TELKOM
NIKA
, Vol. 11, No. 5, May 2013, pp. 2641 ~ 2647
ISSN: 2302-4
046
2641
Re
cei
v
ed
Jan
uary 17, 201
3
;
Revi
sed Ma
rch 1
5
, 2013;
Acce
pted Ma
rch 2
5
, 2013
Voice Collection under Different Spectrum
Min Li*, Yu-duo Wang
Schoo
l of Information a
nd C
o
mmunicati
on E
ngi
neer
in
g
Beiji
ng Informa
tion Scie
nce a
nd T
e
chnol
og
y Universit
y
, Bei
jing
100
10
1, Chin
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: rongb
ing-
limi
n
@1
63.com
A
b
st
r
a
ct
Accordi
ng to
th
e sh
ort-time
F
o
urier tra
n
sfor
m theor
y
an
d pr
i
n
cipl
e
of di
gital
filterin
g, this
p
aper
establ
ishe
d a math
e
m
atic
al mo
de
l
ca
lle
d c
o
llecti
o
n
of v
o
i
c
e si
gna
l c
o
ll
e
c
tion
at d
i
ffere
nt spectru
m
. T
h
e
voice si
gna
l w
a
s a non-stati
onary pr
ocess
,
w
h
ile t
he stand
ard F
ouri
e
r transform on
ly app
lie
d to th
e
peri
odic s
i
gn
al,
transie
nt sign
als or
stati
ona
ry rand
om s
i
g
nal. T
her
ef
ore,
the stand
ard
F
ourier tra
n
sfo
r
m
coul
d not b
e
di
rectly used for
the
speec
h si
gna
l. By controlli
ng th
e inp
u
t
different types
and p
a
ra
mete
rs,
this p
aper
an
al
y
z
e
d
th
e co
llec
t
ed ori
g
i
nal
voi
c
e sig
n
a
l
sp
ectrum w
i
th the
us
e of MAT
L
AB s
o
ftw
are platfor
m
.
At the sa
me ti
me, it re
al
i
z
e
d
the extractio
n
,
record
i
ng a
n
d
pl
ayback
of
the spe
e
ch s
i
gn
al at d
i
ffere
nt
freque
ncies. T
herefor
e, the
w
a
veforms c
o
uld
be
dis
p
l
a
y
ed
obvi
ously
o
n
the
gra
p
h
i
c
user i
n
terface
and
voice effect co
uld b
e
more cl
early. Mea
n
w
h
ile, the re
su
lt w
a
s verified b
y
the hardw
ar
e platfor
m
s, whic
h
consiste
d of T
M
S320VC
5
5
09A [1] chi
p
and T
L
V
320
AIC23 vo
ice ch
i
p
. T
he results
show
ed that
th
e
extraction of vo
ice sig
nal u
n
d
e
r
different spec
tr
um
mod
e
l w
a
s scientific, rati
ona
l an
d effective.
Ke
y
w
ords
: DS
P, speech sig
n
a
l, Graphic Us
er
Interface, di
gital filter
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
As the hot rese
arch topi
cs in the field
of high-te
ch
application, voice si
gnal
has a
more
clo
s
ely
relation
ship
with many fields. Such as telep
hon
e
s
in the are
a
s of indu
strial
prod
uctio
n
, automatic
di
al of teleco
mmuni
cati
on
system
s, spee
ch recog
n
ition [2], sp
eech
con
s
ultatio
n
and m
ana
ge
ment, health
care a
nd a
ssisted q
u
e
r
y in the a
r
ea
s
of living . Sp
eech
sign
al processing
ca
n be
more
efficient
to pr
od
uce, tran
spo
r
t, store and
get sp
eech me
ssag
e,
whi
c
h ha
s a g
r
eat sig
n
ifica
n
t for prom
oting so
cial d
e
velopme
n
t.
Digital sig
nal
processo
r (DSP) is spe
c
ia
lly desi
g
n
ed for digital
signal p
r
ocessing
algorith
m
of
real-time
an
d rapid im
pleme
n
tation [3].
