I
nte
rna
t
io
na
l J
o
urna
l o
f
Adv
a
nces in Applie
d Science
s
(
I
J
AAS)
Vo
l.
4
,
No
.
4
,
Dec
em
b
er
201
5
,
p
p
.
151
~
15
6
I
SS
N:
2252
-
8814
151
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
jo
u
r
n
a
l.c
o
m/o
n
lin
e/in
d
ex
.
p
h
p
/I
J
AAS
Speech
Recog
niti
o
n Using
MFCC
a
nd VQ
LB
G
M
.
Su
m
a
n,
K
.
H
a
rish
,
K
.
M
a
no
j
K
u
m
a
r,
S.
Sa
m
ra
j
y
a
m
El
e
c
tro
n
ics
a
n
d
Co
m
p
u
ters
De
p
t.
,
K L
Un
iv
e
rsit
y
,
V
ij
a
y
a
wa
d
a
,
In
d
ia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Sep
1
8
,
2
0
1
5
R
ev
i
s
ed
No
v
14
,
2
0
1
5
A
cc
ep
ted
No
v
2
7
,
2
0
1
5
S
p
e
a
k
e
r
Re
c
o
g
n
it
io
n
is
t
h
e
c
o
m
p
u
ti
n
g
tas
k
o
f
c
o
n
f
ir
m
a
to
ry
a
u
se
r‟s
c
lai
m
e
d
id
e
n
ti
ty
m
istre
a
t
m
e
n
t
c
h
a
ra
c
teristics
e
x
trac
ted
f
ro
m
th
e
ir
v
o
ice
s.
T
h
is
tec
h
n
iq
u
e
is
o
n
e
o
f
th
e
m
o
st
h
e
lp
f
u
l
a
n
d
i
n
sty
le
b
io
m
e
tri
c
re
c
o
g
n
it
io
n
tec
h
n
i
q
u
e
s
in
t
h
e
w
o
rld
p
a
rt
icu
larly
c
o
n
n
e
c
ted
to
a
re
a
s
i
n
th
a
t
se
c
u
rit
y
c
o
u
ld
b
e
a
m
a
jo
r
c
o
n
c
e
rn
.
It
a
re
o
f
ten
u
se
d
f
o
r
a
u
t
h
e
n
ti
c
a
ti
o
n
,
p
o
li
c
e
w
o
rk
,
rh
e
to
rica
l
sp
e
a
k
e
r
re
c
o
g
n
it
io
n
a
n
d
v
a
riety
o
f
c
o
n
n
e
c
ted
a
c
ti
v
it
ies
.
T
h
e
m
e
th
o
d
o
f
S
p
e
a
k
e
r
re
c
o
g
n
it
io
n
c
o
n
sists
o
f
tw
o
m
o
d
u
les
p
a
rti
c
u
larly
f
e
a
tu
re
e
x
trac
ti
o
n
a
n
d
h
a
v
e
m
a
tch
in
g
.
F
e
a
tu
re
e
x
trac
t
io
n
is
t
h
a
t
th
e
m
e
th
o
d
d
u
ri
n
g
w
h
ich
w
e
h
a
v
e
a
ten
d
e
n
c
y
to
e
x
tra
c
t
a
ti
n
y
lo
w
q
u
a
n
ti
ty
o
f
k
n
o
w
led
g
e
f
ro
m
th
e
v
o
ice
sig
n
a
l
th
a
t
w
il
l
late
r
b
e
u
se
d
to
re
p
re
se
n
t
e
v
e
r
y
sp
e
a
k
e
r.
F
e
a
tu
re
m
a
tch
in
g
in
v
o
lv
e
s
id
e
n
ti
f
ic
a
ti
o
n
o
f
th
e
u
n
k
n
o
w
n
sp
e
a
k
e
r
b
y
sc
ru
ti
n
y
th
e
e
x
trac
ted
o
p
ti
o
n
s
f
ro
m
h
is/h
e
r
v
o
ice
i
n
p
u
t
w
it
h
th
o
se
f
ro
m
a
c
o
ll
e
c
ti
o
n
o
f
id
e
n
ti
f
ied
sp
e
a
k
e
rs.
Ou
r
p
r
o
jec
ted
w
o
r
k
c
o
n
sists
o
f
tru
n
c
a
ti
n
g
a
re
c
o
rd
e
d
v
o
ice
sig
n
a
l,
f
ra
m
in
g
it
,
p
a
ss
in
g
it
th
ro
u
g
h
a
w
in
d
o
w
p
e
rf
o
r
m
,
c
o
n
n
iv
in
g
th
e
S
h
o
r
t
T
e
r
m
F
F
T
,
e
x
tra
c
ti
n
g
it
s
o
p
ti
o
n
s
a
n
d
M
a
tch
in
g
it
w
it
h
a
h
o
ld
o
n
g
u
id
e
.
Ce
p
stra
l
c
o
n
sta
n
t
Ca
lcu
latio
n
a
n
d
M
e
l
f
re
q
u
e
n
c
y
Ce
p
stra
l
C
o
e
ff
icie
n
ts
(M
F
CC)
a
re
a
u
n
it
a
p
p
li
e
d
f
o
r
f
e
a
tu
re
e
x
tr
a
c
ti
o
n
p
u
r
p
o
se
.
V
QL
B
G
(V
e
c
to
r
Qu
a
n
ti
z
a
ti
o
n
v
ia
Li
n
d
e
-
Bu
z
o
-
G
ra
y
)
a
l
g
o
rit
h
m
ic
ru
le
is
u
se
d
fo
r
g
e
n
e
ra
ti
n
g
g
u
id
e
a
n
d
f
e
a
tu
re
m
a
tch
in
g
p
u
r
p
o
se
.
