
r
ec
o
g
n
itio
n
:
a
r
ev
iew
(
I
b
r
a
h
i
m
A
h
med
A
l
-
Oma
r
i
)
1689
4
.
1
.
T
ra
ditio
na
l
m
et
ho
ds
Kh
elif
a
et
a
l
.
[
1
3
]
d
ev
elo
p
ed
a
n
ef
f
ec
tiv
e
ASR
-
b
ased
Qu
r
an
i
c
s
o
u
n
d
r
ec
o
g
n
itio
n
s
y
s
tem
to
r
ec
o
g
n
ize
an
d
id
e
n
tify
Qu
r
an
ic
s
o
u
n
d
s
.
T
h
e
f
ir
s
t
p
h
ase
s
tar
ts
with
a
b
aselin
e
HM
M
b
ased
s
y
s
te
m
f
o
r
f
u
n
d
am
en
tal
Qu
r
an
ic
s
o
u
n
d
s
.
T
h
e
s
ec
o
n
d
p
h
ase
im
p
r
o
v
es
th
is
b
aselin
e
s
y
s
tem
u
s
in
g
Qu
r
an
ic
s
o
u
n
d
d
u
r
atio
n
m
o
d
elin
g
tech
n
iq
u
es.
Gam
m
a,
Gau
s
s
ian
,
an
d
Po
is
s
o
n
d
is
tr
ib
u
tio
n
s
wer
e
ex
am
in
ed
a
n
d
in
co
r
p
o
r
ate
d
in
to
HM
M
tr
ain
in
g
an
d
d
ec
o
d
in
g
t
o
m
o
d
el
s
tate
d
u
r
atio
n
s
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
elin
g
tech
n
i
q
u
es
h
a
v
e
i
m
p
r
o
v
e
d
t
h
e
r
esu
lts
s
ig
n
if
ican
tly
with
9
9
% a
cc
u
r
a
cy
.
Mo
r
eo
v
er
,
B
aig
et
a
l
.
[
4
5
]
p
r
esen
ted
a
Qu
r
an
ic
r
ec
itatio
n
s
y
s
tem
th
at
r
ec
o
g
n
izes
r
ec
it
atio
n
o
n
a
p
h
o
n
e
m
e
b
asis
.
Fo
r
au
d
io
m
o
d
els,
m
in
im
u
m
p
h
o
n
e
er
r
o
r
(
MPE
)
,
an
d
m
ac
h
i
n
e
lea
r
n
in
g
,
ac
co
r
d
in
g
ly
,
non
-
d
is
cr
im
in
ativ
e
an
d
d
is
cr
im
in
ativ
e
tr
ain
in
g
m
eth
o
d
s
ar
e
u
s
ed
an
d
e
v
alu
ated
.
MPE
m
in
im
izes
p
h
o
n
e
er
r
o
r
s
to
p
r
o
d
u
ce
b
etter
r
esu
lts
wit
h
an
8
5
%
ac
cu
r
ac
y
r
ate.
M
PE
g
en
er
alize
s
u
n
s
ee
n
d
ata
b
etter
th
an
m
ac
h
in
e
lear
n
in
g
.
MPE
o
u
tp
er
f
o
r
m
s
m
ac
h
in
e
lear
n
in
g
wh
en
t
r
ain
ed
a
n
d
test
ed
o
n
n
o
is
y
d
atasets
.
L
ik
ewise,
a
tech
n
iq
u
e
f
o
r
id
e
n
tify
in
g
Q
u
r
an
ic
r
ec
iter
s
was
p
r
esen
ted
b
y
Gu
n
awa
n
et
a
l
.
[
1
6
]
.
T
h
e
GM
M
class
if
ier
an
d
MFC
C
f
ea
tu
r
es
wer
e
u
s
ed
.
Usi
n
g
f
if
t
ee
n
s
am
p
les
f
r
o
m
ea
ch
o
f
th
e
f
iv
e
r
ec
iter
s
,
th
e
y
b
u
ild
an
au
d
i
o
d
atab
ase
o
f
th
e
Qu
r
an
f
o
r
th
e
e
x
p
er
im
e
n
t.
ten
wer
e
u
tili
ze
d
f
o
r
tr
ain
in
g
p
u
r
p
o
s
es
an
d
f
iv
e
f
o
r
test
in
g
.
I
n
a
d
d
itio
n
,
th
ey
e
m
p
lo
y
an
ad
d
itio
n
al
u
n
k
n
o
wn
r
e
citer
to
ass
ess
th
e
ef
f
ec
tiv
en
e
s
s
o
f
th
e
p
r
o
p
o
s
ed
s
y
s
tem
.
Du
r
in
g
th
e
tr
ain
in
g
a
n
d
test
in
g
p
h
ase,
th
e
p
r
o
p
o
s
e
d
s
y
s
tem
d
em
o
n
s
tr
ated
1
0
0
%
ac
cu
r
ac
y
,
b
ased
o
n
th
e
r
esu
lts
.
1
0
0
% r
ejec
tio
n
r
at
e
was a
ls
o
r
ea
ch
ed
f
o
r
u
n
k
n
o
wn
s
am
p
les.
I
n
a
d
if
f
er
e
n
t
co
n
tex
t,
Ham
m
a
m
i
et
a
l
.
[
4
6
]
ad
v
an
ce
d
a
s
y
s
tem
th
at
ca
n
r
ep
ly
to
u
s
er
v
o
ic
e
r
eq
u
ests
b
y
r
ec
itin
g
a
p
o
r
tio
n
o
f
th
e
Q
u
r
an
in
lin
e
with
th
e
u
s
er
i
n
p
u
t.
T
h
e
s
y
s
tem
aim
s
to
en
ab
le
u
s
er
ex
p
lo
r
atio
n
an
d
n
av
ig
atio
n
b
etwe
en
Qu
r
a
n
ic
v
er
s
es o
r
p
ag
es u
s
in
g
v
o
ice
r
ec
o
g
n
itio
n
o
f
Ar
ab
ic
n
u
m
er
als.
T
h
e
s
y
s
tem
is
u
s
ef
u
l
f
o
r
th
o
s
e
wh
o
ar
e
b
li
n
d
o
r
v
is
u
ally
im
p
air
ed
.
GM
M
is
u
s
ed
with
MFC
C
as
a
clas
s
if
ier
.
T
h
e
p
r
o
p
o
s
ed
s
o
lu
tio
n
ac
h
iev
ed
a
9
9
.
3
1
% a
cc
u
r
ac
y
r
ate.
Al
-
J
ar
r
ah
et
a
l.
[
4
7
]
p
r
o
p
o
s
ed
an
en
h
an
ce
d
L
in
d
e
–
B
u
zo
–
Gr
ay
(
L
B
G)
al
g
o
r
ith
m
f
o
r
r
ec
o
g
n
izin
g
Ho
ly
Qu
r
a
n
r
ec
iter
s
.
T
h
e
o
r
i
g
in
al
L
B
G
alg
o
r
ith
m
is
a
n
it
er
ativ
e
VQ
alg
o
r
ith
m
to
i
m
p
r
o
v
e
a
s
m
all
s
et
o
f
v
ec
to
r
s
(
co
d
eb
o
o
k
)
t
o
r
ep
r
esen
t
a
lar
g
er
s
et
o
f
v
ec
to
r
s
(
tr
ai
n
in
g
s
et)
,
s
u
ch
th
at
it
will
b
e
l
o
ca
lly
o
p
tim
al.
