Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
10
,
No.
5
,
Octo
be
r
2020
,
pp
.
4752
~
4758
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v10
i
5
.
pp
4752
-
47
58
4752
Journ
al h
om
e
page
:
http:
//
ij
ece.i
aesc
or
e.c
om/i
nd
ex
.ph
p/IJ
ECE
Creation
of speec
h corpus
for emo
tion anal
ysis in G
ujarati
languag
e and its
evalu
ation by va
rious sp
eech para
meters
Visha
l
P. T
ank
1
,
S
.
K
.
H
ad
i
a
2
1
V T
Pat
el Departm
ent
of
Elec
tr
onic
s
and
Com
muni
cation Engi
ne
eri
ng,
Chandubh
ai
S Pa
te
l
Insti
tu
te
of
T
ec
hnolog
y
(CSP
IT)
,
Charotar
Univer
si
t
y
of
Scie
nc
e and
T
echnolog
y
(CHA
RUS
AT),
India
2
Gujar
at Techno
logi
c
al
Univ
ersity
,
Indi
a
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
N
ov
18
, 201
9
Re
vised
Ma
r
23
,
2020
Accepte
d
Apr
3
, 2
020
In
the
l
ast
coup
le
of
y
e
ars
emotion
re
cogni
t
ion
has
prove
n
it
s
signifi
c
ance
in
the
area
of
art
if
ic
i
al
in
tell
ige
nc
e
and
m
a
n
m
ac
hine
co
m
m
unic
at
ion.
Emotion
rec
ogni
ti
on
ca
n
be
done
using
spee
ch
and
image
(fa
ci
a
l
e
xpre
ss
ion),
thi
s
paper
de
al
s
with
SER
(spe
ec
h
emotion
re
c
ognit
ion)
on
l
y
.
For
emotion
rec
ogni
ti
on
emo
ti
onal
spee
ch
d
at
ab
ase
is
essen
ti
al.
In
th
is
pap
er
we
hav
e
proposed
emotio
nal
d
at
ab
ase
whi
ch
is
d
eve
lop
ed
i
n
Gujar
a
ti
la
ngu
age
,
one
of
the
offi
cial’
s
l
a
nguage
of
Ind
ia.
Th
e
proposed
spee
ch
cor
pus
b
ifurc
a
te
six
emotiona
l
st
at
e
s
as:
sadne
ss
,
surprise,
a
ng
e
r,
disgust,
f
ea
r
,
happi
n
ess.
To
observe
eff
e
c
t
of
diffe
ren
t
emotions,
anal
y
s
is
of
proposed
Gujar
ati
spee
ch
dat
ab
ase
is
c
arr
i
ed
out
using
eff
ic
i
ent
spe
ec
h
p
a
ramet
ers
l
ike
p
itch,
ene
r
g
y
and
MF
CC usin
g
MA
TL
AB Software
.
Ke
yw
or
d
s
:
Em
otion
d
et
ect
ion
from
sp
eec
h
Energy
Guja
rati
langu
age
MATLAB
soft
war
e
MFC
C
Pit
ch
Copyright
©
202
0
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Vish
al
P.
Ta
nk
,
V
T
Patel
D
e
pa
rtm
ent o
f
Ele
ct
ronics an
d
C
omm
un
ic
at
ion
En
gin
eeri
ng,
Chan
dubhai
S
Pate
l In
sti
tute
of Tech
nolo
gy
(CSPIT),
Char
otar U
nive
rsity
o
f
Scien
ce an
d
Tec
hnol
og
y
(CH
ARU
S
AT),
Chan
ga
-
3884
21,
A
nand, G
uja
rat,
I
ndia
.
Em
a
il
:
vish
al
ta
nk.ec@c
harusa
t.ac.i
n
1.
INTROD
U
CTION
Sp
eec
h
an
d
fa
ci
al
exp
ressi
on
m
ai
ny
two
m
od
e
by
wh
ic
h
people
interact
and
com
m
un
ic
at
e
to
each
oth
e
r,
bet
wixt
sp
eec
h
is
best
m
od
e
fo
r
inf
or
m
at
ion
exch
a
nge.
S
peech
is
a
com
pu
nd
sig
nal
wh
ic
h
co
ntains
the
sh
ar
p
detai
ls
of
la
ngua
ge
,
sp
ea
ker,
em
otion,
an
d
m
e
ssag
e
[
1].
It
is
i
m
po
rtance
t
o
unde
rstan
d
r
ole
of
diff
e
re
nt
em
ot
ion
s
in
s
peec
h
becau
se
pr
e
s
ecnce
of
em
ot
ion
s
m
ake
sp
e
ech
m
or
e
natu
ral.
Wo
r
d
“O
KAY”
sp
oke
n
with
di
ff
e
ren
t
em
otion
s
hav
e
dif
fere
nt
m
eaning
s
and
i
nerp
retat
ion.
H
um
an
robo
t
i
nteracti
on
can
be
po
s
sible
in
bett
er,
e
ff
ect
ive
a
nd
natur
al
way
i
f
valid
em
otion
gets
in
vo
l
ve
d
in
a
sp
e
ech
.F
inall
y
this
help
s
in
to
area
of
a
rtific
ia
l i
ntell
ie
nce.
As
m
ention
e
arli
er
em
otion
s
can
be
per
c
ei
ved
ei
the
r
f
r
om
sp
eech
or
facial
ex
pr
es
sion
(
im
ag
e
processi
ng),
but
di
gnos
ti
c
at
e
from
the
sp
ee
ch
is
c
om
plicated
ta
sk.
By
re
cogniti
on
i
ng
e
m
ot
ion
s
of
us
e
rs
a
dd
values
in
day
to
day
li
fe.
Em
ot
ion
rec
ogni
ti
on
ta
sk
is
us
efu
l
in
day
to
day
li
fe
in
sever
al
ways
li
ke,
li
e
detect
ion
syst
e
m
[2
]
,
au
dio/
vid
eo
retrie
va
l
[3
,
4],
a
rtific
ia
l
intel
li
gen
ce
an
d
r
oboti
cs,
assig
n
pr
i
or
it
y
to
custom
ers
in
va
rio
us
cal
l
-
cen
te
rs,
im
pr
ov
e
d
diag
no
sti
c
to
ol,
intel
li
gen
t
te
aching
/t
uto
ri
ng
syst
em
,
la
ng
ua
ge
conve
rsion,
im
pro
ved
c
om
pu
te
r
gam
es,
s
m
art
car
boar
d
s
yst
e
m
and
so
rt
ing
of
voic
em
ai
l/
m
essages.
Su
c
h
util
is
at
ion
s m
a
ke
em
otion
rec
ogniti
on fro
m
sp
eec
h
as
best
researc
h
t
op
ic
i
n
the
f
ie
ld
of s
peech p
r
ocessi
ng.