K
ey
w
o
r
d
:
E
u
clid
ea
n
d
is
ta
n
ce
Featu
r
e
ex
t
r
ac
tio
n
Featu
r
e
m
atch
in
g
MFC
C
Sig
n
al
Vec
to
r
q
u
an
tizatio
n
Co
p
y
rig
h
t
©
201
5
In
s
t
it
u
te o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
rin
g
a
n
d
S
c
ien
c
e
.
Al
l
rig
h
ts
re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
M.
Su
m
an
,
E
lectr
o
n
ics an
d
C
o
m
p
u
ter
s
De
p
t.
,
K
L
U
n
iv
er
s
it
y
,
Vij
ay
a
w
ad
a,
I
n
d
ia
.
E
m
ail:
s
u
m
a
n
.
m
alo
j
i@
k
l
u
n
i
v
e
r
s
it
y
.
i
n
1.
I
NT
RO
D
UCT
I
O
N
Sp
ee
ch
p
r
o
ce
s
s
is
o
n
e
i
n
e
v
er
y
o
f
m
o
s
t
s
i
g
n
i
f
ica
n
t
b
r
an
ch
e
s
in
d
ig
ita
l
s
ig
n
al
p
r
o
ce
s
s
.
Sp
e
ec
h
s
ig
n
al
s
ar
e
o
f
ten
u
s
ed
f
o
r
s
p
ee
ch
r
ec
o
g
n
itio
n
,
s
p
ea
k
er
r
ec
o
g
n
itio
n
o
r
v
o
ice
co
m
m
a
n
d
r
ec
o
g
n
itio
n
s
y
s
te
m
s
.
T
h
e
task
o
f
r
ec
o
g
n
itio
n
is
to
s
ee
t
h
e
id
e
n
ti
t
y
o
f
a
s
p
ea
k
er
.
T
o
ac
k
n
o
w
led
g
e
v
o
ice,
t
h
e
v
o
ices
s
h
o
u
ld
b
e
ac
q
u
ain
ted
j
u
s
t
in
ca
s
e
o
f
p
er
s
o
n
al
ities
s
till
as
m
ac
h
i
n
e
s
.
T
h
e
s
ec
o
n
d
ele
m
e
n
t
o
f
r
ec
o
g
n
itio
n
i
s
test
in
g
,
p
ar
ticu
lar
l
y
t
h
e
tas
k
o
f
s
cr
u
ti
n
y
AN
u
n
id
en
tifie
d
au
d
it
o
r
y
co
m
m
u
n
icatio
n
to
t
h
e
co
ac
h
in
g
k
n
o
w
led
g
e
an
d
cr
ea
tin
g
th
e
id
en
ti
f
icat
io
n
.
Dep
en
d
in
g
u
p
o
n
th
e
ap
p
lian
c
e
th
e
r
ea
l
m
o
f
s
p
ea
k
er
r
ec
o
g
n
itio
n
is
s
p
lit
i
n
to
2
co
m
p
o
n
e
n
ts
.
On
e
i
s
id
en
ti
f
icatio
n
an
d
d
i
f
f
er
en
t
is
v
er
if
ica
tio
n
.
I
n
r
ec
o
g
n
itio
n
t
h
er
e
ar
e
a
u
n
it
2
s
o
r
ts
,
o
n
e
i
s
t
ex
t
d
ep
en
d
e
n
t
a
n
d
an
o
t
h
er
is
tex
t
f
r
ee
la
n
ce
.
R
ec
o
g
n
i
tio
n
i
s
s
p
lit
i
n
to
2
co
m
p
o
n
e
n
ts
:
f
ea
t
u
r
e
ex
tr
ac
t
io
n
an
d
h
av
e
class
i
f
icatio
n
.
I
n
r
ec
o
g
n
itio
n
t
h
e
s
p
ea
k
er
ar
e
o
f
ten
k
n
o
w
n
b
y
h
is
v
o
ice,
w
h
er
ev
er
j
u
s
t
i
n
ca
s
e
o
f
s
p
ea
k
er
v
er
i
f
icatio
n
t
h
e
s
p
ea
k
er
is
v
er
i
f
ied
m
is
tr
ea
t
m
e
n
t in
f
o
.
T
h
e
m
ain
p
u
r
p
o
s
e
to
g
r
asp
c
o
n
ce
r
n
i
n
g
s
p
ee
ch
is
t
h
at
th
e
s
o
u
n
d
s
g
e
n
er
ated
b
y
a
p
er
s
o
n
's
ar
ea
u
n
it
f
ilter
ed
b
y
t
h
e
f
o
r
m
o
f
th
e
v
o
c
al
tr
ac
t to
g
eth
er
w
it
h
to
n
g
u
e,
t
ee
th
etc.
T
h
is
f
o
r
m
d
eter
m
i
n
es
w
h
at
s
o
u
n
d
co
m
e
s
o
u
t.
I
f
w
e
w
ill
co
n
f
ir
m
th
e
f
o
r
m
ac
cu
r
atel
y
,
t
h
is
c
o
u
ld
p
r
o
v
i
d
e
No
r
th
Am
er
ican
n
atio
n
a
c
o
r
r
ec
t
illu
s
tr
atio
n
o
f
th
e
s
o
u
n
d
b
ein
g
cr
ea
ted
.
T
h
e
f
o
r
m
o
f
th
e
v
o
ca
l
tr
ac
t
m
a
n
i
f
est
s
its
el
f
w
it
h
i
n
th
e
e
n
v
elo
p
e
o
f
th
e
s
h
o
r
t
ti
m
e
p
o
w
er
s
p
ec
tr
u
m
,
a
n
d
th
er
e
f
o
r
e
th
e
j
o
b
o
f
MFC
C
s
i
s
to
ac
cu
r
atel
y
r
ep
r
esen
t t
h
is
e
n
v
elo
p
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8814
IJ
AA
S
Vo
l.
4
,
No
.
4
,
Dec
em
b
er
201
5
:
1
5
1
–
1
5
6
152
Fig
u
r
e
1
.