T
h
e
co
d
eb
o
o
k
is
a
co
m
p
r
ess
ed
r
ep
r
esen
tatio
n
o
f
th
e
o
r
i
g
in
al
d
ata
.
T
h
e
b
asic
id
ea
o
f
L
B
G
is
to
d
iv
id
e
th
e
g
r
o
u
p
o
f
tr
ain
in
g
v
ec
to
r
s
an
d
u
s
e
it to
f
in
d
th
e
m
o
s
t r
ep
r
esen
tativ
e
v
ec
to
r
f
r
o
m
o
n
e
g
r
o
u
p
.
T
h
ese
r
ep
r
esen
tativ
e
v
ec
to
r
s
f
r
o
m
ea
ch
g
r
o
u
p
ar
e
g
at
h
er
e
d
to
f
o
r
m
t
h
e
co
d
e
b
o
o
k
.
T
h
e
en
h
an
ce
d
alg
o
r
ith
m
th
at
p
r
o
p
o
s
ed
in
[
4
7
]
is
a
m
eth
o
d
to
alig
n
th
e
ce
n
tr
o
id
s
o
f
co
d
eb
o
o
k
s
m
o
r
e
ac
cu
r
ate
th
an
th
e
o
r
ig
in
al
L
B
G.
T
h
e
s
ig
n
als
wer
e
d
er
iv
ed
f
r
o
m
v
o
ices
o
f
1
4
ex
p
e
r
t
r
ec
iter
s
.
T
h
e
r
ec
iter
s
m
ain
ly
r
ec
ite
d
f
r
o
m
Su
r
ah
“
Al
-
Kaw
th
ar
”
f
o
r
th
is
p
u
r
p
o
s
e.
T
h
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
ac
h
iev
e
d
h
ig
h
lev
el
ac
c
u
r
ac
y
with
9
8
.
2
1
.
As
a
way
to
d
ete
r
m
in
e
w
h
o
r
ec
ites
th
e
Ho
ly
Qu
r
an
,
th
is
s
t
u
d
y
p
r
esen
ts
a
r
ec
iter
r
ec
o
g
n
itio
n
s
y
s
tem
[
4
8
]
.
Fo
r
r
ec
iter
class
if
icatio
n
,
th
e
MFC
C
an
aly
s
is
an
d
L
B
G
-
VQ
wer
e
u
s
ed
.
T
h
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
h
as
th
e
p
o
ten
tial
to
id
en
tif
y
th
e
Ho
l
y
Qu
r
an
r
e
citer
in
b
o
t
h
liv
e
-
s
tr
ea
m
ed
an
d
r
ec
o
r
d
e
d
b
r
o
a
d
ca
s
ts
.
A
d
atab
ase
co
n
tain
in
g
2
0
0
s
am
p
les,
co
m
p
r
is
in
g
r
ec
o
r
d
in
g
s
m
ad
e
b
y
2
0
r
ec
iter
s
,
was
u
tili
ze
d
in
t
h
e
ex
p
e
r
im
en
t.
T
h
e
r
esu
lts
s
h
o
w
th
at
f
o
r
clea
n
s
a
m
p
les,
a
r
ec
o
g
n
itio
n
r
ate
o
f
8
6
.
5
% h
as b
ee
n
ac
h
iev
ed
.
4
.
2
.
M
a
chine le
a
rning
m
et
h
o
ds
Alk
h
atee
b
[
4
9
]
in
tr
o
d
u
ce
d
a
m
o
d
el
b
ased
o
n
ANN
an
d
KN
N
as c
las
s
if
ier
s
,
MF
C
C
is
u
s
e
d
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
.
Usi
n
g
th
e
ANN,
th
e
p
r
o
p
o
s
ed
s
y
s
tem
g
iv
es
9
7
.
6
2
%
ac
cu
r
ac
y
f
o
r
ch
ap
ter
1
8
a
n
d
9
6
.
7
%
ac
cu
r
ac
y
f
o
r
c
h
ap
ter
3
6
.
On
th
e
o
th
er
h
an
d
,
th
e
p
r
o
p
o
s
ed
s
y
s
tem
g
i
v
es
9
7
.
0
3
%
ac
cu
r
ac
y
f
o
r
c
h
a
p
ter
1
8
an
d
9
6
.
0
8
%
ac
cu
r
ac
y
f
o
r
ch
a
p
ter
3
6
b
y
u
s
in
g
th
e
KNN.
Asd
a
et
a
l.
[
1
8
]
d
esig
n
ed
a
s
y
s
tem
th
at
ca
n
id
en
tif
y
th
e
s
p
e
ak
er
’
s
v
o
ice
as
a
u
n
iq
u
e
b
io
m
etr
ic
s
ig
n
al
to
r
ec
o
g
n
ize
th
e
r
ec
iter
.
T
h
is
r
esear
ch
u
s
ed
ANN
an
d
MFC
C
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
.
T
h
e
m
is
m
atch
er
r
o
r
r
ate
h
as
d
ec
r
ea
s
ed
b
y
r
etr
ain
in
g
t
h
e
n
etwo
r
k
an
d
in
cr
ea
s
in
g
th
e
s
ize
o
f
th
e
h
id
d
en
lay
er
.
T
h
e
r
esu
lts
ac
h
iev
ed
9
1
.
2
%
ac
cu
r
ac
y
.
Nah
ar
et
a
l
.
[
5
0
]
in
v
esti
g
ated
th
e
r
ec
o
g
n
iti
o
n
o
f
Qu
r
a
n
r
ec
iter
s
u
s
in
g
SVM
an
d
ANN.
T
h
is
r
esear
ch
u
s
ed
a
c
o
r
p
u
s
c
o
n
tai
n
s
ten
r
ec
itatio
n
ty
p
es.
T
h
e
r
e
s
u
lts
d
em
o
n
s
tr
ated
th
at
th
e
b
e
s
t
r
esu
lts
h
av
e
b
ee
n
ac
h
iev
ed
u
s
in
g
SVM
with
9
6
% a
cc
u
r
ac
y
.
A
Ho
ly
Qu
r
an
r
ec
iter
id
e
n
tific
atio
n
s
y
s
tem
th
at
u
tili
ze
s
s
o
u
n
d
wav
es
to
r
ep
r
esen
t
th
e
p
r
o
n
u
n
ciatio
n
o
f
v
er
s
es
was
in
tr
o
d
u
ce
d
in
[
5
1
]
.
Af
ter
f
ea
tu
r
es
ex
tr
ac
tio
n
s
u
s
in
g
MFC
C
,
SVM
,
an
d
AN
N
u
s
ed
in
d
iv
id
u
ally
to
id
en
tify
t
h
e
r
ec
iter
.
T
h
e
o
b
tain
ed
f
in
d
i
n
g
s
d
em
o
n
s
tr
ate
th
at
th
e
SVM
o
u
tp
er
f
o
r
m
ed
th
e
ANN
with
9
6
.
5
9
%
ac
cu
r
ac
y
r
ate.
Sh
ah
an
d
Ah
s
an
[
5
2
]
u
s
ed
a
co
m
b
in
atio
n
o
f
L
PC
an
d
d
is
cr
ete
wav
elet
tr
an
s
f
o
r
m
(
DW
T
)
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
.
R
an
d
o
m
f
o
r
est
(
R
F)
class
if
ier
is
u
s
ed
.