To
ha
ve
a
spe
ech
databa
se
is
essenti
al
i
n
process
of
sp
eec
h
em
ot
io
n
recog
niti
on
as
sh
ow
n
in
Figure
1
[
5].
Re
searche
rs
a
nd
sci
e
ntist
s
hav
e
de
velo
pe
d
sp
eec
h
c
orpor
a
inv
ari
ou
s
la
ngua
ges
li
ke
En
glish,
Ger
m
an,
Chin
ese,
Spanis
h,
J
apan
e
se,
Ru
ssian,
S
we
dish,
a
nd
Ital
ia
n
et
c
[
6].
The
re
are
f
ew
sp
e
ech
dat
abases
avail
able
f
or
offici
al
Indian
la
nguag
es
li
ke
Hindi,
Tel
ugua
a
ndMa
ly
al
a
m
[7
]
.
As
pe
r
auth
or
pe
rc
eption,
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Creati
on o
f
spe
ech c
orpu
s
fo
r emoti
on
anal
ysi
s in Gu
j
arati
langua
ge
... (
Visha
l
P.
Tank
)
4753
sta
nd
a
rd
s
peec
h
c
our
pu
s
is
not
a
vaila
ble
in
G
uj
a
rati
l
la
ng
uag
e
(
offici
al
la
nguag
e
of
I
ndia
)
[
8,
9].
In
this
pap
e
r, we
are
i
ntr
oducin
g
t
he speec
h databse
r
ec
orded in
G
uj
a
rati
lan
gu
a
ge
for
em
otion r
ecognit
ion.
It
is
ver
y
i
m
po
rta
nt
to
know
how
well
the
sp
eec
h
c
o
r
up
s
is
pr
e
par
e
d
an
d
he
nce
it
s
analy
sis
is
pr
im
e
ta
sk
[10
]
.
To
a
naly
se
s
peech
feat
ur
es
are
ext
racted
a
s
desc
ribe
d
la
te
r.
Em
otion
s
pe
ci
fic
inf
or
m
ation
i
s
al
ways
present
at
excit
at
ion
s
ource,
vocal
tr
act
,
an
d
li
ngui
sti
c
le
vels.
I
ndividu
al
em
otion
s
hav
e
a
cl
ear
eff
ec
t
on
sp
eec
h
s
poke
n
by
hum
an
an
d
it
can
be
obse
rv
e
d
by
evluati
ong
va
rio
us
para
m
et
ers/featur
es
li
ke
MFC
C
[11,
12]
,
pitch
,
ene
r
gy
et
c.
The
m
ention
featu
res
are
extracte
d
from
pr
posed
Guja
rati
datab
ase
to
see
the
ef
fects
of
var
i
ous
e
m
ot
ion
s
a
nd
it
’s
descr
i
be
d
i
n
t
his
pap
e
r.
The
pa
per
is
m
et
ho
dical
as
fo
ll
ows:
1)
i
ntrod
uctio
n
,
2)
process
of
s
peec
h
e
m
ot
ion
rec
ogni
ti
on
(S
ER
)
,
3)
Guja
rati
la
ngua
ge
an
d
it
s
roots
,
4)
creati
on
of
e
m
otion
al
sp
ee
ch
datab
ase
f
or
G
uj
a
rati
la
nguag
e
,
5)
analy
sis
of
sp
eec
h
c
orp
us
us
i
ng
s
pe
ech
par
am
et
ers,
6)
discuss
i
on a
nd
con
cl
ud
i
ng r
e
m
ark
s,
a
nd
7) re
fer
e
nces.
2.
BASI
C
F
RME
WORK OF
SPEE
CH E
M
OTION
R
E
C
OGNIT
IO
N
(
SER)
T
h
e
p
r
o
c
e
s
s
o
f
s
p
e
e
c
h
r
e
c
o
g
n
i
t
i
o
n
f
r
om
s
p
e
e
c
h
c
a
n
b
e
u
n
d
e
r
s
t
o
o
d
a
s
s
h
o
w
n
i
n
F
i
g
u
r
e
1.
T
h
e
p
r
e
r
e
q
u
i
s
i
t
e
o
f
a
n
y
S
E
R
s
y
s
t
e
m
i
s
s
u
i
t
a
b
l
e
e
m
ot
i
o
n
a
l
s
p
e
e
c
h
d
a
t
a
b
a
s
e
.
C
o
u
p
l
e
o
f
r
e
s
e
a
r
c
h
e
r
s
h
a
v
e
d
o
n
e
r
e
v
i
e
w
f
o
r
a
v
a
i
l
a
b
l
e
s
p
e
e
c
h
d
a
t
a
b
a
s
e
f
o
r
S
E
R
i
n
v
a
r
i
o
u
s
l
a
n
g
u
a
g
e
s
l
i
ke
G
e
r
m
a
n
,
S
p
a
n
i
s
h
,
E
n
g
l
i
s
h
,
e
t
c
a
n
d
m
a
i
n
l
y
i
t
i
s
c
a
t
e
g
o
r
i
z
e
d
i
n
s
i
m
u
l
a
t
e
d
d
a
t
a
b
a
s
e
(
a
c
t
o
r
g
e
n
e
r
a
t
e
d
)
,
e
l
i
c
i
t
e
d
d
a
t
a
b
a
s
e
a
n
d
n
a
t
u
r
a
l
d
a
t
a
b
a
s
e
[
1
3
]
.
Fr
om
this
a
vaila
ble
data
base
on
e
has
t
o
e
xtr
act
featu
res
fro
m
database.
S
uitable
f
eat
ur
e
sel
ect
ion
i
s
an
im
po
rtant
ta
sk
beca
us
e
i
t
carries
inten
ded
i
nfor
m
at
i
on
a
nd
it
dec
ides
overall
e
ff
ic
ie
ncy
of
s
yst
e
m
.
Gen
e
rall
y
thre
e
kinds
of
feat
ur
es
a
re
ext
rac
te
d
from
database
1)
E
xcita
ti
on
Sour
ce
feat
ur
es
li
ke
L
P
re
sidu
al
,
glo
tt
al
excit
at
ion
sig
nal,
2)
V
ocal
trac
k
featur
es
li
ke
MFC
C,
LPCC
3)
pr
os
odic
featu
res
li
ke
pitch,
f
orm
ants
4) H
yb
rid
f
eat
u
res
[1
4,
15]
.
Var
i
ou
s
cl
assif
ie
rs
li
ke
GMM
(G
a
us
sia
n
m
ixtur
e
m
od
el
),
H
MM
(
hidde
n
Ma
rkov
m
od
el
)
are
trai
ne
d
by
extra
ct
ed
f
utures
in
it
will
decide
the
s
pecific
em
otion
.