It involves m
any
su
bje
c
ts
and
is
wid
e
ly use
d
in many are
a
s
of emerging
disci
pline
s
. It has mo
re ad
vantage
s tha
n
simulatio
n
system, such
as
predi
ctability, prog
ram
m
ab
le, high p
r
e
c
i
s
ion, g
ood
st
ability, reliabil
i
ty and rep
e
a
t
ability, easy to
reali
z
e the a
daptive algo
ri
thms, large-scale in
te
gration etc. Thi
s
pape
r u
s
e
s
T
M
S320VC55
09A
chip
lau
n
che
d
by TI, thi
s
chip
ha
s the
cha
r
a
c
t
e
risti
c
s of
e
fficiency, p
o
rtable a
nd l
o
w
con
s
um
ption.
It also ad
opt
s the u
n
iform
addressin
g
ways to pa
rtition the sto
r
ag
e
spa
c
e,
whi
c
h
is
conve
n
ient to
a prog
ram o
p
timization, a
s
well a
s
the
reali
z
ation d
a
ta pro
c
e
s
sin
g
. It provides a
c
onvenienc
e
for the
s
p
eech s
i
gnal process
i
ng
by us
ing the IIC, Mc
BSP [4
] or
RTC peripheral
interface.
The d
e
si
gn
of the
spe
e
c
h
sig
nal
sy
stem inte
rfa
c
e is ba
se
d
on MATLAB
GUI
environ
ment.
It design
s
a
digital sig
nal
filter wi
th the
cha
r
a
c
teri
stics of h
andli
n
g low fre
que
ncy
sign
als, n
o
drift and the
ideal fre
q
u
ency respon
se, an
d fina
lly complete
spe
e
ch si
g
nal
acq
u
isitio
n, extraction a
n
d
pl
ayba
ck.
Users ca
n d
i
rectly loadi
n
g
and listen
to the spe
e
ch
resou
r
ce file
s. By cli
cki
n
g
on th
e
correspon
di
ng
b
u
tton, we
ca
n co
mplete t
he data
re
su
lts
stora
ge an
d waveform display after the extraction.
2. Rese
arch
Metho
d
2.1. Short-time Fourier T
r
ansform the
o
r
y
Standard Fo
urie
r tran
sform only appli
e
s to tr
an
si
ent
sign
als, perio
dic sign
al
or
stationa
ry ra
ndom
sign
al. The
spe
e
ch sig
nal b
e
l
ong
s to no
n
-
station
a
ry p
r
ocess,
with
the
c
h
ar
ac
te
r
i
s
t
ics
o
f
time
-
v
aryin
g
.
Bu
t w
i
th
in
a s
hort
time, its
ch
aracteri
st
ics remain
relatively
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 5, May 2013 : 2641 – 264
7
2642
stable. T
here
f
ore it could
be
seen
as a qua
si
-ste
ady-state
pro
c
e
ss [5]. Sh
ort-time F
o
u
r
ier
transfo
rm
(ST
FT) is
a math
ematics tra
n
sform rel
a
t
ed t
o
the Fou
r
ie
r transfo
rm, wh
ich i
s
to sel
e
ct
a time-f
req
u
e
n
cy lo
cali
zati
on
wind
ow fu
nction. A
s
su
mes that the
wind
ow fun
c
tion g
(
t) is sm
ooth
within
a sho
r
t time inte
rval
(pseud
o
smo
o
th), mova
ble
win
dow fun
c
tion, ma
ke
s f
(t)g(t) a
smo
o
th
sign
al
within
different time,
and
calculat
es
out
p
o
wer
spe
c
tru
m
at
d
i
fferent mo
m
ents. Sh
ort-ti
me
Fouri
e
r tra
n
sf
orm wi
ndo
w f
unctio
n
s
can’
t be cha
nge
d
,
once
sele
ct
ed, its sh
ape
will not chan
ge,
and the resol
u
tion is dete
r
mined a
c
cord
ingly. To cha
nge the resol
u
tion, we ne
e
d
to resele
ct th
e
wind
ow fun
c
ti
on.
2.2. Design
of FIR filter
s
The pap
er a
dapts wi
ndo
w function
s
method to
desig
n FIR filter, it can be reali
z
e
d
arithmeti
c
by
differential e
q
uation
s
1
0
)
(
)
(
N
k
k
n
x
k
h
n
y
[6]. In
the formula,
k
n
x
is the
k
sampli
ng
pe
ri
od d
e
lays of i
nput
sign
al,
N
is the
nu
mbe
r
of filter
order,
k
h
is the
k tim
e
-del
ay
weig
hted valu
e (filter co
efficient),
n
y
is the filter’s output
sign
al of
nT
t
.