Acc
o
r
d
i
n
g
to
th
e
s
tu
d
y
,
tr
ain
in
g
th
e
R
F
cla
s
s
if
ier
u
s
in
g
L
PC
o
r
D
W
T
s
ep
ar
ately
r
ed
u
ce
d
th
e
ac
cu
r
ac
y
,
wh
ile
tr
ain
in
g
u
s
in
g
f
ea
tu
r
es e
x
tr
ac
ted
b
y
b
o
th
L
PC
an
d
DW
T
im
p
r
o
v
e
d
th
e
r
ec
o
g
n
itio
n
ac
cu
r
ac
y
.
T
h
e
r
esu
lts
ac
h
iev
ed
9
0
.
9
0
% a
cc
u
r
ac
y
r
ate.
A
n
o
v
el
ap
p
r
o
ac
h
f
o
r
r
ec
o
g
n
i
zin
g
Ho
ly
Qu
r
an
r
ec
iter
s
was
in
tr
o
d
u
ce
d
in
[
5
3
]
.
T
h
e
a
u
th
o
r
s
u
s
ed
two
d
if
f
er
en
t
tech
n
iq
u
es
f
o
r
au
d
i
o
r
ep
r
esen
tatio
n
.
T
h
e
f
ir
s
t
o
n
e
is
th
e
au
d
io
s
ig
n
al
i
n
th
e
f
r
eq
u
en
cy
d
o
m
ain
.
T
h
e
s
ec
o
n
d
o
n
e
was
co
n
s
id
er
in
g
th
e
au
d
io
as
im
ag
es
b
y
u
tili
zin
g
t
h
e
Sp
ec
tr
o
g
r
am
.
T
h
r
ee
class
if
ier
s
,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
14
,
No
.
3
,
J
u
n
e
20
25
:
1
6
8
3
-
1
6
9
5
1690
in
clu
d
in
g
n
aïv
e
B
ay
es
(
NB
)
,
J
4
8
,
an
d
R
F,
wer
e
em
p
lo
y
ed
t
o
ef
f
ec
tiv
ely
lear
n
th
e
d
is
tin
ctio
n
b
etwe
en
class
es.
A
d
atab
ase
o
f
1
2
0
au
d
io
f
iles
o
f
Q
u
r
an
ic
r
ec
itatio
n
s
f
r
o
m
1
2
d
i
f
f
er
en
t
r
ec
iter
s
was
u
tili
ze
d
.
T
h
e
r
esu
lts
in
d
icate
d
th
at
th
e
h
ig
h
est
lev
el
o
f
ac
cu
r
ac
y
,
with
8
8
%,
wa
s
ac
h
iev
ed
u
s
in
g
R
F
class
if
ier
an
d
b
y
u
s
in
g
th
e
au
d
io
s
ig
n
als f
o
r
r
ep
r
esen
tin
g
th
e
d
ata.
4
.
3
.
Dee
p lea
rning
m
e
t
ho
ds
Q
a
y
y
u
m
e
t
a
l
.
[
5
4
]
l
e
v
e
r
a
g
e
t
h
e
p
o
w
e
r
o
f
R
N
N
a
b
i
li
t
ie
s
f
o
r
ev
a
l
u
a
t
i
n
g
a
n
d
m
o
d
e
li
n
g
s
e
q
u
e
n
t
i
a
l
s
p
e
e
c
h
d
a
t
a
,
n
a
m
e
l
y
b
i
d
i
r
ec
t
i
o
n
a
l
l
o
n
g
s
h
o
r
t
-
t
e
r
m
m
e
m
o
r
y
(
B
L
S
T
M
)
.
A
s
i
n
p
u
t
,
t
h
e
m
o
d
e
l
w
a
s
g
i
v
e
n
r
e
c
i
t
a
ti
o
n
s
f
r
o
m
f
i
v
e
w
e
l
l
-
k
n
o
w
n
r
e
c
i
t
e
r
s
.
M
FC
C
i
s
u
s
e
d
f
o
r
f
e
a
t
u
r
e
e
x
t
r
a
ct
i
o
n
.
S
e
v
e
r
a
l
r
e
c
i
ta
t
i
o
n
l
e
n
g
t
h
s
a
n
d
p
e
r
f
o
r
m
a
n
c
e
i
n
d
i
c
a
t
o
r
s
w
e
r
e
u
t
il
i
z
e
d
t
o
c
a
r
r
y
o
u
t
a
c
o
m
p
a
r
a
t
i
v
e
a
n
a
l
y
s
is
o
f
B
L
S
T
M
i
n
c
o
n
j
u
n
c
t
i
o
n
wi
t
h
v
a
r
i
o
u
s
b
as
e
l
i
n
e
c
l
a
s
s
i
f
ie
r
s
.
B
a
s
e
d
o
n
t
h
e
f
i
n
d
i
n
g
s
,
B
L
S
T
M
w
a
s
b
e
t
t
e
r
wi
t
h
9
9
.
8
9
%
a
c
c
u
r
a
c
y
t
h
a
n
o
t
h
e
r
c
o
m
m
o
n
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
s
(
D
NN
s
)
i
n
t
e
r
m
s
o
f
th
e
i
r
a
b
i
l
it
y
t
o
r
e
p
r
e
s
e
n
t
t
h
e
t
e
m
p
o
r
a
l
d
y
n
a
m
i
c
s
o
f
Q
u
r
a
n
r
e
c
it
at
i
o
n
s
.
A
d
ee
p
lear
n
in
g
m
o
d
el
th
at
ca
n
d
is
tin
g
u
is
h
b
etwe
en
au
th
en
tic
an
d
f
alse
Qu
r
an
r
ec
itatio
n
s
was
p
r
esen
ted
in
[
5
5
]
.
T
h
e
p
r
o
p
o
s
ed
s
o
lu
tio
n
ac
h
iev
ed
a
r
em
ar
k
ab
le
ac
cu
r
ac
y
r
ate
with
9
9
.
6
6
%.
T
h
e
r
esu
lts
al
s
o
d
em
o
n
s
tr
ated
th
at
t
h
e
d
ee
p
lea
r
n
in
g
p
r
o
p
o
s
ed
m
o
d
el
o
u
tp
er
f
o
r
m
ed
th
e
class
ical
m
ac
h
in
e
lear
n
in
g
m
eth
o
d
s
o
n
th
e
s
am
e
d
ataset.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
was
u
tili
ze
d
to
co
n
d
u
ct
s
ev
er
al
co
m
p
ar
is
o
n
s
o
f
s
eg
m
en
t
len
g
th
an
d
f
ea
tu
r
e
n
u
m
b
e
r
to
id
en
tif
y
th
e
id
ea
l
v
alu
es
th
at
wo
u
ld
r
esu
lt
in
t
h
e
m
a
x
im
u
m
ac
h
iev
ab
le
ac
c
u
r
ac
y
.
T
h
e
ex
p
er
im
e
n
tal
f
in
d
i
n
g
s
in
d
icate
th
at
a
s
eg
m
en
t
len
g
th
o
f
4
s
ec
o
n
d
s
an
d
a
f
ea
tu
r
e
co
u
n
t
o
f
3
0
y
ield
ex
ce
llen
t r
esu
lts
.
T
all
et
a
l
.
[
5
6
]
in
t
r
o
d
u
ce
d
a
n
i
n
n
o
v
ativ
e
ap
p
r
o
ac
h
u
tili
zin
g
d
ee
p
lear
n
i
n
g
tec
h
n
iq
u
es
to
r
ec
o
g
n
ize
t
h
e
r
ec
iter
,
s
u
r
a,
an
d
v
er
s
e
o
f
a
Qu
r
an
ic
r
ec
itatio
n
.