B
asic B
lo
ck
Diag
r
am
o
f
a
B
io
m
etr
ic
S
y
s
te
m
2
.
I
DE
N
T
I
F
I
CA
T
I
O
N
V
S VE
RIFICA
T
I
O
N
T
h
is
class
o
f
clas
s
i
f
icatio
n
is
t
h
at
t
h
e
m
o
s
t
s
ig
n
i
f
ica
n
t
a
m
o
n
g
t
h
e
h
ea
p
.
Au
to
m
a
tic
r
ec
o
g
n
i
tio
n
an
d
v
er
if
ica
tio
n
ar
ea
u
n
it
u
s
u
all
y
t
h
o
u
g
h
t
-
ab
o
u
t
to
b
e
th
e
m
o
s
t
n
atu
r
al
a
n
d
ec
o
n
o
m
ical
s
tr
ate
g
ies
f
o
r
av
o
id
in
g
u
n
a
u
t
h
o
r
ized
ac
ce
s
s
to
p
h
y
s
ic
al
lo
ca
tio
n
s
o
r
p
c
s
y
s
te
m
s
.
(
a)
S
p
ea
k
er
id
en
tif
icat
io
n
(
b
)
Sp
ea
k
er
Ver
if
icatio
n
Ou
r
p
ap
er
is
o
n
r
ec
o
g
n
itio
n
.
E
ac
h
th
e
f
i
g
u
r
es
r
ep
r
esen
t
t
h
e
A
SI
(
au
to
m
atic
s
p
ea
k
er
r
ec
o
g
n
itio
n
)
s
y
s
te
m
s
.
T
h
e
o
n
to
p
o
f
2
ar
ea
u
n
i
t t
h
e
b
lo
c
k
d
iag
r
a
m
s
o
f
ea
c
h
t
h
e
p
r
o
ce
s
s
e
s
w
h
er
ea
s
f
ig
u
r
e
a
p
air
o
f
r
ep
r
esen
t
th
e
s
e
n
s
ib
le
i
m
p
le
m
en
ta
tio
n
o
f
th
e
s
y
s
te
m
s
.
2
.
1
.
P
ra
ct
ica
l Ex
a
m
ples
o
f
I
n
den
t
if
ica
t
io
n
a
nd
Ver
if
ica
t
io
n S
y
s
t
e
m
Fig
u
r
e
2
.
P
r
ac
tical
ex
a
m
p
les o
f
id
en
ti
f
icat
io
n
a
n
d
v
er
if
ica
tio
n
s
y
s
te
m
s
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
AA
S
I
SS
N:
2252
-
8814
S
p
ee
ch
R
ec
o
g
n
itio
n
Usi
n
g
MFCC
a
n
d
V
QLBG
(
M.
S
u
ma
n
)
153
2
.
2
.
M
O
DULE
S
Fig
u
r
e
3
.
Mo
d
u
le
MFC
C
a
n
d
VQL
B
G
W
e
ar
e
u
s
in
g
MF
C
C
a
n
d
VQL
B
G
f
o
r
f
ea
t
u
r
e
ex
tr
ac
t
io
n
an
d
f
ea
tu
r
e
m
a
tch
i
n
g
p
u
r
p
o
s
e.
3.
SPEAK
E
R
I
D
E
NT
I
F
I
C
AT
I
O
N
T
h
e
m
ai
n
ai
m
o
f
t
h
is
p
r
o
j
ec
t
is
r
ec
o
g
n
itio
n
t
h
at
co
n
s
i
s
ts
o
f
s
cr
u
tin
y
a
s
p
ee
ch
s
i
g
n
a
l
f
r
o
m
AN
u
n
k
n
o
w
n
s
p
ea
k
er
to
an
in
f
o
o
f
n
o
tab
le
s
p
ea
k
er
.
T
h
e
s
y
s
t
e
m
w
i
ll
ac
k
n
o
w
led
g
e
t
h
e
s
p
ea
k
er
t
h
at
h
as
b
ee
n
tr
ain
ed
w
it
h
v
ar
iet
y
o
f
s
p
ea
k
er
s
.
Fig
u
r
e
h
al
f
d
o
ze
n
s
h
o
w
s
t
h
e
ele
m
e
n
tal
f
o
r
m
atio
n
o
f
r
ec
o
g
n
itio
n
an
d
v
er
if
ica
tio
n
s
y
s
te
m
s
.
W
h
er
e
v
e
r
th
e
r
ec
o
g
n
itio
n
is
t
h
at
th
e
m
e
th
o
d
o
f
cr
u
cial
th
at
r
eg
is
ter
ed
s
p
ea
k
er
p
r
o
v
id
es
a
g
iv
e
n
s
p
ee
c
h
.
On
t
h
e
o
p
p
o
s
ite
h
an
d
,
s
p
ea
k
er
v
er
if
icatio
n
i
s
th
at
t
h
e
m
et
h
o
d
o
f
r
ej
ec
tin
g
o
r
ac
ce
p
tiv
e
th
e
id
en
tit
y
clai
m
o
f
s
p
ea
k
er
.
I
n
m
an
y
ap
p
licatio
n
s
,
v
o
ice
i
s
u
s
ed
b
ec
au
s
e
th
e
k
e
y
to
s
u
b
s
ta
n
t
iate
th
e
id
en
titi
es
o
f
a
s
p
ea
k
er
ar
ea
u
n
it c
la
s
s
i
f
ied
a
s
s
p
ea
k
er
v
er
i
f
icatio
n
.
3
.
1
.
M
F
CC
(
M
el
F
re
qu
ency
Cepstr
a
l C
o
ef
f
icient
s
)
1.
Fra
m
e
t
h
e
s
i
g
n
al
i
n
to
s
h
o
r
t f
r
a
m
es.
2.