3. MATL
AB
Model De
sign and Simulation
This de
sig
n
acq
u
ire
s
a .wav format’
s
audio file throug
h the co
mputer
sou
n
d card.
First,
con
d
u
c
t a
sh
ort-time
Fouri
e
r tran
sf
orm, the
n
de
sign
lo
w-p
a
ss, hig
h
-pa
ss,
band
-pa
s
s a
n
d
band
-sto
p filter for the tra
n
sform results. We
can pl
ayback the
spee
ch afte
r e
x
traction
of the
high fre
que
ncy, low frequ
e
n
cy, mediu
m
freque
ncy
th
rough th
e bro
adcast fun
c
ti
on, therefo
r
e
th
e
effec
t
after filter extr
ac
tion is
more c
l
early.
3.1. Design
of FIR Filters
The pap
er u
s
es the functio
n
1
fir
in the MATLAB signal proce
s
sing To
o
l
box to desig
n
FIR filter. The
1
fir
call
s f
o
rmat
:
window
ftype
Wn
n
fir
,
'
'
,
,
1
b
[7].
The
n
is FIR fil
t
er orde
r, whi
c
h i
s
a
even
for hig
h
-p
ass and
ban
d-p
a
s
s filter;
Wn
is
cutoff fre
que
ncy: Fo
r
ba
nd-p
a
ss an
d
ban
d-stop
filter:
2
,
1
W
W
Wn
,
1
W
and
2
W
are
respe
c
tively the lo
we
r
cut
-
off freq
uen
cy and the
u
pper cut-off f
r
equ
en
cy.
ftype
is the filter
type
,
it is a low-pa
ss o
r
b
and-pa
ss
filte
r
without emp
hasi
z
e
d
. The
igh'
'
h
is high-pa
ss
filter and
the
'
'
stop
is band
-stop filter;
window
is the type of window fun
c
tion.
3.2. Main Functions
for
MATL
AB In
terfa
ce Desig
n
1)
[y, FS, bits] = wavre
ad ('file
name'
)
‘wavre
ad’
su
pport
s
m
u
ltichann
el d
a
ta,
with u
p
to
32
bits pe
r
sa
mple, a
nd
su
pport
s
readi
ng 24
-
and 32
-bit .wav files. The
.wav ex
tension is a
ppen
ded if no extensi
on is giv
en.
Amplitude values a
r
e in the
range [-1,+1]
.
Return
s the sampl
e
rate (Fs) in
Hert
z a
nd the numb
e
r
of bits p
e
r
sa
mple (bits) u
s
ed to
en
co
d
e
the d
a
ta
in
the file. ‘y’ repre
s
e
n
ts
a
string
of data
,
in
whi
c
h
store
s
the sam
p
le v
a
lue
s
. ‘fs’ is .
w
av form
at file ’s
sampli
ng
freque
ncy
(Hz); ‘bits’ i
s
the
quantification
length in the A/D conve
r
ter, that is samp
ling bits [8].
2)
s
o
und (x, fs
, bits
)
It plays the
sou
nd usi
n
g
bits numbe
r of bi
ts/sam
ple, if possib
l
e. Most platforms
sup
port bits
= 8 or bits = 1
6
.
wavplay is a
synchro
nou
s
operation, M
A
TLAB
will stop other ta
sks until it has
finishe
d
playing. If yo
u requi
re oth
e
r ope
ratio
n
s
while pl
aying
the spe
e
ch, sound o
r
soun
dsc functio
n
can
be used for n
on-synchrono
us play
. The
differen
c
e bet
wee
n
them is that sound
sc calibrates th
e
values in the
x ranging from -1 to 1, for formali
z
in
g pro
c
e
ssi
ng,
in order to
achi
eve the best
perfo
rman
ce.
3) wavreord
(n,fs
)
It reco
rd
s n
sampl
e
s of a
n
audi
o
sign
al, sam
p
led
at a rate of
Fs
Hz (sam
p
l
es
per
se
con
d
). The
default value
for Fs is 1
102
5 Hz.
4)
wavwrite (x, fs, n, ’filename’)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Voice Colle
ction und
er Diff
erent Spe
c
tru
m
(Min Li)
2643
It writes the data s
t
ored in t
he variable y
to a WAVE file c
a
lled filename. The
data has
a sampl
e
rat
e
of
Fs Hz
a
nd i
s
N-bit, where
N i
s
8, 16,
24,
or
3
2
.