Fu
r
th
er
m
o
r
e,
th
ey
d
em
o
n
s
tr
ated
th
e
ef
f
i
ca
cy
o
f
em
p
lo
y
in
g
p
r
e
-
tr
ain
ed
em
b
ed
d
in
g
m
o
d
el
s
to
ca
teg
o
r
ize
v
ar
io
u
s
r
ec
ite
r
s
.
Mo
r
eo
v
er
,
th
ey
in
d
icate
d
th
at
t
h
e
a
p
p
r
o
ac
h
ac
h
iev
ed
r
e
m
ar
k
ab
le
p
r
ec
is
io
n
in
r
ec
o
g
n
izin
g
th
e
r
ec
iter
o
f
a
p
r
o
v
i
d
ed
p
ass
ag
e.
T
h
e
e
x
p
er
i
m
en
tal
d
ataset
was
ac
q
u
ir
ed
f
r
o
m
m
u
ltip
le
o
p
en
-
s
o
u
r
ce
p
latf
o
r
m
s
.
Ad
d
itio
n
all
y
,
a
class
co
m
p
r
is
in
g
ir
r
elev
a
n
t
s
o
u
n
d
s
,
s
u
ch
as
m
u
s
ic
o
r
d
is
cu
s
s
io
n
,
was
ap
p
en
d
ed
to
th
e
r
ec
iter
s
to
p
r
ev
en
t
th
e
m
o
d
el
f
r
o
m
m
is
class
i
f
y
in
g
a
n
u
n
r
elate
d
s
o
u
n
d
as
a
r
ec
iter
.
T
h
e
ex
p
e
r
im
en
ts
ac
h
iev
ed
an
ex
ce
llen
t
r
esu
lt
b
y
u
s
in
g
th
e
T
R
I
p
L
et
lo
s
s
n
etwo
r
k
(
T
r
ill)
m
o
d
el,
wh
ich
r
ea
ch
ed
an
ac
cu
r
ac
y
o
f
9
8
%.
Sab
er
et
a
l
.
[
5
7
]
p
r
o
p
o
s
ed
an
ap
p
r
o
ac
h
u
tili
zin
g
d
ee
p
tr
a
n
s
f
er
lear
n
in
g
m
o
d
els
to
r
ec
o
g
n
i
ze
a
Ho
ly
Qu
r
an
r
ec
iter
.
T
h
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
ev
alu
ates
s
ix
p
r
e
-
tr
ain
ed
d
ee
p
lear
n
in
g
m
o
d
e
ls
th
at
h
av
e
b
ee
n
in
co
r
p
o
r
ated
s
ep
ar
ately
.
Fu
r
th
er
m
o
r
e,
a
d
ataset
o
f
th
e
Ho
ly
Qu
r
an
is
co
n
s
tr
u
cted
b
y
co
n
v
er
tin
g
au
d
i
o
f
iles
in
to
im
ag
e
-
b
ased
v
is
u
al
r
ep
r
e
s
en
tatio
n
s
u
s
in
g
MFC
C
f
ea
tu
r
es
ex
tr
ac
ted
f
r
o
m
1
1
,
0
0
0
au
d
i
o
f
iles
.
T
h
e
d
ataset
in
clu
d
es
r
ec
o
r
d
in
g
s
f
r
o
m
2
0
d
if
f
er
en
t
r
ec
iter
s
.
Alo
n
g
with
th
e
p
r
e
-
tr
ain
e
d
m
o
d
els,
NASNetL
ar
g
e
ac
h
iev
ed
th
e
b
est
ac
cu
r
ac
y
r
ate
o
f
9
8
.
5
0
%
ac
co
r
d
i
n
g
t
o
th
e
r
esu
lt.
T
a
b
le
4
p
r
o
v
id
es
a
s
u
m
m
ar
y
o
f
r
esear
ch
p
ap
e
r
s
th
a
t
ex
p
lo
r
ed
th
e
r
ec
o
g
n
itio
n
o
f
Q
u
r
an
r
ec
iter
s
.
T
ab
le
4
.
Su
m
m
a
r
y
o
f
tech
n
i
q
u
es u
s
ed
in
Qu
r
an
r
ec
iter
s
r
ec
o
g
n
itio
n
s
tu
d
ies
R
e
f
.
Y
e
a
r
C
l
a
s
si
f
i
c
a
t
i
o
n
t
e
c
h
n
i
q
u
e
D
a
t
a
s
e
t
F
e
a
t
u
r
e
s
e
x
t
r
a
c
t
i
o
n
t
e
c
h
n
i
q
u
e
B
e
st
a
c
c
u
r
a
c
y
r
a
t
e
[
4
8
]
2
0
1
2
LB
G
-
VQ
2
0
0
s
a
m
p
l
e
s fr
o
m
2
0
r
e
c
i
t
e
r
s
M
F
C
C
8
6
.
5
%
[
4
6
]
2
0
1
3
G
M
M
-
M
F
C
C
9
9
.
3
1
%
[
5
2
]
2
0
1
4
RF
9
6
a
u
d
i
o
sam
p
l
e
s fr
o
m
6
r
e
c
i
t
e
r
s
D
W
T,
LP
C
9
0
.
9
0
%
[
4
5
]
2
0
1
5
H
M
M
-
M
F
C
C
8
5
%
[
1
8
]
2
0
1
6
ANN
Q
u
r
a
n
i
c
r
e
c
i
t
a
t
i
o
n
f
r
o
m
5
d
i
f
f
e
r
e
n
t
r
e
c
i
t
e
r
s
M
F
C
C
9
1
.
2
%
[
1
3
]
2
0
1
7
H
M
M
-
G
M
M
5
9
3
5
a
u
d
i
o
s
a
mp
l
e
s
f
r
o
m
1
0
r
e
c
i
t
e
r
s
M
F
C
C
9
9
%
[
1
6
]
2
0
1
8
G
M
M
1
5
a
u
d
i
o
sam
p
l
e
s fr
o
m
5
r
e
c
i
t
e
r
s
M
F
C
C
1
0
0
%
[
5
4
]
2
0
1
8
K
N
N
,
R
F
,
D
T
,
L
R
,
S
V
M
,
B
LST
M
Q
u
r
a
n
i
c
r
e
c
i
t
a
t
i
o
n
f
r
o
m
5
d
i
f
f
e
r
e
n
t
r
e
c
i
t
e
r
s
M
F
C
C
B
LST
M
=
9
9
.
8
9
%
[
5
3
]
2
0
1
9
N
B
,
J
4
8
,
R
F
1
2
0
s
a
m
p
l
e
s fr
o
m
1
2
r
e
c
i
t
e
r
s
M
F
C
C
R
F
=
8
8
%
[
5
1
]
2
0
1
9
S
V
M
,
A
N
N
C
o
r
p
u
s
o
f
1
5
d
i
f
f
e
r
e
n
t
r
e
c
i
t
e
r
s
M
F
C
C
S
V
M
=
9
6
.
5
9
%
[
5
0
]
2
0
2
0
S
V
M
,
A
N
N
C
o
r
p
u
s
o
f
1
0
d
i
f
f
e
r
e
n
t
r
e
c
i
t
e
r
s
M
F
C
C
S
V
M
=
9
6
%
[
4
9
]
2
0
2
0
K
N
N
,
A
N
N
C
o
r
p
u
s
o
f
1
0
d
i
f
f
e
r
e
n
t
r
e
c
i
t
e
r
s
M
F
C
C
A
N
N
=
9
7
.