Fo
r
ea
ch
f
r
a
m
e
ca
lc
u
late
t
h
e
p
er
io
d
g
r
am
e
s
ti
m
ate
o
f
t
h
e
p
o
w
er
s
p
ec
tr
u
m
.
3.
A
p
p
l
y
t
h
e
Me
l
f
ilter
b
an
k
to
th
e
p
o
w
er
s
p
ec
tr
u
m
a
n
d
s
u
m
t
h
e
en
er
g
y
in
ea
c
h
f
i
lter
.
4.
T
ak
e
th
e
lo
g
ar
ith
m
o
f
all
f
il
ter
b
an
k
en
er
g
ies.
5.
T
ak
e
th
e
DC
T
o
f
th
e
lo
g
f
ilter
b
an
k
en
er
g
ies.
6.
Kee
p
DC
T
co
ef
f
icien
t
s
2
-
1
3
,
d
is
ca
r
d
th
e
r
est.
B
u
t
n
o
tice
t
h
at
o
n
l
y
1
2
o
f
t
h
e
2
6
DC
T
co
ef
f
icien
t
s
ar
e
k
ep
t.
T
h
is
is
b
ec
au
s
e
h
i
g
h
er
DC
T
co
ef
f
icie
n
t
s
r
ep
r
esen
t
f
a
s
t
ch
a
n
g
es
i
n
t
h
e
f
ilter
b
an
k
e
n
er
g
ie
s
an
d
it
t
u
r
n
s
o
u
t
th
e
s
e
f
ast
c
h
a
n
g
e
s
ac
t
u
all
y
d
eg
r
ad
e
A
S
R
p
er
f
o
r
m
a
n
ce
.
So
w
e
g
et
a
s
m
a
ll i
m
p
r
o
v
e
m
en
t b
y
d
eg
r
ad
in
g
t
h
e
m
.
Fig
u
r
e
4
.
P
ip
elin
e
o
f
MFC
C
Fig
u
r
e
5
.
B
lo
ck
d
iag
r
a
m
o
f
M
FC
C
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8814
IJ
AA
S
Vo
l.
4
,
No
.
4
,
Dec
em
b
er
201
5
:
1
5
1
–
1
5
6
154
T
h
e
Me
l
-
f
r
eq
u
e
n
c
y
C
ep
s
tr
u
m
co
n
s
tan
t
(
M
FC
C
)
tech
n
iq
u
e
is
u
s
u
all
y
ac
cu
s
to
m
ed
p
r
o
d
u
ce
th
e
f
i
n
g
er
p
r
in
t
o
f
t
h
e
s
o
u
n
d
f
ile
s
.
T
h
e
MFC
C
s
q
u
ar
e
m
ea
s
u
r
e
s
u
p
p
o
r
ted
th
e
w
e
ll
-
k
n
o
w
n
v
ar
i
atio
n
o
f
t
h
e
h
u
m
a
n
ea
r
‟
s
v
ital i
n
f
o
r
m
atio
n
m
ea
s
u
r
e
f
r
eq
u
e
n
cies
w
i
th
f
i
lter
s
s
p
ac
ed
lin
ea
r
l
y
a
t lo
w
f
r
eq
u
en
cie
s
an
d
lo
g
ar
it
h
m
icall
y
at
h
i
g
h
f
r
eq
u
e
n
cies
ac
c
u
s
to
m
ed
ca
p
tu
r
e
th
e
n
ec
es
s
ar
y
c
h
a
r
ac
ter
is
tics
o
f
s
p
ee
ch
.
St
u
d
ie
s
h
a
v
e
s
h
o
w
n
t
h
at
h
u
m
a
n
p
er
ce
p
tio
n
o
f
t
h
e
f
r
eq
u
en
c
y
co
n
te
n
ts
o
f
s
o
u
n
d
s
f
o
r
s
p
ee
ch
s
i
g
n
a
ls
d
o
es
n
'
t
f
o
llo
w
a
lin
ea
r
s
ca
le.
So
f
o
r
ev
er
y
to
n
e
w
i
th
AN
ac
tu
a
l
f
r
e
q
u
en
c
y
,
f
,
m
ea
s
u
r
ed
in
cp
s
,
a
s
u
b
j
ec
tiv
e
p
itch
i
s
m
ea
s
u
r
ed
o
n
a
s
ca
le
r
ef
er
r
ed
to
as
th
e
Me
l
s
ca
le.
T
h
e
Me
l
-
f
r
eq
u
en
c
y
s
ca
le
i
s
li
n
ea
r
f
r
eq
u
en
c
y
s
p
ac
in
g
b
elo
w
a
th
o
u
s
a
n
d
cp
s
an
d
a
p
o
w
er
s
p
ac
in
g
h
i
g
h
er
th
a
n
a
t
h
o
u
s
a
n
d
cp
s
.
A
s
a
p
o
in
t
o
f
r
ef
er
e
n
ce
,
th
e
p
itch
o
f
a
o
n
e
k
ilo
h
er
tz
to
n
e,
f
o
r
t
y
d
ec
ib
el
h
ig
h
er
t
h
an
t
h
e
s
e
n
s
o
r
y
ac
ti
v
it
y
h
ea
r
i
n
g
t
h
r
es
h
o
ld
,
is
o
u
tli
n
e
d
as a
th
o
u
s
a
n
d
Me
l‟
s
.
T
h
e
f
o
llo
w
i
n
g
f
o
r
m
u
la
is
em
p
lo
y
ed
to
ca
lcu
late
th
e
M
el‟
s
f
o
r
a
s
p
ec
if
ic
f
r
eq
u
en
c
y
:
Me
l
(
f
)
=
2
5
9
5
*
lo
g
1
0
(
1
+
f
/ 7
0
0
)
.
A
d
iag
r
a
m
o
f
t
h
e
MF
C
C
p
r
o
ce
s
s
es
is
s
h
o
w
n
i
n
Fi
g
u
r
e
4
.