N
or
m
al
ly
c
hoos
i
ng
of
a
pa
rtic
ular
cl
assi
fier
is
base
d
on
e
xperim
ental
resul
ts
or
thu
m
b
r
ule.
G
ene
rall
y
cl
assifi
ers
are
cat
egorized
in
li
near
cl
assifi
ers
(N
ai
ve
Ba
ye
s c
la
ssifie
r)
a
nd
nonlinea
r
cl
assi
f
ie
rs
(
GMM,
H
MM
)
[
16, 17].
3.
GUJA
R
ATI L
ANGU
AGE
A
ND ITS
ROO
TS
Twe
nty
two
of
fici
al
la
ng
ua
ge
s
are
re
ported
accor
ding
to
ei
gh
t
h
Sc
hedule
of
Indian
c
o
ns
ti
tuti
on
an
d
Guja
rati
is
par
t
of
it
an
d
m
ajo
rity
sp
oke
n
in
Guja
rat
sta
te
in
India.
As
per
li
te
ratur
e,
G
uj
a
r
at
i
is
approxim
at
el
y
700
ye
ars
old
and
or
i
gn
ia
te
s
from
ind
o
Eu
r
op
ea
n
fam
il
y
b
efor
11
00
to
1500
AD
.
G
uj
a
r
at
i
is
widely
s
poke
n
la
nguag
e
in
India
by
n
um
ber
of
native
s
pe
aker
s
,
s
poke
n
by
55.
5
m
illi
on
s
peak
e
rs
whic
h
in
her
e
ntly
about
4.5%
of
t
he
to
ta
l
In
dia
n
popula
ti
on
with
6th
ra
nk.
It
is
th
e
m
os
t
widely
spok
e
n
la
ng
ua
ge
in
t
he
w
or
ld
by
nu
m
ber
of
native
s
pea
ker
s
as
of
2007
with
a
ra
nk
of
26
[
18
]
.
G
ujarati
i
s
the
offici
al
la
ngua
ge
i
n
c
ount
ry
of
India.
G
ujarati
is
the
24
th
ra
nk
e
d
la
ng
uag
e
sp
oke
n
by
56
.4
m
illi
on
people
in
the
w
orl
d
an
d
w
hich
m
akes
0.732 %
of
t
otal world
po
pu
l
at
ion
of worl
d as o
n
Ma
rch 2
019.
T
he
locat
i
on of
Guja
rat is sh
own
in Fi
gure
2.
Accor
ding
to
ref
ere
nce
an
d
avail
able
li
te
rat
ur
es
,
ACS
(
A
m
erican
Com
m
un
it
y
Su
rv
ey
)
data
by
USA
Ce
ns
us
Bu
rea
u
repo
rted
t
hat
4.34
la
kh
of
popula
ti
on
s
pea
k
Gujarat
i
la
nguag
e
as
on
20
17.
O
ut
side
t
he
India,
Guja
rati
is also w
idely
sp
oke
n i
n
countries li
ke
U
nite
d
Stat
es, Cana
da,
Bri
ti
sh
an
d
s
poke
n
to a lesser ext
ent in
China
(
par
ti
c
ul
arly
Ho
ng
K
ong),
I
ndonesi
a,
Sin
gapor
e
,
Au
st
rali
a.Th
is
m
akes
groun
d
trut
h
to
car
ry
ou
t
researc
h w
ork i
n
G
ujarati
lan
gu
a
ge [
19
]
.
Figure
1.
Ba
sic
f
ram
ework f
or em
otion
sp
eec
h
recog
niti
on
fro
m
sp
eech
Figure
2.
Locat
ion
of
Gujar
at
Stat
e in In
dia
avail
able in
W
i
kip
e
dia
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
5
,
Oct
ob
e
r 2
020
:
47
52
-
47
58
4754
4.
CR
E
ATIO
N OF EM
OTIO
NA
L
GUJ
ARATI SPEE
CH CO
RP
US
Sp
eec
h
co
rpus
can
be
c
reated
in
var
i
ous
way
s
li
ke
act
ed,
na
tural,
in
duced
et
c
[2
0,
21]
.
I
n
this
pap
e
r
create
d
G
uj
a
rat
i speec
h
c
orpus is act
ed
m
od
e
[22
]
. T
he
pro
c
ess is de
scri
be
d belo
w.
4.1.
Reco
r
ding
me
th
od
As
s
how
n
in
Figure
3
re
cordin
g
is
pe
rfor
m
ed
us
in
g
m
ob
il
e
phon
es.
The
dista
nce
bet
wee
n
the
sp
ea
ker
a
nd
the
m
ob
il
e
phone
is
m
ai
nta
ined
ar
ound
20
centim
et
er.
To
m
ake
a
com
mo
n
platfo
rm
between
al
l
reco
rd
i
ngs
on
ly
Len
ovo
(
VI
BE
K
5
note
)
is
util
iz
ed.
Fo
r
rec
ordi
ng,
aud
i
o
rec
order
app
li
cat
ion
is
us
e
d
wh
ic
h
is
in
buil
t
in
phone
.
Re
c
ordin
g
is done w
it
h
rate o
f
sa
m
pl
ing
f
re
qu
e
ncy
is
44
100H
z
an
d
s
up
e
rio
r
qu
al
it
y
with
wav file
(
*.wav).
4.2.
Co
m
posi
tion
of pr
op
ose
d
G
uj
ar
ati
em
ot
io
na
l spe
ech
C
or
pus:
The
pro
po
se
d
database
is
re
corde
d
us
in
g
9
arti
sts
(6
m
al
e
and
3
Fe
m
al
e)
wh
o
a
re
exp
e
rtise
in
DRAM
ATI
CS
.
Th
e
rec
ordin
g was
done
i
n qui
te
sing
le
ro
om at
Anand,
Gu
ja
rat (stat
e),
In
di
a. A
ll
the
s
pea
ker
s
are
in
t
he
a
ge
gro
up
of
20
-
25
ye
ars.
For
analy
zi
ng
em
otion
s,
24
diff
e
ren
t
wor
ds
ar
e
recor
ded
wit
h
six
diff
e
re
nt
em
oti
on
s
a
s
sho
wn
Table
1
.
Each
of
the
sp
ea
kers/
arti
sts
has
to
sp
eak
t
he
24
words
in
6
em
otions.
Sp
ea
ker
s
w
e
re
well
aw
a
re a
bout the t
r
ai
ning
set
s and
wor
ds
.
Table
1
c
on
ta
i
ns
in
form
at
ion
about
rec
orde
d
24
w
ords
i
n
Gu
a
j
arati
la
ngua
ge
a
nd
it
s
sequ
e
nce
.