6
2
%
[
4
7
]
2
0
2
2
LB
G
-
VQ
S
a
mp
l
e
s fr
o
m1
4
r
e
c
i
t
e
r
s
M
F
C
C
9
8
.
2
1
%
[
5
5
]
2
0
2
3
C
N
N
s
S
a
mp
l
e
s fr
o
m
7
r
e
c
i
t
e
r
s
M
F
C
C
9
9
.
6
6
%
[
5
6
]
2
0
2
3
TR
I
LL,
V
G
G
i
s
h
1
7
0
s
a
m
p
l
e
s fr
o
m
d
i
f
f
e
r
e
n
t
r
e
c
i
t
e
r
M
F
C
C
Tr
i
l
l
=
9
8
%
[
5
7
]
2
0
2
4
N
A
S
N
e
t
M
o
b
i
l
e
,
N
A
S
N
e
t
La
r
g
e
,
Ef
f
i
c
i
e
n
t
N
e
t
B
7
,
Ef
f
i
c
i
e
n
t
N
e
t
V
2
S
,
Ef
f
i
c
i
e
n
t
N
e
t
V
2
M
,
Ef
f
i
c
i
e
n
t
N
e
t
V
1
1
K
sam
p
l
e
s fr
o
m
2
0
r
e
c
i
t
e
r
s
M
F
C
C
N
A
S
N
e
t
La
r
g
e
=
9
8
.
5
0
%
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Tech
n
iq
u
es o
f Q
u
r
a
n
r
ec
iter
s
r
ec
o
g
n
itio
n
:
a
r
ev
iew
(
I
b
r
a
h
i
m
A
h
med
A
l
-
Oma
r
i
)
1691
5.
Q
URAN
I
C
DA
T
A
SE
T
S
I
n
a
s
y
s
tem
d
esig
n
ed
f
o
r
t
h
e
r
ec
o
g
n
itio
n
o
f
Q
u
r
an
ic
r
ec
iter
,
th
e
d
ata
m
ay
b
e
o
b
tain
ed
b
y
eith
er
th
e
r
ec
o
r
d
in
g
o
f
in
d
i
v
id
u
al
r
ec
itat
io
n
s
o
r
s
o
u
r
ce
d
f
r
o
m
a
p
r
e
-
ex
is
tin
g
co
r
p
u
s
o
f
s
tan
d
ar
d
ized
Qu
r
an
ic
r
ec
itatio
n
s
[
1
4
]
.
T
h
e
wav
ef
o
r
m
au
d
io
f
il
e
f
o
r
m
at,
g
e
n
er
ally
r
ef
e
r
r
ed
t
o
as
.
wav
,
is
th
e
m
ajo
r
au
d
io
f
o
r
m
at
p
r
e
f
er
r
e
d
b
y
r
esear
ch
er
s
.
A
p
r
o
p
er
c
h
o
ice
o
f
d
ata
is
n
ec
ess
ar
y
f
o
r
a
n
y
r
ec
iter
r
ec
o
g
n
itio
n
s
y
s
tem
t
o
d
ec
r
ea
s
e
th
e
tim
e
r
eq
u
ir
ed
f
o
r
a
d
d
itio
n
al
p
r
e
-
p
r
o
ce
s
s
in
g
.
Ma
n
y
r
esear
ch
e
r
s
r
ely
o
n
tr
ad
itio
n
al
an
d
n
ativ
e
l
an
g
u
ag
e
d
atab
ases
,
d
ep
en
d
i
n
g
o
n
th
e
n
ee
d
s
o
f
th
e
s
tu
d
y
to
p
ic.
Par
ticu
lar
r
eq
u
ir
e
m
en
ts
m
u
s
t
b
e
m
et
wh
en
s
elec
tin
g
a
d
atab
ase
f
o
r
a
r
ec
iter
r
ec
o
g
n
itio
n
s
y
s
tem
.
As
clar
if
ied
in
s
ec
tio
n
4
,
it
wa
s
o
b
s
er
v
ed
th
at
s
ev
er
al
s
tu
d
ies
m
en
tio
n
th
e
u
s
e
o
f
s
p
ec
if
ic
d
atasets
f
o
r
th
eir
r
esear
ch
.
Ho
wev
er
,
d
eta
iled
s
o
u
r
ce
s
o
r
r
ef
er
en
c
es
f
o
r
th
ese
d
atasets
wer
e
n
o
t
co
n
s
is
ten
tly
p
r
o
v
id
ed
.
Du
e
to
th
e
d
is
tin
ctiv
e
ch
ar
ac
ter
o
f
Qu
r
an
ic
r
ec
itatio
n
s
,
a
f
ew
d
atasets
o
f
Qu
r
an
ic
r
ec
itatio
n
s
h
av
e
b
ee
n
co
llected
an
d
p
u
b
lis
h
ed
in
th
e
p
ast
f
ew
y
ea
r
s
.
T
h
ese
d
atasets
ca
n
b
e
u
s
ed
f
o
r
v
ar
i
o
u
s
task
s
,
in
clu
d
in
g
b
u
t
n
o
t
lim
ited
to
ASR
.
Mo
s
t
o
f
t
h
e
a
v
ailab
le
Qu
r
a
n
ic
au
d
io
d
atasets
ar
e
ju
s
t
f
o
ld
er
s
o
f
au
d
io
f
il
es,
th
at
m
a
y
n
ee
d
a
lo
t
o
f
p
r
e
-
p
r
o
ce
s
s
in
g
b
ef
o
r
e
th
ey
ca
n
b
e
em
p
l
o
y
ed
i
n
d
ata
-
d
r
iv
e
n
d
e
v
elo
p
m
e
n
t.
T
ab
le
5
p
r
esen
ts
a
co
m
p
ar
is
o
n
o
f
s
o
m
e
p
u
b
licly
av
ailab
le
d
atasets
.
6.
DIS
CU
SS
I
O
N
AND
O
P
E
N
I
SS
UE
S
T
h
is
s
ec
tio
n
in
tr
o
d
u
ce
s
a
d
is
cu
s
s
io
n
ab
o
u
t
wh
at
h
as
b
ee
n
in
tr
o
d
u
ce
d
in
s
ec
tio
n
s
4
an
d
5
.
I
t
also
h
ig
h
lig
h
ts
s
o
m
e
r
esear
ch
g
ap
s
an
d
o
p
e
n
is
s
u
es
in
th
e
f
ield
.
Star
tin
g
with
d
atasets
,
as
th
ey
ar
e
th
e
b
asis
th
at
th
e
d
if
f
e
r
en
t
p
r
o
ce
s
s
in
g
p
h
as
es
b
u
ilt
u
p
o
n
.
As
s
h
o
wn
i
n
s
ec
tio
n
5
a
n
d
T
a
b
le
5
,
th
e
r
e
ar
e
f
ew
d
atasets
av
ailab
le
an
d
it
s
u
f
f
er
s
m
an
y
p
r
o
b
lem
s
th
at
m
ak
e
it
n
o
t
s
u
itab
le
f
o
r
s
o
lid
e
x
p
er
im
e
n
ts
.