T
h
e
s
p
ee
ch
w
av
e
f
o
r
m
is
cr
o
p
p
ed
t
o
g
et
r
id
o
f
s
ilen
ce
o
r
ac
o
u
s
tic
in
ter
f
er
e
n
ce
w
h
ic
h
w
il
l
b
e
g
if
t
w
it
h
i
n
t
h
e
s
tar
ti
n
g
o
r
f
i
n
is
h
o
f
th
e
s
o
u
n
d
f
i
le.
T
h
e
w
in
d
o
w
i
n
g
b
lo
ck
m
i
n
i
m
izes
t
h
e
d
i
s
co
n
ti
n
u
ities
o
f
t
h
e
s
ig
n
al
b
y
tap
er
in
g
t
h
e
s
tar
t
an
d
f
in
i
s
h
o
f
e
v
er
y
f
r
a
m
e
to
ze
r
o
.
T
h
e
FF
T
b
l
o
ck
co
n
v
er
ts
e
v
er
y
f
r
a
m
e
f
r
o
m
t
h
e
ti
m
e
d
o
m
ai
n
to
t
h
e
f
r
eq
u
e
n
c
y
d
o
m
ai
n
.
W
it
h
i
n
t
h
e
Me
l
-
f
r
eq
u
en
c
y
w
r
ap
p
in
g
b
lo
ck
,
th
e
s
i
g
n
al
is
p
lan
n
ed
ag
ain
s
t
t
h
e
Me
l
s
p
ec
tr
u
m
to
m
i
m
ic
h
u
m
a
n
h
ea
r
in
g
.
W
it
h
i
n
t
h
e
f
in
al
s
tep
,
t
h
e
C
ep
s
tr
u
m
,
t
h
e
Me
l
-
s
p
ec
tr
u
m
s
ca
le
is
r
eg
e
n
er
ate
b
ac
k
to
s
tr
aig
h
tf
o
r
w
ar
d
f
r
eq
u
en
c
y
s
ca
le.
T
h
is
s
p
ec
tr
u
m
p
r
o
v
id
es
a
d
e
ce
n
t
il
lu
s
tr
atio
n
o
f
th
e
s
p
ec
tr
al
p
r
o
p
er
ties
o
f
th
e
s
ig
n
al
t
h
at
is
v
ital
f
o
r
r
ep
r
esen
ti
n
g
a
n
d
r
ec
o
g
n
iz
in
g
ch
ar
ac
ter
is
tics
o
f
t
h
e
s
p
ea
k
er
.
Af
ter
t
h
e
f
i
n
g
er
p
r
in
t
is
f
o
r
m
ed
,
w
e
ar
e
g
o
in
g
to
ad
d
itio
n
all
y
s
tated
as
A
N
ac
o
u
s
tic
v
ec
to
r
.
T
h
is
v
ec
to
r
ar
e
k
ee
p
as
a
r
ef
er
en
ce
w
it
h
i
n
th
e
i
n
f
o
r
m
atio
n
.
o
n
ce
A
N
u
n
k
n
o
w
n
s
o
u
n
d
f
ile
i
s
f
o
r
eig
n
in
to
Ma
t
r
esear
ch
lab
,
a
f
in
g
er
p
r
in
t
ar
e
cr
ea
ted
o
f
it
ad
d
itio
n
ally
a
n
d
its
r
esu
lta
n
t
v
ec
to
r
ar
e
co
m
p
ar
ed
ag
ain
s
t
t
h
o
s
e
w
it
h
i
n
th
e
i
n
f
o
r
m
atio
n
,
o
n
ce
m
o
r
e
m
is
tr
ea
t
m
e
n
t
th
e
g
eo
m
etr
icia
n
d
is
tan
ce
tec
h
n
iq
u
e,
an
d
an
ac
ce
p
tab
le
m
atc
h
ar
e
d
eter
m
in
ed
.
T
h
is
m
eth
o
d
is
as st
ated
as
f
ea
t
u
r
e
m
atch
i
n
g
.
3
.
2
.
Vect
o
r
Q
ua
ntiz
a
t
io
n
A
s
p
ea
k
er
r
ec
o
g
n
i
tio
n
s
y
s
te
m
s
h
o
u
ld
ab
le
to
esti
m
ate
c
h
an
ce
d
is
tr
ib
u
tio
n
s
o
f
t
h
e
co
m
p
u
t
ed
f
ea
tu
r
e
v
ec
to
r
s
.
Sto
r
in
g
ea
c
h
s
i
n
g
le
v
ec
to
r
th
at
g
e
n
er
ate
f
r
o
m
t
h
e
co
ac
h
i
n
g
m
o
d
e
is
n
o
t
p
o
s
s
ib
le,
s
i
n
ce
t
h
ese
d
is
tr
ib
u
tio
n
s
s
q
u
ar
e
m
ea
s
u
r
e
o
u
tli
n
ed
o
v
er
a
h
ig
h
-
d
i
m
e
n
s
io
n
al
ar
ea
.
I
t‟
s
u
s
u
all
y
ea
s
ier
to
b
eg
in
b
y
q
u
a
n
tizi
n
g
ev
er
y
f
ea
t
u
r
e
v
ec
to
r
to
at
leas
t
o
n
e
o
f
a
co
m
p
ar
ati
v
el
y
t
in
y
v
ar
iet
y
o
f
m
o
d
el
v
ec
to
r
s
,
w
it
h
a
m
et
h
o
d
r
ef
er
r
e
d
to
as
v
ec
to
r
q
u
a
n
tizat
io
n
.