Fo
r N < 32, amplitude
val
ues
outsid
e
the ra
nge [-1,
+1] are clipp
ed.
5)
w
a
vplay (
y
,Fs)
It plays the audio si
gnal
stored in th
e vect
or y on a
PC-b
ased au
dio output de
vice.
You sp
ecify the audi
o sig
nal sa
mpling
rate with
th
e intege
r Fs
in sam
p
les
p
e
r second. T
he
default
valu
e for
F
s
i
s
110
25 Hz
(sam
pl
es per se
co
n
d
).
wavpl
a
y
sup
port
s
o
n
ly 1- or 2
-
chan
nel
(mon
o or ste
r
eo) au
dio si
g
nals.
3.3. Use of G
r
aphic User Inter
f
ac
e
GUI [9], [10] can
be u
s
ed t
o
create
cont
rols
nee
ded i
n
this pa
pe
r. Thre
e dynam
ic text
box FH, FL, Fs are u
s
e
d
, they are respe
c
tive
ly the upp
er cut-off freque
ncy, lower cut
-
off
freque
ncy
an
d the
samplin
g fre
que
ncy.
A pop
-up
me
nu
i
s
used to
sele
ct the
FI
R filter type.
Add
five axes: two wavefo
rms on the left are re
sp
ectivel
y
for the orig
inal sig
nal time-d
omain
a
n
d
freque
ncy
-
do
main, wavef
o
rm
s on the
right
sh
ow the time-do
m
ain and freque
ncy-dom
ain
waveforms af
ter filtering. T
he Middle i
s
the amplit
ud
e-freque
ncy for filtering. Buttons o
n
the rig
h
t
are fun
c
tion selectio
n key
s
.
3.4 Simulation Res
u
lts a
nd Analy
s
is
(1) De
sign
of high-pa
ss
filter
If the hig
h
fre
quen
cy p
a
rt
of sp
ee
ch
si
g
nal i
s
to
be
e
x
tracted,
put
the si
gnal
in
the hi
gh
-
pass
filter. Fs is
s
e
t for
400Hz
, FH is
s
e
t
for 100H
z. At this time, hig
h
freq
uen
cy i
s
sele
cted
wh
ile
low fre
que
ncy is filtered. Comp
ared wi
th the
origin
a
l
signal, the
sou
nd be
co
mes m
o
re
acute
after high
-pa
s
s filter.
Figure 1. Hig
h
-pa
s
s filter
(2) De
sign
of band
-sto
p
filter
Re
set the up
per a
nd lo
wer cut-off freq
u
ency, low fre
quen
cy and h
i
gh frequ
en
cy band
is reserved,
interme
d
iate
freque
ncy i
s
filter
ed th
ro
ugh the
ban
d stop
filter. The
spe
c
trum
cha
nge
s of the spee
ch b
e
fore a
nd aft
e
r filteri
ng
sh
ows: To som
e
extent, spe
e
ch
after filter
become
s
lower, but it is cl
ose to the o
r
iginal spee
ch
sign
al.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 5, May 2013 : 2641 – 264
7
2644
Figure 2. Band-sto
p
filter
(3)
De
sign of
low pa
ss filter
Select lo
w-pa
ss filter, m
odi
fy the param
et
ers
of high
freque
ncy a
n
d
low frequ
en
cy. The
waveform sh
ows that: the low freque
n
c
y is re
se
rve
d
while the h
i
gh frequ
en
cy is filtered, the
spe
e
ch after low-pa
ss filter becom
es lo
w and bori
ng.
Figure 3. Low-pa
ss filter
(4)
Real time
recording of spee
ch si
gnal
The p
ape
r in
cre
a
ses the
real
-time recordin
g
spe
e
ch sig
nal fu
nction. Click th
e sta
r
t-
record
butto
n,
spe
e
ch signal can be
re
corded
a
nd saved. S
e
lect filter ty
pe, modify the
para
m
eters, we ca
n
extra
c
t
the
ce
rtain
freq
uen
cy ra
nge
by the fu
nction
of
real
-time
re
cordi
ng
spe
e
ch sig
nal
. This pa
pe
r a
dopts
ban
d-p
a
ss filter
, extracting th
e fre
quen
cy from
50Hz to 15
0Hz.