So
m
e
o
f
th
ese
p
r
o
b
lem
s
ad
d
r
ess
ed
ar
e:
th
e
lim
it
s
ize,
th
e
n
u
m
b
er
o
f
r
ec
iter
s
d
id
n
o
t
ex
ce
ed
2
0
r
ec
iter
s
in
th
e
b
est
d
ataset,
r
ec
iter
s
alm
o
s
t
wer
e
Ar
ab
ic
r
ec
iter
s
,
an
d
s
o
m
e
s
tu
d
ies
co
llect
th
e
d
ata
b
ased
o
n
in
d
iv
id
u
al
ef
f
o
r
t
f
o
r
s
o
m
e
v
o
lu
n
teer
r
ec
iter
s
wh
o
a
r
e
n
o
t
h
ig
h
ly
s
k
illed
.
I
n
a
d
d
itio
n
,
th
er
e
is
n
o
d
iv
er
s
ity
o
f
r
ep
r
esen
tatio
n
f
o
r
m
ats,
n
o
in
v
o
lv
in
g
a
r
ea
s
o
n
a
b
le
am
o
u
n
t
o
f
n
o
is
e,
a
n
d
n
o
in
clu
d
in
g
v
ar
io
u
s
len
g
th
s
o
f
au
d
i
o
f
iles
.
T
h
er
ef
o
r
e,
a
m
ajo
r
in
s
titu
tio
n
al
ef
f
o
r
t
is
n
ee
d
ed
t
o
b
u
ild
h
ig
h
-
q
u
ality
Qu
r
an
au
d
io
d
atasets
th
at
ar
e
s
u
itab
le
f
o
r
s
cien
tific
r
esear
ch
p
u
r
p
o
s
es
an
d
s
u
itab
le
f
o
r
t
h
e
u
s
e
o
f
m
o
d
er
n
an
d
em
er
g
in
g
co
m
p
u
tin
g
tech
n
o
lo
g
ies
s
u
ch
as
tr
ain
in
g
lear
n
i
n
g
an
d
d
ee
p
lea
r
n
in
g
alg
o
r
ith
m
s
.
T
h
e
s
ize,
d
i
v
er
s
ity
,
an
d
d
if
f
er
e
n
t
m
eth
o
d
s
o
f
r
e
p
r
esen
tin
g
au
d
io
d
ata,
as
well
as
th
e
in
clu
s
io
n
o
f
d
if
f
er
e
n
t
ty
p
e
s
o
f
n
o
is
e,
s
h
o
u
ld
b
e
tak
en
in
t
o
ac
co
u
n
t
wh
en
co
n
s
tr
u
ctin
g
t
h
is
ty
p
e
o
f
d
ataset
s
o
th
at
we
ca
n
s
ay
th
at
th
i
s
d
ata
is
a
r
ep
r
esen
tativ
e
s
am
p
le
s
u
itab
le
f
o
r
ap
p
l
y
in
g
s
cien
tific
m
eth
o
d
o
l
o
g
ies
an
d
ac
h
iev
in
g
h
o
n
est r
esu
lts
with
o
u
t b
ias.
T
ab
le
5
.
C
o
m
p
a
r
is
o
n
o
f
av
aila
b
le
d
atasets
N
a
me
F
o
r
mat
N
u
mb
e
r
o
f
f
i
l
e
s
R
e
f
.
Q
u
r
a
n
.
c
o
m
a
u
d
i
o
M
P
3
1
3
6
8
[
5
8
]
QDAT
w
a
v
1
5
0
0
[
5
9
]
Q
u
r
a
n
r
e
c
i
t
a
t
i
o
n
s f
o
r
a
u
d
i
o
c
l
a
ssi
f
i
c
a
t
i
o
n
w
a
v
6
6
8
8
[
6
0
]
Th
e
t
a
r
t
e
e
l
d
a
t
a
s
e
t
M
P
3
2
5
,
0
0
0
[
6
1
]
H
o
l
y
Q
u
r
a
n
d
a
t
a
se
t
w
a
v
2
0
0
[
6
2
]
R
eg
ar
d
in
g
t
h
e
f
ea
t
u
r
e
e
x
tr
ac
ti
o
n
m
et
h
o
d
s
,
th
e
m
aj
o
r
ity
o
f
s
tu
d
ies
u
tili
ze
d
MFC
C
,
as
h
ig
h
lig
h
ted
i
n
T
ab
le
4
.
C
o
n
d
u
ctin
g
i
n
-
d
e
p
th
ex
p
er
im
en
tal
r
esear
ch
o
n
o
th
e
r
f
ea
tu
r
e
ex
tr
ac
ti
o
n
m
eth
o
d
s
,
p
ar
ticu
lar
ly
m
o
d
e
r
n
tech
n
iq
u
es
s
u
ch
as
th
o
s
e
u
s
ed
in
d
ee
p
lear
n
i
n
g
,
a
n
d
a
p
p
ly
in
g
th
em
t
o
d
iv
e
r
s
e
an
d
ex
ten
s
i
v
e
d
atasets
,
wo
u
ld
lik
ely
y
ield
m
o
r
e
ac
c
u
r
ate
r
esu
lts
.
L
ig
h
tweig
h
t
m
o
d
els
c
o
u
ld
b
e
p
r
o
p
o
s
ed
b
y
ex
clu
d
i
n
g
n
o
n
-
s
ig
n
if
ican
t
f
ea
tu
r
es a
n
d
tailo
r
e
d
m
o
d
els f
o
r
Qu
r
an
r
ec
itatio
n
co
u
ld
b
e
in
tr
o
d
u
ce
d
.
I
n
ad
d
itio
n
,
t
h
is
r
ev
iew
r
e
v
e
aled
s
ev
er
al
s
ig
n
if
ican
t
f
in
d
i
n
g
s
r
eg
ar
d
in
g
th
e
class
if
icatio
n
p
h
ase
in
tr
o
d
u
ce
d
in
v
a
r
io
u
s
liter
atu
r
e.
T
r
ad
itio
n
al
tec
h
n
iq
u
es
s
u
c
h
as
HM
M,
GM
M,
an
d
L
B
G
-
VQ
d
em
o
n
s
tr
ated
im
p
r
ess
iv
e
r
esu
lts
with
s
m
all
d
atasets
b
u
t sh
o
wed
lo
w
ac
cu
r
ac
y
with
lar
g
er
d
atasets
.
T
h
is
alig
n
s
with
f
in
d
in
g
s
f
r
o
m
p
r
ev
io
u
s
s
tu
d
ies
[
1
6
]
,
[
4
6
]
–
[
4
8
]
.
I
n
c
o
n
tr
ast,
m
ac
h
in
e
lear
n
in
g
tech
n
iq
u
es
lik
e
SVM
s
[
5
0
]
,
[
5
1
]
an
d
R
F
[
5
3
]
ac
h
iev
ed
g
o
o
d
r
esu
lts
;
h
o
wev
er
,
wh
en
th
ese
tech
n
iq
u
es
wer
e
ev
al
u
ated
alo
n
g
s
id
e
d
ee
p
lear
n
in
g
alg
o
r
ith
m
s
o
n
th
e
s
am
e
d
atas
ets,
th
eir
p
er
f
o
r
m
a
n
ce
d
eter
io
r
ated
co
m
p
a
r
ed
to
d
ee
p
lear
n
i
n
g
m
o
d
els
s
u
ch
as
lo
n
g
s
h
o
r
t
-
te
r
m
m
em
o
r
y
(
L
STM
)
[
5
4
]
.
T
h
is
o
b
s
er
v
ati
o
n
s
u
g
g
ests
th
at
d
ee
p
lear
n
i
n
g
al
g
o
r
ith
m
s
m
ay
o
f
f
er
s
u
p
er
io
r
p
er
f
o
r
m
an
ce
i
n
lar
g
e
r
,
m
o
r
e
co
m
p
lex
d
atasets
,
a
tr
en
d
n
o
ted
in
o
th
er
co
m
p
ar
ativ
e
s
tu
d
ies,
alth
o
u
g
h
s
u
ch
s
tu
d
ies
ar
e
s
till
lim
ited
.