VQ
m
a
y
b
e
a
m
et
h
o
d
o
f
t
ak
in
g
a
n
o
v
er
s
ized
s
et
o
f
f
ea
tu
r
e
v
ec
to
r
s
an
d
m
an
u
f
ac
t
u
r
in
g
a
s
m
aller
s
et
o
f
liv
e
v
ec
to
r
s
t
h
at
r
ep
r
esen
t
s
th
e
ce
n
tr
o
id
s
o
f
th
e
d
is
tr
ib
u
tio
n
.
T
h
e
tech
n
iq
u
e
o
f
VQ
co
n
s
i
s
t
s
o
f
e
x
tr
ac
ti
n
g
litt
le
v
ar
iet
y
o
f
r
ep
r
esen
tati
v
e
f
ea
t
u
r
e
v
ec
t
o
r
s
as
A
N
ec
o
n
o
m
ical
m
ea
n
s
t
h
at
o
f
c
h
ar
ac
ter
izin
g
t
h
e
s
p
ea
k
er
s
p
ec
i
f
ic
o
p
tio
n
s
.
B
y
m
ea
n
s
t
h
at
o
f
VQ,
s
to
r
in
g
ea
ch
s
in
g
le
v
ec
to
r
th
at
w
e
h
a
v
e
a
te
n
d
en
c
y
to
g
en
er
ate
f
r
o
m
t
h
e
c
o
ac
h
in
g
is
n
o
t p
o
s
s
ib
le.
B
y
m
i
s
tr
ea
t
m
e
n
t
t
h
ese
co
ac
h
i
n
g
k
n
o
w
led
g
e
o
p
tio
n
s
s
q
u
ar
e
m
ea
s
u
r
e
clu
s
ter
ed
to
cr
ea
te
a
co
d
eb
o
o
k
f
o
r
ev
er
y
s
p
ea
k
er
.
W
ith
i
n
th
e
r
ec
o
g
n
itio
n
s
tag
e,
t
h
e
in
f
o
f
r
o
m
th
e
tes
ted
s
p
ea
k
er
is
co
m
p
ar
ed
to
th
e
co
d
eb
o
o
k
o
f
ev
er
y
s
p
ea
k
er
an
d
liv
e
t
h
e
d
is
tin
ctio
n
.
T
h
ese
v
ar
iatio
n
s
s
q
u
ar
e
m
ea
s
u
r
e
t
h
en
u
s
e
to
f
o
r
m
t
h
e
p
o
p
u
lar
it
y
ca
ll.
3
.
3
.
K
-
M
ea
ns
Alg
o
rit
h
m
T
h
e
K
-
m
ea
n
s
f
o
r
m
u
la
m
a
y
b
e
a
th
a
n
k
s
to
cl
u
s
ter
t
h
e
c
o
ac
h
in
g
v
ec
to
r
s
to
u
r
g
e
f
ea
t
u
r
e
v
ec
to
r
s
.
Du
r
in
g
t
h
is
f
o
r
m
u
la
clu
s
ter
ed
th
e
v
ec
to
r
s
s
u
p
p
o
r
ted
attr
ib
u
tes
in
to
k
p
ar
titi
o
n
s
.
I
t
u
s
e
t
h
e
k
m
ea
n
s
t
h
at
o
f
k
n
o
w
led
g
e
g
e
n
er
ated
f
r
o
m
m
ath
e
m
aticia
n
d
is
tr
ib
u
tio
n
s
t
o
clu
s
ter
t
h
e
v
ec
to
r
s
.
T
h
e
tar
g
et
o
f
th
e
k
-
m
ea
n
s
is
to
atten
u
ate
to
tal
in
tr
a
-
cl
u
s
ter
v
ar
ian
ce
,
V
.
T
h
e
p
r
o
ce
s
s
o
f
k
-
m
ea
n
s
f
o
r
m
u
la
u
s
ed
lea
s
t
-
s
q
u
ar
es
p
ar
ti
tio
n
in
g
m
e
th
o
d
o
lo
g
y
to
d
i
v
id
e
th
e
i
n
p
u
t
v
ec
to
r
s
in
to
k
in
i
tial
s
et
s
.
I
t
th
en
ca
lcu
la
tes
t
h
e
m
ea
n
p
u
r
p
o
s
e,
o
r
ce
n
ter
o
f
m
a
s
s
,
o
f
e
v
er
y
s
et.
I
t
co
n
s
tr
u
ct
s
a
b
r
an
d
n
e
w
p
ar
titi
o
n
b
y
ass
o
ci
atin
g
ev
er
y
p
u
r
p
o
s
e
w
it
h
t
h
e
h
ig
h
e
s
t
ce
n
ter
o
f
m
as
s
.
T
h
en
t
h
e
ce
n
tr
o
id
s
s
q
u
ar
e
m
ea
s
u
r
e
r
ec
alcu
la
ted
f
o
r
th
e
n
e
w
cl
u
s
ter
s
,
an
d
f
o
r
m
u
la
r
ec
u
r
r
en
t
till
o
n
ce
t
h
e
v
ec
to
r
s
n
o
w
n
o
t
s
w
itc
h
cl
u
s
ter
s
o
r
in
s
tead
ce
n
tr
o
id
s
ar
en
'
t a
n
y
lo
n
g
er
m
o
d
i
f
ied
.
3
.
4
.
E
ucli
dea
n Dista
nce
I
n
th
e
s
p
ea
k
er
r
ec
o
g
n
itio
n
s
ec
tio
n
,
AN
u
n
k
n
o
w
n
s
p
ea
k
er
‟
s
v
o
ice
is
d
iag
r
a
m
m
atic
b
y
a
s
e
q
u
en
ce
o
f
f
ea
t
u
r
e
v
ec
to
r
th
e
n
it
's
co
m
p
ar
ed
w
ith
th
e
co
d
eb
o
o
k
s
f
r
o
m
t
h
e
in
f
o
r
m
a
t
io
n
.