Selc
ti
on
of
ea
ch
w
ord
is
done
s
o
that
it
cov
e
rs
e
ntire
range
of
Guja
rat
i
phonem
e
s
and
it
s
var
i
abili
ty
.
Each
em
otional
file
is
store
d,
num
ber
ed
a
nd
la
belle
d
i
n
the
com
pu
te
r
with
a
ppr
opriat
e
e
m
otion
al
sta
te
.
Table
2
c
on
ta
i
ns
si
x
em
otional
sta
te
s an
d i
ts seq
ue
nce
nu
m
ber, li
ke:
02
-
03
-
05.
wa
v (02 is s
pea
ker
-
2,
03 is em
otion
-
3, 05 is
w
ord
-
5)
04
-
05
-
20.
wa
v (04 is s
pea
ker
-
4,
05 is em
otion
-
5, 20 is
w
ord
-
20)
Total
1296
e
m
ot
ion
al
sp
ee
ch
sam
ples
(9
us
ers
*24
w
ords
*6
em
otion
s)
are
r
eco
rd
e
d
with
si
x
di
ff
e
ren
t
e
m
otion
al
clas
ses.
Table
1.
Rec
orded 2
4 w
ords
i
n Guja
rati
languag
e
and its
En
glish m
eaning
Serial
No
.
Gu
jarati
W
o
rd
Ap
p
rop
iate E
n
g
lish
M
eani
n
g
1
Pride
2
Tr
y
3
Proo
f
4
Ou
t
5
Go
d
6
Co
u
n
ten
an
ce
7
An
y
How
8
Ksh
triy
a
9
Rev
o
lu
tio
n
10
Ego
11
W
eapo
n
12
Un
eq
u
al
13
Garnis
h
ee Or
d
e
r
14
Co
n
tin
u
o
u
s
15
False
16
Glo
ry
17
Manif
esto
18
Moles
tatio
n
19
Fast
20
Step
21
I
m
p
o
ster
22
Co
m
p
ain
t
23
Ch
eatin
g
24
Bo
tto
m
Price
Table
2.
Six e
m
ot
ion
s i
n Guajarati
a
nd it
s
En
glish m
eaning
Serial
No
.
E
m
o
tio
n
in Gu
jara
ti
Ap
p
rop
iate E
n
g
lish
M
eani
n
g
1
Sad
n
ess
2
Su
rprise
3
An
g
er
4
Disg
u
st
5
Fear
6
Hap
p
in
ess
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Creati
on o
f
spe
ech c
orpu
s
fo
r emoti
on
anal
ysi
s in Gu
j
arati
langua
ge
... (
Visha
l
P.
Tank
)
4755
Figure
3. Re
co
rd
i
ng of s
peec
h database
u
si
ng m
ob
il
e phon
e
5.
E
X
PERI
MEN
T RES
ULTS
Sp
eec
h
sig
nal
is
a
de
nouem
e
nt
of
ti
m
e
var
yi
ng
vo
cal
tra
ct
syst
e
m
agitated
by
t
he
ti
m
e
var
yi
ng
excit
at
ion
s
our
ce
sign
al
.
Hence
forw
a
r
d
sp
e
ech
feat
ur
es
a
r
e
pr
ese
nt
in
both
vocal
tract
syst
e
m
and
exci
ta
ti
on
so
urce
cha
ract
erist
ic
s.
In
this
pap
er
cl
assifi
c
at
ion
is
ob
se
rved
by
three
di
f
fer
e
nt
par
am
eter
s
Ene
rg
y,
M
FCC
1
to
13
an
d
Pit
ch
[23].
E
ach
pa
ram
et
er
is
ev
al
uated
a
nd
s
how
n
Sp
ee
ch
pa
ram
et
er
Vs
U
sers
(Spea
ker
s
)
f
or
diff
e
re
nt
em
oti
on
in Grap
h
f
orm
and
in
div
i
dual
s
peak
e
rs
a
nd
it
s
in
div
i
du
al
value
s
a
re p
re
sented
in
t
a
ble f
orm
.
Com
plete
ev
al
uation i
s car
rie
d
in
MAT
LAB
softwa
re
par
ti
cularly
R2
015b
ver
si
on.
Ener
gy:
T
his
can
be
c
onsid
ered
as
c
ru
ci
a
l
par
am
et
er
fo
r
S
ER
(Spee
ch
em
otion
re
cogniti
on).
Norm
al
l
y
ener
gy
range
or
val
ue
is
low
f
or
th
e
sadn
es
s
(em
otion
-
1),
dis
gu
st
(em
otion
-
4),
f
ear
(em
otion
-
5)
an
d
high
ra
nge
or
value
f
or
t
he
joy
(em
otion
-
6),
a
nger
(e
m
ot
ion
-
3)
an
d
surp
rise
(em
otion
-
2).
En
er
gy
le
vel
evaluati
on
of
i
nd
i
vidual
us
er
is
show
n
in
gr
aph
a
nd
val
ues
are
pl
otted
in
norm
al
scal
e.
Durin
g
a
naly
sis
f
ram
e
siz
e
is
k
ept
with
160
sam
ples
and
norm
al
i
zed
in
the
ra
nge
of
(+1,
-
1).
Energy
of
a
s
peech
si
gn
al
c
an
be
fin
ding
ou
t
by
us
in
g
t
his equa
ti
on
:
=
∑
|
[
]
|
2
−
∞
∞
MFC
C
(
Mel
fre
qu
e
ncy
c
est
ru
m
c
oeffi
ci
ents
):
MFC
C
is
widely
use
d
feat
ur
e
f
or
em
otion
cl
assifi
cat
ion
.
MFC
C
purely
descr
i
bed
the
s
hap
e
of
vo
cal
t
rack
i
n
f
orm
of
sho
rt
pow
er
s
pectr
um
.
Evaluati
on
of
MFC
C
is
carried
ou
t
as
fo
l
lows
:
1)
div
i
de
the
sp
eech
sig
nal
into
short
f
ram
es,
2)
fo
r
e
ach
fr
am
e
fo
re
cast
the
pe
rio
dogra
m
and
est
im
at
e
the
po
wer
s
pe
ct
ru
m
,
3)
a
ff
ix
the
m
el
fi
l
te
r
ba
nk
to
the
po
w
er
spe
ct
ru
m
,
4)
su
m
the
energy
in
each
filt
er
finall
y
ta
ke
the
log
arit
hm
of
al
l
fi
lt
er
bank
energies.
ta
ke
the
discret
e
cosine
trans
form
of
th
e
log
filt
er
bank
e
nergies,
5)
keep
DCT
coe
ff
ic
ie
nts
2
-
13,
and
ab
dicat
e
re
st
,
an
d
6)
the
outp
u
t
of
the
filt
ers
f
r
om
each
fram
e
is
us
e
d
a
s
feat
ur
es
an
d
ce
nter
fr
e
quencies
of
the
filt
ers
are
us
e
d
in
Me
l
sc
al
e
by
us
in
g
t
he
e
qu
at
ion
[
24
]
.