C
o
n
d
u
ctin
g
ex
ten
s
iv
e
co
m
p
ar
ativ
e
s
tu
d
ies
u
s
in
g
em
er
g
in
g
in
ar
tific
ial
in
tellig
en
ce
,
s
u
ch
as
co
m
p
l
ex
d
ee
p
lear
n
in
g
m
o
d
els,
t
r
an
s
f
er
lear
n
in
g
,
an
d
h
y
b
r
i
d
f
r
am
ewo
r
k
s
an
d
o
p
tim
izatio
n
m
et
h
o
d
s
,
w
o
u
ld
m
ak
e
a
s
ig
n
if
ican
t
d
i
f
f
er
en
ce
i
n
r
esu
lts
b
ased
o
n
a
s
o
lid
s
cie
n
tific
m
eth
o
d
o
lo
g
y
,
as su
g
g
ested
b
y
th
e
c
u
r
r
e
n
t lim
ited
s
tu
d
ies in
th
is
f
ield
[
5
7
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
14
,
No
.
3
,
J
u
n
e
20
25
:
1
6
8
3
-
1
6
9
5
1692
Fu
r
th
er
m
o
r
e
,
th
er
e
is
s
ig
n
if
ican
t
o
v
er
lap
an
d
d
ep
e
n
d
en
c
y
b
et
wee
n
th
e
th
r
ee
is
s
u
es:
th
e
q
u
ality
o
f
th
e
d
atasets
,
th
e
m
eth
o
d
s
u
s
ed
to
ex
tr
ac
t
f
ea
tu
r
es,
a
n
d
th
e
class
if
ier
s
u
s
ed
to
id
en
tify
th
e
Qu
r
an
r
ec
iter
.
Ho
wev
er
,
th
e
m
o
s
t
im
p
o
r
tan
t
f
ac
to
r
as
a
p
r
elim
in
a
r
y
s
tep
is
to
wo
r
k
h
ar
d
to
cr
ea
te
lar
g
e
a
n
d
d
iv
e
r
s
e
d
atasets
f
o
r
a
lar
g
e
n
u
m
b
er
o
f
r
ec
iter
s
ar
o
u
n
d
th
e
w
o
r
ld
.
On
ce
a
v
ai
lab
le
to
th
e
r
esear
ch
co
m
m
u
n
ity
,
th
ese
d
atasets
wo
u
ld
allo
w
f
o
r
c
o
m
p
r
e
h
en
s
iv
e
s
tu
d
ies
an
d
th
e
g
en
er
al
izatio
n
o
f
r
esu
lts
b
ased
o
n
a
r
o
b
u
s
t
s
cien
tific
f
o
u
n
d
atio
n
.
On
th
e
o
th
er
h
an
d
,
th
e
m
eth
o
d
s
u
s
ed
f
o
r
d
ata
p
r
e
-
p
r
o
ce
s
s
in
g
ar
e
n
o
t
co
v
er
ed
i
n
th
is
s
tu
d
y
as
we
d
id
n
o
t
f
in
d
a
s
u
f
f
icien
t in
f
o
r
m
atio
n
in
p
r
ev
io
u
s
liter
atu
r
e
.
C
o
n
s
id
er
in
g
all
t
h
at
h
as
b
ee
n
p
r
esen
ted
ab
o
v
e,
as
well
as
wh
at
was
m
is
s
in
g
in
p
r
ev
i
o
u
s
s
tu
d
ies,
s
o
m
e
r
esear
ch
ch
allen
g
es a
n
d
r
ec
o
m
m
en
d
atio
n
s
ca
n
b
e
r
ea
c
h
ed
th
at
ca
n
b
e
s
u
m
m
a
r
ized
as f
o
llo
ws:
‒
A
m
ajo
r
in
s
titu
tio
n
al
e
f
f
o
r
t
is
n
ee
d
ed
t
o
b
u
ild
h
i
g
h
-
q
u
ality
Qu
r
an
au
d
io
d
atasets
.
So
m
e
k
ey
cr
iter
ia
s
h
o
u
ld
b
e
id
en
tifie
d
to
cr
ea
te
th
ese
d
a
tasets
s
u
ch
as
s
ize,
d
iv
e
r
s
ity
,
a
n
d
m
u
lti
-
m
o
d
al
d
ata,
as
well
a
s
th
e
in
clu
s
io
n
o
f
d
if
f
e
r
en
t ty
p
es o
f
n
o
is
e.
‒
C
o
n
d
u
ctin
g
e
x
ten
s
iv
e
co
m
p
a
r
ativ
e
s
tu
d
ies
u
s
in
g
e
m
er
g
in
g
tech
n
o
lo
g
ies
in
a
r
tific
ial
in
tellig
en
ce
,
d
ee
p
lear
n
in
g
m
o
d
els,
tr
an
s
f
er
lear
n
in
g
,
an
d
h
y
b
r
id
f
r
am
ewo
r
k
s
an
d
o
p
tim
izatio
n
m
eth
o
d
s
.
‒
C
o
n
d
u
ctin
g
in
-
d
ep
th
ex
p
e
r
im
en
tal
r
esear
ch
o
n
o
t
h
er
f
ea
tu
r
e
ex
tr
ac
tio
n
m
eth
o
d
s
,
p
ar
tic
u
lar
ly
m
o
d
er
n
tech
n
iq
u
es
s
u
ch
as
th
o
s
e
u
s
ed
in
d
ee
p
lear
n
in
g
(
e
n
d
-
to
-
en
d
tech
n
iq
u
es),
a
n
d
ap
p
ly
in
g
th
e
m
to
d
iv
er
s
e
an
d
ex
ten
s
iv
e
d
atasets
,
wo
u
ld
lik
ely
y
ield
m
o
r
e
ac
cu
r
ate
r
esu
lts
.
L
ig
h
tweig
h
t
m
o
d
els
co
u
l
d
b
e
p
r
o
p
o
s
ed
b
y
ex
clu
d
in
g
n
o
n
-
s
ig
n
if
ican
t f
ea
t
u
r
es a
n
d
tailo
r
e
d
m
o
d
els f
o
r
Q
u
r
an
r
ec
itatio
n
co
u
ld
b
e
in
tr
o
d
u
ce
d
.
‒
T
h
e
ef
f
icac
y
o
f
d
ee
p
l
ea
r
n
in
g
,
s
p
ec
if
ically
C
NNs,
R
NNs,
an
d
L
STM
in
v
o
ice
r
ec
o
g
n
itio
n
h
as
b
ee
n
s
u
b
s
tan
tial.
Ho
wev
er
,
th
er
e
is
a
n
ee
d
to
alg
o
r
ith
m
s
th
at
a
r
e
tailo
r
ed
to
ad
d
r
ess
th
e
u
n
iq
u
e
d
if
f
icu
lties
en
co
u
n
ter
e
d
in
Qu
r
an
ic
r
ec
itat
io
n
,
s
u
ch
as th
e
d
iv
er
s
e
r
an
g
e
o
f
s
ty
les an
d
in
to
n
atio
n
s
.