The sp
ee
ch b
e
com
e
s a
c
ut
e after the ba
nd pa
ss f
ilter,
but is lowe
r than hig
h
pa
ss filter.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Voice Colle
ction und
er Diff
erent Spe
c
tru
m
(Min Li)
2645
Figure 4. Re
cordin
g sp
ee
ch sign
al
(5) Sou
nd effects of differe
nt FIR filters
With
the com
pari
s
on of
wa
veforms and sou
nd
effe
cts,
we
can see:
The spe
e
ch sou
n
d
s
lowe
r if the low freq
uen
cy is extra
c
ted .In the
cont
ra
st, the spee
ch
beco
m
e
s
sh
arp. The
re
su
lts
of the compa
r
ison am
ong t
he four f
ilters are sho
w
n in
the table 1.
Table1. Sou
n
d
effects of di
fferent FIR filters
Filter t
y
pe
Freque
nc
y
selection
Sound effects
High pass
High frequenc
y
pitch
、、
sharp
har
sh
Band stop
Band freque
nc
y
low
、
but close to
original signal
Lo
w
pass
Lo
w
freq
uenc
y
deep
、、
lo
w
borin
g
Band pass
Middle frequenc
y
sharp
、
but close to original signal
4. Sy
stem Implementa
tio
n
4.1. Hard
w
a
r
e
Design Sc
heme
The sy
stem
hard
w
a
r
e d
e
s
ign
schem
e
is sho
w
n in
figure 5, it m
a
inly co
nsi
s
ts of DSP
pro
c
e
ssi
ng m
odule an
d sp
eech sign
al acq
u
isitio
n
module. Its wo
rkin
g pro
c
e
ss is as follows:
mike a
c
qui
re
s a spe
e
ch
signal, put
it into
the
sign
al con
d
itioning ci
rcuit. Speech
sig
nal
acq
u
isitio
n a
nd processin
g
m
odul
e (T
LV320AIC23) [11] mainly
compl
e
tes sp
eech
sig
nal A/D
and D/A conversion [
12]; TMS320VC55
09A (DSP)
p
r
o
c
e
ssi
ng mod
u
l
e
reali
z
e
s
the
comm
uni
cati
on
with a
udio
chi
p
, an
d
co
mpletes the
spee
ch
sig
nal
pro
c
e
ssi
ng fu
nction
[13]. T
h
e
spe
e
ch si
gna
l after p
r
o
c
e
ssi
ng i
s
filtered an
d
ca
rri
ed on
po
we
r amplifier, th
en playb
a
ck
the
spe
e
ch by the head
pho
ne.
Figure 5. Hardwa
re de
sig
n
sch
eme
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 5, May 2013 : 2641 – 264
7
2646
4.2. Soft
w
a
r
e
Flo
w
Char
t
System software co
ntain
s
the main
prog
ra
m a
n
d
interru
ption
sub
r
outin
e. The main
prog
ram
com
p
letes the
sy
stem
hardware initiali
zation
, key
scanni
n
g
an
d
key p
r
oce
s
sing.
Wh
ile
interrupt pro
g
ram mai
n
ly complete
s
data acq
u
is
it
ion, softwa
r
e
filtering, sp
eech re
cove
ry
pro
c
e
ssi
ng.
Figure 6. Software flow cha
r
t
Parts of the software p
r
o
g
ram cod
e
are as follo
ws:
void main()
{
uint16 AIC23
data = 0 ;
// Initialization CSL library
CSL_init
();
// Set system freque
ncy : 144 MHZ
PLL_config
(&myConfig
);
hMcbsp =
M
C
BSP_open(MCBSP_PO
RT1,M
C
BSP_OPEN_RES
ET);
//s
etMc
BSP1
MCBSP_conf
ig(hMc
bs
p
,
&
M
c
b
s
p
1Config);
//s
tart Mc
BSP1
MCBSP_s
t
art
(
hMc
b
s
p
,MCBSP_RCB_S
TART| MCBS
P_XMIT_START,0);
....
..