(
)
=
2595
×
log
10
(
1
+
/
700
)
Fo
r
t
he
ap
pr
ai
s
al
of
the
Me
l
f
reque
ncy
sp
ect
ru
m
,
24
tria
ng
ular
filt
er
ba
nk
s
are
accum
ulate
d.
The
se
filt
ers
com
pu
t
e
the
s
pectr
um
ar
ound
eac
h
center
fr
e
quen
cy
with
i
ncr
ea
sing
ba
ndwi
dth
s.
I
n
t
his
ev
a
luati
on
of
MFC
C
1
to
13
feat
ur
es
is
sh
own
in
Ta
bl
e
s
3
–
5
.
For
r
efere
nce
he
re
ob
s
er
vation
of
MFC
C
-
1,
MF
CC
-
2,
and
MFC
C
-
3
are
show
n
ot
he
rw
ise
in
act
ua
l
al
l
the
par
am
et
ers
MFC
C
-
1
to
MFC
C
-
13
are
e
valuat
ed
in
sam
e
m
ann
er.
Pit
ch:
Pit
ch
is
al
so
an
oth
e
r
us
ef
ul
pa
ram
eter
w
hich
c
onve
ys
con
si
der
a
bl
e
inform
ation
for
em
otion
cl
assifi
cat
ion
.
Table
6
an
d
Figure
4
sho
w
evaluati
on
of
it
[25].
In
fi
gure
ind
ivi
du
al
use
r
wise
energy
values
cal
culat
ed
an
d
plo
tt
ed
agai
nst
resp
ect
ive
e
m
otion
s.
As
sh
ow
n
in
fig
ure
three
em
oti
on
s
a
nger,
s
urpr
ise
,
happine
ss
a
re
higher
ene
rg
y
band
em
otion
s
an
d
disgust,
s
adn
e
ss,
fear
ar
e
fall
s
un
der
lowe
r
e
nergy
ba
nd
.
Figure
cl
ea
rly
show
s
the
se
par
at
io
n
of
e
m
ot
ion
s
us
in
g
ene
rg
y
pa
ra
m
et
er
of
sp
ee
ch.
Seco
nd
i
m
po
rtant
par
am
et
er
pitch
val
ues
are
e
valuated
a
nd
it
s
values
are
repor
te
d
as
s
how
n
in
fig
ur
e
.
Her
e
the
range
of
the
pitch
valu
es
are
as
f
ollow
i
ng.Sa
ndnes
s
(21
4
-
118
H
z),
S
urpr
ise
(
208
-
15
2
Hz
),
Ange
r
(
211
-
16
5
Hz
),
Disgust
(19
4
-
140
Hz
),
Fea
r
(2
18
-
13
9
Hz),
Happine
ss
(
194
-
13
8
Hz).
Me
ntion
value
s
cl
early
distin
guis
h
e
m
otion
s.
Our
intenti
on of cal
culat
ing
M
FCC
is em
otion
classi
ficat
ion
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
5
,
Oct
ob
e
r 2
020
:
47
52
-
47
58
4756
Table
3
.
E
val
ua
ti
on
of MFC
C
-
1 val
ues for
G
uj
a
rati
sp
e
ech
database
Sp
eaker
1
Sp
eaker
2
Sp
eaker
3
Sp
eaker
4
Sp
eaker
5
Sp
eaker
6
Sp
eaker
7
Sp
eaker
8
Sp
eaker
9
Sad
n
ess
5
4
.14
6
5
5
5
.63
4
5
5
1
.55
1
9
5
0
.27
6
7
5
1
.83
4
6
5
5
.13
5
7
4
9
.07
4
5
6
2
.81
9
9
5
6
.07
8
9
Su
rprise
5
5
.53
1
4
5
9
.13
9
9
5
4
.69
4
9
5
0
.95
5
6
5
3
.62
9
7
5
3
.53
2
9
5
1
.42
2
9
6
2
.61
1
7
5
6
.09
0
3
An
g
er
5
9
.96
1
1
5
3
.94
5
1
5
3
.35
9
7
5
5
.37
2
6
5
6
.62
7
0
4
9
.62
9
3
5
3
.94
3
0
5
8
.22
1
3
5
7
.59
5
9
Disg
u
st
5
6
.21
8
1
5
2
.74
5
5
5
5
.19
1
1
5
0
.37
9
0
5
1
.70
3
5
5
1
.44
7
4
5
0
.67
2
9
5
9
.87
5
0
5
8
.83
6
7
Fear
5
7
.74
0
5
5
4
.66
1
1
5
1
.53
4
3
5
3
.12
2
8
5
5
.72
4
6
4
9
.18
6
5
5
1
.13
9
2
5
3
.03
3
7
5
8
.35
6
5
Hap
p
in
ess
6
0
.19
3
7
5
4
.61
3
1
5
9
.14
6
6
5
4
.38
8
0
5
4
.59
6
6
4
9
.10
5
5
5
1
.64
7
1
5
2
.73
5
3
5
7
.03
8
8
Table
4
.
E
val
ua
ti
on
of MFC
C
-
2 f
or Gu
j
arati
sp
eec
h databas
e
Sp
eaker
1
Sp
eaker
2
Sp
eaker
3
Sp
eaker
4
Sp
eaker
5
Sp
eaker
6
Sp
eaker
7
Sp
eaker
8
Sp
eaker
9
Sad
n
ess
-
2
.22
3
8
0
.97
0
6
-
2
.93
3
0
-
1
.85
9
8
-
0
.57
9
6
-
1
.60
8
0
-
0
.50
7
1
-
3
.98
2
8
-
4
.85
2
8
Su
rprise
0
.32
8
2
0
.98
2
4
-
0
.93
3
7
-
0
.37
3
2
-
0
.94
6
0
-
3
.26
3
5
-
1
.08
8
8
-
2
.10
7
0
-
4
.23
2
4
An
g
er
-
0
.33
5
5
-
2
.28
8
4
-
2
.38
8
9
-
1
.44
4
3
-
1
.00
1
8
-
5
.79
8
1
-
0
.59
1
3
-
3
.95
9
0
-
4
.41
2
1
Disg
u
st
0
.29
9
9
-
1
.00
2
4
-
2
.15
2
2
-
1
.14
7
2
-
0
.90
0
3
-
2
.11
0
4
-
0
.90
3
0
-
2
.66
2
5
-
5
.95
7
2
Fear
1
.70
6
6
-
3
.89
3
5
-
3
.07
0
8
-
2
.27
1
3
-
0
.63
9
3
-
3
.42
0
3
-
2
.91
6
9
-
4
.34
1
3
-
5
.03
4
5
Hap
p
in
ess
0
.43
1
1
0
.07
4
1
-
2
.67
3
1
-
2
.23
0
2
-
1
.21
0
4
-
4
.44
2
7
-
2
.61
0
7
-
4
.11
2
4
-
5
.87
4
7
Table
5
.