‒
F
o
c
u
s
i
n
g
o
n
d
a
t
a
p
r
e
-
p
r
o
c
e
s
s
i
n
g
m
e
t
h
o
d
s
e
s
p
e
c
i
a
ll
y
n
o
i
s
e
p
r
e
-
p
r
o
c
e
s
s
i
n
g
,
d
a
t
a
s
e
g
m
e
n
t
a
t
i
o
n
,
r
e
p
r
e
s
e
n
t
a
t
i
o
n
,
a
n
d
t
r
a
n
s
f
o
r
m
a
t
i
o
n
s
t
i
ll
a
v
e
r
y
im
p
o
r
t
a
n
t
i
s
s
u
e
t
h
a
t
c
o
u
l
d
b
e
t
ac
k
l
e
d
b
y
r
e
s
e
a
r
c
h
e
r
s
i
n
t
h
is
f
i
e
ld
.
‒
C
o
n
d
u
ctin
g
s
ev
er
al
ex
p
er
im
e
n
tal
s
tu
d
ies
u
s
in
g
all
p
r
ev
io
u
s
f
ac
to
r
s
wo
u
ld
r
esu
lt
with
in
tr
o
d
u
cin
g
a
r
o
b
u
s
t
tailo
r
ed
f
r
am
ewo
r
k
f
o
r
Qu
r
an
r
ec
iter
s
r
ec
o
g
n
itio
n
.
Fin
ally
,
we
b
eliev
e
th
at
esta
b
lis
h
in
g
th
e
co
m
p
u
tatio
n
al
r
e
s
o
u
r
ce
s
u
s
in
g
a
s
cien
tific
ap
p
r
o
ac
h
a
n
d
m
ak
in
g
th
em
av
ailab
le
f
o
r
all
r
esear
ch
er
s
’
au
d
ien
ce
s
is
th
e
f
ir
s
t
s
tep
f
o
r
an
y
s
er
io
u
s
attem
p
ts
to
im
p
r
o
v
e
th
e
p
er
f
o
r
m
an
ce
o
f
an
d
th
e
r
esu
lts
in
th
is
ar
ea
.
7.
CO
NCLU
SI
O
N
T
h
is
p
ap
er
p
r
o
v
id
es
a
c
o
m
p
r
eh
en
s
iv
e
r
ev
iew
o
f
th
e
Qu
r
a
n
r
ec
iter
r
ec
o
g
n
itio
n
s
y
s
tem
.
T
h
e
m
ai
n
m
eth
o
d
s
o
f
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
class
if
icatio
n
tech
n
iq
u
es
ar
e
d
is
cu
s
s
ed
.
B
esid
es,
A
r
e
v
iew
o
f
a
v
ailab
le
d
atasets
is
in
tr
o
d
u
ce
d
.
An
a
n
a
ly
tical
d
is
cu
s
s
io
n
is
also
p
r
o
v
id
ed
to
h
ig
h
lig
h
t
th
e
m
ain
g
ap
s
an
d
p
o
ten
tial o
p
en
is
s
u
es.
I
t
ca
n
b
e
s
aid
th
at
th
e
m
ain
co
n
clu
s
io
n
o
f
th
is
s
tu
d
y
is
th
at
to
d
ev
elo
p
a
n
ef
f
ec
tiv
e
s
y
s
tem
to
r
ec
o
g
n
iz
e
Qu
r
an
ic
r
ec
iter
s
,
it
is
im
p
er
ati
v
e
to
in
co
r
p
o
r
ate
s
o
m
e
k
ey
f
a
cto
r
s
.
T
h
ese
k
ey
f
ac
to
r
s
s
tar
t
with
th
e
u
tili
za
tio
n
o
f
h
ig
h
-
q
u
ality
d
atasets
f
o
r
b
o
th
tr
ain
in
g
a
n
d
ass
ess
in
g
r
ec
o
g
n
itio
n
m
o
d
els.
T
h
e
d
atasets
s
h
o
u
ld
en
co
m
p
ass
s
ev
er
al
d
if
f
er
e
n
t
r
ec
iter
s
,
s
ty
l
es,
an
d
ac
ce
n
ts
,
c
o
n
s
tr
u
ctin
g
s
u
ch
d
atasets
an
d
m
a
k
in
g
th
e
m
av
ailab
le
to
th
e
r
esear
ch
.
co
m
m
u
n
ity
is
p
er
h
ap
s
th
e
m
o
s
t
im
p
o
r
tan
t
p
r
io
r
ity
in
th
is
f
ield
.
Fu
r
th
er
m
o
r
e
,
f
ea
tu
r
e
ex
tr
ac
tio
n
m
eth
o
d
s
ar
e
a
v
ital k
ey
with
in
th
e
en
tire
r
ec
o
g
n
itio
n
s
y
s
tem
.
Ma
n
y
r
esear
ch
er
s
em
p
lo
y
M
FC
C
in
th
eir
s
tu
d
ies
d
u
e
to
th
e
h
ig
h
ef
f
icac
y
o
f
th
i
s
tech
n
iq
u
e
in
ca
p
tu
r
in
g
th
e
s
p
ec
tr
al
ch
ar
ac
ter
is
tics
o
f
au
d
io
s
ig
n
als,
h
o
wev
er
,
th
er
e
is
a
n
ee
d
to
co
n
d
u
ct
ex
p
er
im
en
tal
s
tu
d
ies
to
i
n
v
e
s
tig
ate
an
d
co
m
p
ar
e
th
e
d
if
f
er
en
t
m
eth
o
d
s
an
d
in
tr
o
d
u
ce
a
r
eliab
le
m
et
h
o
d
wi
th
th
e
ab
ilit
y
to
ca
p
tu
r
e
th
e
u
n
iq
u
e
ch
ar
ac
ter
is
tics
o
f
Qu
r
an
ic
r
ec
itatio
n
,
s
u
ch
as
p
itch
,
in
to
n
atio
n
,
an
d
r
h
y
th
m
.
R
eg
ar
d
in
g
t
h
e
class
if
icatio
n
tech
n
i
q
u
es,
t
h
er
e
a
r
e
a
wid
e
r
a
n
g
e
o
f
alg
o
r
ith
m
s
/tech
n
iq
u
es
a
r
e
u
til
ized
.
T
h
e
ef
f
icac
y
o
f
d
ee
p
le
ar
n
in
g
,
s
p
ec
if
ically
C
NNs,
R
NNs,
an
d
L
STM
in
v
o
ice
r
ec
o
g
n
itio
n
h
as
b
ee
n
s
u
b
s
tan
tial.
Ho
wev
er
,
th
er
e
is
a
n
ee
d
to
an
al
g
o
r
ith
m
th
at
is
tailo
r
ed
to
ad
d
r
ess
th
e
u
n
iq
u
e
d
if
f
icu
lties
en
co
u
n
ter
e
d
in
Qu
r
an
ic
r
ec
itatio
n
,
s
u
ch
as
th
e
d
iv
er
s
e
r
an
g
e
o
f
s
ty
l
es
an
d
in
to
n
atio
n
s
.
Fin
ally
,
co
n
d
u
ctin
g
s
ev
er
al
ex
p
er
im
en
tal
s
tu
d
ies
u
s
in
g
all
p
r
ev
io
u
s
f
ac
to
r
s
wo
u
ld
r
esu
lt
with
in
tr
o
d
u
cin
g
a
r
o
b
u
s
t ta
ilo
r
ed
f
r
am
ewo
r
k
f
o
r
Qu
r
an
r
ec
iter
s
r
ec
o
g
n
itio
n
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
Au
th
o
r
s
s
tate
n
o
f
u
n
d
in
g
in
v
o
lv
ed
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
id
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.