So
as
to
s
p
o
t
th
e
u
n
k
n
o
w
n
s
p
ea
k
er
,
th
i
s
co
u
ld
b
e
d
o
n
e
b
y
ac
tiv
it
y
th
e
d
i
s
to
r
tio
n
d
is
t
an
ce
o
f
2
v
ec
to
r
s
ets
s
u
p
p
o
r
ted
m
in
i
m
izin
g
t
h
e
E
u
clid
ea
n
d
is
ta
n
ce
.
T
h
e
E
u
clid
ea
n
d
is
tan
ce
i
s
th
a
t
th
e
"
o
r
d
in
ar
y
"
d
is
tan
ce
b
et
w
e
en
th
e
2
p
o
in
ts
t
h
at
o
n
e
w
o
u
ld
liv
e
w
i
t
h
a
r
u
ler
,
w
h
ich
m
a
y
b
e
estab
lis
h
ed
b
y
r
ec
u
r
r
en
t a
p
p
licatio
n
o
f
th
e
p
h
ilo
s
o
p
h
e
.
T
h
e
s
p
ea
k
er
w
it
h
t
h
e
lo
w
e
s
t d
i
s
to
r
tio
n
d
is
ta
n
ce
is
ch
o
s
en
to
b
e
id
en
tif
ied
as t
h
e
u
n
k
n
o
w
n
p
er
s
o
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
AA
S
I
SS
N:
2252
-
8814
S
p
ee
ch
R
ec
o
g
n
itio
n
Usi
n
g
MFCC
a
n
d
V
QLBG
(
M.
S
u
ma
n
)
155
4
.
E
XP
E
R
I
M
E
NT
A
L
RE
SUL
T
S
T
o
im
p
le
m
en
t
p
r
o
j
ec
ted
s
p
ea
k
er
r
ec
o
g
n
itio
n
s
y
s
te
m
,
a
s
y
s
te
m
w
it
h
s
o
m
e
v
o
ice
co
m
m
an
d
s
li
k
e
'
Hel
lo
'
i
s
tak
e
n
i
n
to
ac
co
u
n
t.
Fig
u
r
e
6
.
O
r
ig
in
al
s
p
ee
ch
s
i
g
n
al
Fig
u
r
e
7
.
Sil
en
ce
r
e
m
o
v
al
s
i
g
n
al
Fig
u
r
e
8
.
Fra
m
i
n
g
Fig
u
r
e
9
.
W
in
d
o
w
i
n
g
T
r
ain
in
g
p
ar
t is
co
m
p
leted
i
n
2
f
o
r
m
s
.
I
n
itial
s
y
s
te
m
w
a
s
tr
a
in
ed
w
it
h
o
n
e
r
ep
etitio
n
e
v
er
y
f
o
r
ev
er
y
}
co
m
m
a
n
d
an
d
o
n
ce
i
n
ea
ch
te
s
tin
g
s
es
s
io
n
s
.
Fig
u
r
e
10
.
Fas
t Fo
u
r
ier
T
r
an
s
f
o
r
m
Fig
u
r
e
1
1
. C
o
d
e
v
ec
to
r
s
W
ith
t
h
is
s
o
r
t
o
f
co
ac
h
i
n
g
er
r
o
r
r
ate
is
r
eg
ar
d
in
g
t
h
ir
tee
n
.
I
n
s
ec
o
n
d
k
i
n
d
,
s
p
ea
k
er
p
er
en
n
ia
l
t
h
e
w
o
r
d
s
f
o
u
r
ti
m
es
i
n
an
ex
ce
ed
in
g
l
y
s
in
g
le
co
ac
h
i
n
g
s
e
s
s
io
n
,
an
d
s
o
d
o
u
b
ly
i
n
ev
er
y
tes
tin
g
s
ess
io
n
.
B
y
d
o
in
g
th
is
n
e
g
li
g
ib
le
er
r
o
r
r
ate
in
r
ec
o
g
n
itio
n
o
f
co
m
m
a
n
d
s
is
ac
h
ie
v
ed
.
5
.
CO
NCLU
SI
O
N
T
h
e
g
o
al
o
f
th
is
p
r
o
j
ec
t
w
as
t
o
m
a
k
e
a
s
p
ea
k
er
r
ec
o
g
n
itio
n
s
y
s
te
m
,
ass
o
ciate
d
eg
r
ee
d
ap
p
ly
it
to
a
s
p
ee
ch
o
f
an
u
n
k
n
o
w
n
s
p
ea
k
er
.
B
y
i
n
v
e
s
ti
g
atio
n
t
h
e
ex
t
r
ac
ted
o
p
tio
n
s
o
f
th
e
u
n
k
n
o
w
n
s
p
ee
ch
a
n
d
s
o
co
m
p
ar
e
t
h
e
m
to
t
h
e
h
o
ld
o
n
ex
tr
ac
ted
o
p
tio
n
s
f
o
r
e
v
er
y
to
t
all
y
d
i
f
f
er
en
t
s
p
ea
k
er
s
o
as
to
s
p
o
t
th
e
u
n
k
n
o
w
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8814
IJ
AA
S
Vo
l.
4
,
No
.
4
,
Dec
em
b
er
201
5
:
1
5
1
–
1
5
6
156
s
p
ea
k
er
.
T
h
e
f
ea
tu
r
e
ex
tr
ac
ti
o
n
is
co
m
p
leted
b
y
v
i
cti
m
iza
tio
n
MFC
C
(
Me
l
Fre
q
u
en
c
y
C
o
ef
f
icie
n
ts
)
.
T
h
e
o
p
er
ate
„
m
elce
p
s
t‟
is
e
m
p
lo
y
ed
to
ca
lc
u
late
t
h
e
Me
l
ce
p
s
tr
u
m
o
f
a
s
ig
n
.
T
h
e
s
p
ea
k
er
w
as
m
o
d
elled
v
icti
m
iza
tio
n
Vec
to
r
q
u
an
tizat
io
n
(
VQ)
.