E
val
ua
ti
on
of MFC
C
-
3 f
or Gu
j
arati
sp
eec
h databas
e
Sp
eaker
1
Sp
eaker
2
Sp
eaker
3
Sp
eaker
4
Sp
eaker
5
Sp
eaker
6
Sp
eaker
7
Sp
eaker
8
Sp
eaker
9
Sad
n
ess
0
.19
3
3
1
.08
2
0
-
0
.56
5
6
-
0
.27
3
4
-
0
.14
9
5
0
.35
6
6
1
.69
0
3
-
0
.43
1
3
0
.61
7
2
Su
rprise
-
0
.51
1
4
0
.96
4
9
0
.05
3
7
-
1
.49
3
0
0
.31
2
2
2
.82
2
9
1
.68
4
7
-
2
.26
2
8
0
.08
4
9
An
g
er
-
2
.24
0
6
2
.01
0
9
-
0
.58
7
4
1
.79
5
7
-
0
.59
9
9
1
.19
6
7
2
.35
8
5
1
.47
7
5
1
.68
4
0
Disg
u
st
0
.16
4
3
0
.60
1
5
0
.48
8
7
0
.90
4
6
0
.42
9
3
3
.24
0
5
2
.94
3
0
-
1
.94
9
1
3
.09
9
1
Fear
-
0
.14
4
0
2
.84
5
7
-
1
.13
2
8
2
.53
3
6
-
0
.72
8
0
2
.18
3
8
0
.99
2
0
-
0
.63
3
9
1
.85
7
2
Hap
p
in
ess
2
.62
6
1
0
.96
2
5
2
.02
3
1
0
.08
7
8
2
.44
7
1
2
.42
4
8
2
.59
3
0
1
.50
6
7
2
.53
1
7
Table
6
.
E
val
ua
ti
on
of
pitch
val
ues
f
or
Gujarat
i speech
d
at
a
base
Sp
eaker
1
Sp
eaker
2
Sp
eaker
3
Sp
eaker
4
Sp
eaker
5
Sp
eaker
6
Sp
eaker
7
Sp
eaker
8
Sp
eaker
9
Sad
n
ess
2
1
0
.0982
2
1
4
.0308
1
8
4
.1606
1
6
5
.6608
1
7
3
.5516
1
5
7
.7194
1
1
8
.7337
1
3
5
.0328
2
1
0
.1995
Su
rprise
1
8
3
.9260
2
0
2
.9725
2
0
8
.1153
1
9
9
.6098
1
8
7
.5433
1
7
0
.0165
1
5
2
.1450
1
6
3
.7575
2
0
8
.4589
An
g
er
1
6
8
.5054
2
0
4
.2806
1
8
9
.3183
2
0
7
.5600
2
0
4
.7712
1
9
9
.5164
1
6
5
.5100
1
9
3
.2691
2
1
1
.7395
Disg
u
st
1
8
8
.1440
1
9
1
.2666
1
7
4
.0789
1
9
0
.5393
1
4
0
.9156
1
7
2
.7517
1
2
9
.8313
1
6
0
.4026
1
9
4
.7066
Fear
1
7
6
.5592
1
6
8
.3506
1
6
5
.2658
1
8
1
.8807
1
9
2
.2072
1
7
9
.9400
1
7
6
.5990
1
3
9
.3690
2
1
8
.7620
Hap
p
in
ess
1
6
7
.4267
1
9
1
.5624
1
9
4
.9376
1
9
1
.9864
1
7
6
.6201
1
6
0
.8107
1
3
8
.8605
1
5
0
.5456
1
8
3
.5233
Figure
4. A
verage
e
nergy
values fo
r Guja
rati
sp
eec
h databa
se for
em
otion
s
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Creati
on o
f
spe
ech c
orpu
s
fo
r emoti
on
anal
ysi
s in Gu
j
arati
langua
ge
... (
Visha
l
P.
Tank
)
4757
6.
CONCL
US
I
O
N
In
t
his
pap
e
r
we
hav
e
co
nte
m
pla
te
d
an
e
m
ot
ion
al
s
pee
ch
database
or
sp
eec
h
c
orp
us
in
Guja
rati
la
nguag
e
.
T
he
six
basic
em
otion
s
delibe
rate
d
for
de
velo
pin
g
data
base
a
r
e
sad
ness
,
s
urpr
ise
,
an
ge
r,
di
sg
us
t,
fear
a
nd
ha
pp
i
ness.
E
valuati
on
of
sp
ee
ch
database
is
ca
r
ried
by
m
ai
nly
pa
ram
et
ers
as
ener
gy
,
MFC
C
1
to
13
,
pitch
.
Re
s
ults
cl
early
hav
e
s
how
n
the
di
fference
in
dif
f
eren
t
em
otion
s
.
But
sti
ll
dat
abase
can
be
furthe
r
i
m
pr
oved
a
nd
var
ia
bili
ty
in
sp
eake
rs
a
nd
spok
e
n
wor
ds
m
akes
it
m
os
t
eff
ect
ive.
The
pro
posed
data
ba
se
is
a
n
inte
rm
ixtur
e of c
har
act
erist
ic
s in
te
rm
s o
f dif
fe
re
nt em
oti
on
s
, spea
ker
s
a
nd words
.
Linear
m
od
el
/c
l
assifi
er
an
d
Non
li
nea
r
m
od
el
s/
cl
assifier
s
can
be
e
xplo
red
t
o
f
ur
t
her
im
pr
ove
the
recog
niti
on
pe
rfor
m
ance.
The
im
po
rtance
of
sp
eec
h
e
m
otion
over
i
m
age
is
per
so
n
ca
n
cha
nge
facial
expressi
on
s
ea
sil
y
bu
t
h
a
rd
to
c
hange
s
pee
ch.
In
f
uture,
m
ul
tim
od
al
detect
ion
syst
em
s
can
be
bu
il
t
up
w
hic
h
us
e
d
im
age,
spe
ech si
gn
al
s a
nd
bio
-
sig
nals al
l t
o
gat
her f
or
cl
assifi
cat
ion
of em
otion
stat
e
s of
hu
m
an.
REFERE
NCE
S
[1]
R.
Chakr
abor
t
y
,
M.
Pandhar
ip
a
nde,
and
S.
K.
Koppar
apu,
“
Know
le
dge
b
ase
d
f
ramework
for
in
te
lligent
emotio
n
rec
ogni
ti
on
in
sp
onta
neous
spe
ech
,”
Proce
d
ia
Co
mputer
Scienc
e
,
v
ol
96,
pp.
587
-
596,
2016
.
[2]
K.
S
.
R
ao, Shashidha
r
G.
and
Ko
ola
gudi
,
“
Emoti
on
Rec
ogn
it
ion
using Spee
ch
Fe
at
ure
s
,”
Spring
e
r
,
2013.
[3]
G.
Gaura
v,
et
al
.
,
“
Deve
lopment
of
appl
i
ca
t
ion
spec
if
ic
con
ti
nuou
s
spee
ch
rec
ogn
i
ti
on
s
y
st
em
in
Hindi
,
”
Journal
of
Signal
and
Infor
mation
Proc
essing
,
vol
.
3
,
no
.
3
,
2012.
[4]
F
.
H
.
Rac
hm
an,
R
.
Sarno,
C
.
Fati
ch
ah,
“
Mus
ic
Emotion
Cla
ss
ifi
cation
base
d
on
L
y
ri
cs
-
Audio
using
Corpus
base
d
Emotion
,
”
Int
ernati
onal
Journal
of
E
l
ec
tri
cal
and
C
omputer
Engi
n
ee
ring
(
IJE
CE)
,
v
ol.
8,
n
o
.
3
,
pp.
1720
-
1730
,
2018.
[5]
D.
Prave
na
and
D.
Govind,
“
Deve
lopment
of
sim
ula
te
d
emotio
n
spee
ch
dat
ab
a
se
for
exc
it
a
ti
on
source
ana
l
y
sis
,
”
Inte
rnational
Jo
urnal
of
Sp
eech T
ec
hnology
,
vo
l.
20,
pp.
327
–
338
,
2017
.
[6]
M
.
Mus
faf
a,
M
.
Yus
uf,
and
M
.
Male
kz
ade
h
,
“
S
pee
ch
emotion
r
ec
ogni
ti
on
rese
a
rch
:
an
an
aly
sis
of
rese
arc
h
foc
u
s
,
”
Inte
rnational
Jo
urnal
of
spe
ec
h
Technol
ogy
(
IJS
T)
,
v
ol.
21,
pp.
1
37
-
156,
2018
.
[7]
F
.
Shah,
”
Stud
y
and
an
aly
sis
of
s
pee
ch
emotion
r
ec
ogni
ti
on”
Ph
.
D.
Thesis
,
Sodh
ganga
,
2016.
[O
nli
ne]
.
Avail
ab
le:
htt
p://s
hodhgang
a.
inf
li
bne
t.ac.in/handle/10603/12
2185
.
[8]
V
.
Ta
nk
and
S
.
K
.
Hadia
,
“
Deve
lopment
of
E
m
oti
on
Rec
ogni
ti
on
From
A
Sp
ee
ch
In
Vari
ous
Regi
onal
Indi
a
n
La
nguag
es:
A R
evi
ew
,
”
I
JRECE
,
vol
.
6
,
no
.
2
,
pp
.
2155
-
2161
,
20
18
.
[9]
R.
Kum
ar,
et
al
.,
“
Deve
lopment
of
India
n
l
ang
uage
spee
ch
databa
ses
for
la
rg
e
voca
bu
l
ar
y
spe
ec
h
re
cogni
t
ion
s
y
stems
,
” i
n
In
ter
nati
onal
Conf
ere
nce on
Sp
eech and Compute
r (
SPE
COM)
Proce
edi
ngs
,
2005.
[10]
M
.
Sw
ai
n,
A
.
Routray
,
P.
Ka
bisat
pat
h
y
,
“
Data
base
s,
fe
at
ure
s
and
cl
assifie
rs
for
spee
ch
emotion
rec
ogn
it
io
n:
a
rev
ie
w
,
”
In
te
rn
ati
onal
Journal of
Spe
ec
h
Techn
ology
,
v
ol
.
21
,
p
p.
93
-
120
,
2018
.
[11]
V
.
P
.
Gow
da,
M
.
Murugave
lu
,
S
.
K
.
Tha
nga
m
uthu,
“
Conti
n
uous
kanna
da
spee
ch
segm
ent
a
ti
on
and
spee
ch
rec
ogni
ti
on
b
ase
d
on
thr
eshol
d
using
MF
CC
and
VQ
,
”
Int
ernati
onal
Journal
of
El
e
ct
ri
c
al
and
Comput
er
Engi
ne
ering
(
IJ
ECE
)
, v
ol
.
9
, n
o
.
6,
pp.
4684
-
469
5,
2019
.
[12]
A.
P.
Kum
ar
,
et
al
.
,
“
Conti
nuous
Te
lugu
Speec
h
Rec
ognition
thro
ugh
Com
bine
d
F
ea
tur
e
Ext
r
ac
t
io
n
b
y
MF
CC
and
DW
P
D
Us
ing
HM
M
base
d
DN
N
Te
chn
ique
s,
”
Inte
rnational
Jo
urnal
of
Pure
a
nd
Applie
d
Ma
t
hemati
cs
,
v
ol
11
4,
n
o.
11
,
pp
.
187
–
197,
2017
.
[13]
S.
G.
Koolagudi
and
K
.
Rao,
“
Emotion
rec
og
nit
ion
from
spee
ch:
a
r
evi
ew
,
”
Inte
rnational
j
ournal
of
spee
c
h
Technol
ogy
,
v
o
l. 15, pp. 99
-
117,
2012
.
[14]
B
.
Bashar
ir
ad
and
M
.
Moradha
seli,
“
Speec
h
emotion
rec
og
nit
ion
m
et
hods:
A
rev
ie
w
,”
in
AIP
Confe
ren
ce
Proce
ddings
,
pp.
1
-
7
,
2017
.
[15]
S.
K.
Koolagud
i,
e
t
al
.
,
“
Re
co
gnit
ion
of
emotions
from
spee
c
h
using
exc
i
tati
on
source
fe
at
u
res
,”
in
El
se
vie
r
proce
edi
ngs
for
int
ernati
ona
l
co
nfe
renc
e
on
modeli
ng
,
opt
imizati
on
and
comput
in
g
,
pp
.
3409
-
341
7
,
2012
.
[16]
S.
Ramakrishna
n,
“
Rec
ognition
of
Emotion
fro
m
Speec
h:
A
Revi
ew
,
”
in
Sp
eech
Enhancemen
t,
Mode
li
ng
and
Re
cogn
it
ion
–
Algor
it
hms
and
Ap
pli
cations
,
In
Tec
h
,
pp
.