A
VQ
co
d
eb
o
o
k
is
g
en
er
ated
b
y
cl
u
m
p
t
h
e
co
ac
h
in
g
f
ea
tu
r
e
v
ec
to
r
s
o
f
ev
er
y
s
p
ea
k
er
a
n
d
s
o
h
o
ld
o
n
w
i
th
i
n
th
e
s
p
ea
k
er
in
f
o
r
m
atio
n
.
D
u
r
in
g
t
h
i
s
tec
h
n
iq
u
e,
th
e
K
m
ea
n
s
th
at
f
o
r
m
u
la
i
s
e
m
p
lo
y
ed
to
tr
y
to
to
th
e
cl
u
m
p
.
W
it
h
i
n
t
h
e
R
ec
o
g
n
i
tio
n
s
ta
g
e,
a
d
is
to
r
tio
n
li
v
e
th
at
s
u
p
p
o
r
ted
th
e
m
i
n
i
m
izi
n
g
t
h
e
g
eo
m
e
tr
ician
d
is
tan
ce
w
as
u
s
ed
o
n
ce
m
atc
h
in
g
as
s
o
ciate
d
eg
r
ee
u
n
k
n
o
w
n
s
p
ea
k
er
w
it
h
th
e
s
p
ea
k
er
in
f
o
r
m
atio
n
.
RE
F
E
R
E
NC
E
S
[1
]
M
a
h
d
i
S
h
a
n
e
h
a
n
d
A
z
izo
ll
a
h
T
a
h
e
ri,
"
V
o
ice
Co
m
m
a
n
d
Re
c
o
g
n
it
i
o
n
S
y
ste
m
Ba
s
e
d
o
n
M
F
CC
a
n
d
VQ
a
lg
o
rit
h
m
s"
,
W
o
rld
Aca
d
e
my
o
f
S
c
ien
c
e
,
En
g
in
e
e
rin
g
a
n
d
T
e
c
h
n
o
lo
g
y
,
V
o
l
.
33
,
2
0
0
9
.
[2
]
M
s.
A
ru
n
d
h
a
ti
S
.
M
e
h
e
n
d
a
le
a
n
d
M
rs.
M
.
R.
Dix
it
,
"
S
p
e
a
k
e
r
Id
e
n
ti
f
ica
ti
o
n
S
ig
n
a
ls
a
n
d
Im
a
g
e
P
ro
c
e
ss
in
g
”
,
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
(
S
IPI
J
)
,
Vo
l.
2
,
No
.
2
,
J
u
n
e
2
0
1
1
.
[3
]
Ja
m
e
l
P
rice
,
Dr.
A
li
E
y
d
g
a
h
i
,
"
De
sig
n
o
f
a
n
A
u
to
m
a
ti
c
S
p
e
e
c
h
Re
c
o
g
n
it
io
n
S
y
ste
m
Us
in
g
M
ATLAB"
,
Ch
e
sa
p
e
a
k
e
In
f
o
rm
a
ti
o
n
Ba
se
d
A
e
ro
n
a
u
ti
c
s C
o
n
so
rt
iu
m
,
A
u
g
u
st 2
0
0
5
.
[4
]
E.
Da
rre
n
.
E
ll
is
,
"
De
sig
n
o
f
a
S
p
e
a
k
e
r
R
e
c
o
g
n
it
io
n
Co
d
e
u
si
n
g
M
A
TL
A
B
"
,
De
p
a
rt
m
e
n
t
o
f
Co
m
p
u
ter
a
n
d
El
e
c
tri
c
a
l
En
g
in
e
e
rin
g
-
Un
iv
e
rsity
o
f
T
e
n
n
e
ss
e
e
,
Kn
o
x
v
il
le T
e
n
n
e
ss
e
.
[5
]
J.
S
Ch
it
o
d
e
,
A
n
u
ra
d
h
a
S
.
Nig
a
d
e
,
"
T
h
ro
a
t
M
icro
p
h
o
n
e
S
ig
n
a
ls
f
o
r
Iso
late
d
W
o
rd
Re
c
o
g
n
it
io
n
Us
in
g
L
P
C
"
,
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
Ad
v
a
n
c
e
d
Re
se
a
rc
h
in
Co
mp
u
ter
S
c
ien
c
e
a
n
d
S
o
ft
w
a
re
En
g
i
n
e
e
rin
g
,
V
o
l
.
2
,
No
.
8
,
A
u
g
u
st 2
0
1
2
.
[6
]
B.
G
o
ld
a
n
d
N.
M
o
rg
a
n
,
“
S
p
e
e
c
h
a
n
d
A
u
d
io
S
ig
n
a
l
P
ro
c
e
ss
in
g
”
,
Jo
h
n
W
il
e
y
a
n
d
S
o
n
s
,
Ne
w
Yo
rk
,
NY
,
2
0
0
0
.
[7
]
V
ib
h
a
T
iw
a
ri
,
“
M
F
CC
a
n
d
i
ts
a
p
p
li
c
a
ti
o
n
s
in
s
p
e
a
k
e
r
re
c
o
g
n
it
io
n
”
,
De
p
‟t.
o
f
E
lec
tro
n
ics
En
g
g
.
,
G
y
a
n
G
a
n
g
a
In
stit
u
te
o
f
T
e
c
h
n
o
lo
g
y
a
n
d
M
a
n
a
g
e
m
e
n
t
,
Bh
o
p
a
l.
[8
]
E.
Ka
rp
o
v
,
“
R
e
a
l
T
i
m
e
S
p
e
a
k
e
r
Id
e
n
ti
f
ica
ti
o
n
”
,
M
a
ste
r`
s
th
e
sis,
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter
S
c
ien
c
e
,
Un
iv
e
rsit
y
o
f
Jo
e
n
su
u
,
2
0
0
3
.
Evaluation Warning : The document was created with Spire.PDF for Python.