121
-
138
,
2012
.
[17]
S.
S.
Poo
rna
,
K.
Anuraj
.
,
and
G.
J.
Nair
,
“
A
W
ei
ght
Based
Approac
h
for
E
m
oti
on
Rec
ogni
ti
on
from
Spee
c
h:
An
Anal
y
sis
Us
ing
South
India
n
L
angua
ges
.
Comm
unic
ati
o
ns
in
Computer
and
Information
Scienc
e
So
ft
Computing
Syst
e
m
,
vol
.
837
,
201
8
[18]
“
Hindi
is
the
m
ost
spoken
India
n
la
nguag
e
in
US
A,
foll
owed
b
y
Gujar
ati
and
T
elugu
,”
The
Sen
ti
n
el
of
th
is
land,
fo
r
it
s
people
,
2018.
[Online
]
.
Avai
l
abl
e
:
ww
w.sent
i
nel
assam
.
com/n
ews/hindi
-
is
-
th
e
-
m
ost
-
spoken
-
indi
an
-
l
angua
ge
-
in
-
usa
-
foll
owed
-
by
-
guja
ra
ti
-
and
-
tel
egu/
[19]
“
Li
st
of
la
nguage
s
b
y
num
ber
of
nat
ive
spea
ker
s
,”
Wi
ki
p
ed
ia
,
the
free
en
cy
c
lope
dia
.
[Online
]
.
Avail
ab
le
:
en.
wikip
edi
a
.
org
/wiki
/List_of_languages_b
y
_nu
m
ber
_of_na
ti
ve
_spea
ker
s
.
[20]
S.
S.
Agrawal
,
et
al
.
,
“
Emotion
s
in
Hindi
Spee
ch
-
Anal
y
sis,
Per
ce
pt
ion
and
R
ecogniti
on
,”
in
20
11
Inte
rnationa
l
Confe
renc
e
on
S
pee
ch
Database
and
Assess
ments
(
Or
ie
ntal
COCO
SDA
,
pp.
7
-
13
,
2011
.
[21]
S.
G.
Koolagudi,
et
al
.
,
“
IIT
KG
P
-
SES
C:
Speec
h
Data
bas
e
for
E
m
oti
on
Anal
y
s
is
,”
Comm
unic
a
tions
in
Compute
r
and
Information Sci
en
ce
,
pp
.
485
-
492,
2009
.
[22]
S.
G.
Koolagudi,
et
al
.
,
“
IIT
KG
P
-
SEHS
C:
Hind
i
Speec
h
Corpus
for
Emotion
A
naly
s
is
,”
in
IE
E
E
proce
edi
ngs
o
n
Inte
rnational
Co
nfe
renc
e
on
Dev
ic
es
and
Comm
unic
ati
ons
(
ICDeCom)
,
pp.
1
-
5
,
2
011
.
[23]
A.
Jain,
N.
Prak
ash,
S.
S.
Agraw
al
,
“
Ev
al
ua
ti
on
o
f
MF
CC
for
Em
oti
on
Ide
n
ti
fi
ca
t
i
on
in
Hindi
Spe
ec
h
,
”
2
011
IE
EE
3rd Int
ernati
ona
l
Conf
ere
nce on Com
municat
ion Software
and
Ne
tworks
,
pp.
189
-
193
,
2011
.
[24]
E.
S.
Gopi,
“
Digit
al speech
pro
ces
sing u
sing MATL
AB
,”
Springe
r
,
2013
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
5
,
Oct
ob
e
r 2
020
:
47
52
-
47
58
4758
[25]
S.
S.
Poorna,
et
al
.
,
“
Emotio
n
rec
ogni
ti
on
using
m
ult
i
-
par
amete
r
spe
ec
h
fea
tur
e
c
la
ss
ifi
c
at
ion
,
”
i
n
IE
E
E
Inte
rnational
Co
nfe
renc
e
on
Co
mputers,
Comm
unic
ati
ons
,
and
Syste
ms
,
Ind
ia
,
2
015
.
BIOGR
AP
H
I
ES
OF
A
UTH
ORS
Mr
.
V
P
Tan
k
pursed
Bac
h
el
or
of
Eng
ineeri
ng
from
Dharm
sinh
Desai
Univ
ers
ity
,
Nadi
ad
in
2009
and
Maste
r
of
Engi
ne
eri
n
g
from
Gujar
at
Te
chno
logi
c
al
Univer
sit
y
in
y
e
ar
2011.
He
is
cur
ren
t
l
y
pursui
ng
Ph.D.
and
working
as
As
s
ista
nt
Profess
or
in
V
T
Pat
el
Depa
rtment
of
El
e
ct
roni
cs
&
Com
m
unic
at
ion
Engi
nee
r
ing,
Char
ota
r
Univer
sit
y
o
f
Scie
nce
&
Tec
hnolog
y
,
since
2012.
He
is
a
l
i
fe
ti
m
e
m
ember
of
IET
E
sin
ce
2014.
He
has
p
ubli
shed
m
ore
t
han
10
rese
ar
ch
pape
rs
in
rep
u
ted
int
ern
at
ion
al
j
ourna
ls
and
conf
ere
nc
es
and
i
t’s
al
so
availa
b
l
e
o
nli
ne
.
His
m
ai
n
rese
arc
h
work
f
ocuse
s
on
Digital
Spee
ch
pro
ces
sing,
Bioe
l
ectr
onic
s,
Digi
tal
si
gnal
pro
ce
ss
ing.
He
has
8
y
e
ars
o
f
teac
hing
exp
erienc
e
and
1
y
e
ars
of
Rese
arc
h
Ex
per
ie
n
ce.
Dr.
S
K
Hadia
is
an
assoc
ia
t
e
profe
ss
or
at
Gujar
at
T
echnological
Uni
ver
sit
y
(GTU),
Ahem
daba
d,
Gujar
a
t,
India.
He
has
complet
ed
his
Ph.D.
from
C
har
ota
r
Univer
si
t
y
of
Scie
n
c
e
and
Te
chno
log
y
in
y
ea
r
2016.
His
m
ai
n
rese
ar
ch
work
foc
uses
on
Optic
al
co
m
m
unic
at
ion
computer
net
wo
rk,
Im
age
proc
es
sing.
He
has
m
ore
tha
n
15
y
e
ars
of
te
a
chi
ng
exp
e
rie
nc
e.
He
has
publi
shed
m
ore
tha
n
11
rese
ar
ch
pape
rs
in
re
pute
d
high
impact
in
te
rna
ti
on
al
journa
ls
and
conf
ere
n
ce
s.
Evaluation Warning : The document was created with Spire.PDF for Python.