//set AIC23 di
gital interface
I2C_
write
(
digi
tal_audi
o_int
eface
_
form
at, //pointer to data arrary
2,
//length of data to be transmitted
1,
/
/
mas
t
er
or s
l
aver
CODEC_A
D
DR,
/
/
s
lave address
1,
/
/trans
fer mode
3000
0
//time out for busy bu
s
);
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Voice Colle
ction und
er Diff
erent Spe
c
tru
m
(Min Li)
2647
//setAIC23
s
a
m
pling fre
que
ncy
I2C_
write
(
sa
mple_
r
ate_
co
ntrol,2,1,CODEC_ADDR,1,3000
0);
//s
tart AIC23
I2C_
write
(
digi
tal_interfa
c
e_
activation,2,1
,
CODE
C_A
D
DR,1,3
000
0);
while
(T
RUE
)
{
//spee
ch si
gn
al pro
c
e
ssi
ng
prog
ram
}
5. Conclusio
n
The extra
c
tio
n
of spee
ch
signal of di
fferent frequ
ency ba
nd
has ma
ny extensive
appli
c
ation
s
, su
ch a
s
spe
e
ch
re
cog
n
ition, ca
rtoon sound synthe
sis,
etc.
Thi
s
pape
r combi
nes
MATLAB GUI and M do
cuments , e
s
tablishe
s sp
e
e
ch
sign
al e
x
traction mo
del; reali
z
e
s
the
spe
e
ch sig
n
a
l
acqui
sition,
different fre
quen
cy
ban
d
extraction,
playba
ck fun
c
tion
s ba
sed
on
MATLAB soft
ware pl
atform . It has b
een ve
rified
by the h
a
rd
ware pl
atform, whi
c
h
co
ntains
TLV320AIC2
3
and TMS3
20VC5
509A
.The re
sults
sho
w
that the algorith
m
is efficient a
nd
feasibl
e
, thus reali
z
e th
e i
n
sp
ectio
n
of
theoreti
c
al
kn
owle
dge. At the same tim
e
, it provide
s
a
necessa
ry premise fo
r ne
w
develo
p
me
nt and validat
ions.
Referen
ces
[1]
Shen
gp
eng
Guan, P
e
i
an
He
, Keha
n
Liu,
etc.
Spe
e
ch
recog
n
itio
n
an
d co
ntrol s
y
st
em b
a
sed
o
n
T
M
S320VC55
09A.
Applic
ati
on of Electro
n
i
c
T
e
chnol
ogy
. 200
7; (07): 36~
39.
[2]
N
y
oma
n
Rizkh
a Emill
ia, Su
ya
nto, W
a
rih Ma
hara
n
i.
Isolate
d
W
o
rd Rec
o
g
n
itio
n Usi
ng Er
god
ic Hi
dd
e
n
Markov Mo
del
s and G
enetic
Algorit
hm.
T
E
LKOMNIKA Indones
ian
Jour
n
a
l of
El
ectrica
l
Engi
ne
erin
g
.
201
2; 10(1): 12
9~
136.
[3
]
C
h
un
me
i
Wang
, H
o
ng
bo
Sun
.
T
M
S320V
C
55x DSP
pri
n
ciple
a
nd
app
l
i
catio
n
. Bei
jin
g
:
Publis
hi
ng
hous
e of electr
onics i
ndustr
y. 200
8.
[4]
T
he design of speec
h sign
al
process
i
ng s
y
s
t
em
based o
n
DSP. W
i
reless
Internet
T
e
chnol
og
y. 20
09
;
(09): 99~
99.
[5]
Z
ongfu
W
u
,
M
i
ngs
han Cai, Ri
xin Ch
en.
R
e
searc
h
an
d r
ealiz
atio
n of a
bnorm
a
l n
o
ise
recog
n
itio
n
s
y
stem b
a
sed
on short-term treatment.
Chi
nese Agr
i
cultur
al Mech
ani
z
a
t
i
on
. 201
1; 132(
02): 125~
1
28.
[6]
Yupi
ng Su, Qi
ong
qio
ng Z
h
e
ng, Do
ngj
u Yu
. De
sign
of F
I
R Dig
ital F
ilter
Based
on MA
T
L
AB.
China
Scienc
e an
d T
e
chn
o
lo
gy Info
rmati
o
n
. 20
08;
(08): 144~
1
45.
[7]
Peng
W
e
i. T
he D
e
sig
n
Meth
od
of Assistan
ce of
Dig
ital F
ilter b
a
se
d o
n
MAT
L
AB
w
i
nd
o
w
fu
nctio
n
.
Co
mp
uter Kno
w
ledge a
nd T
e
chno
logy
. 2
012
; 8(18): 456
4~
4
566.
[8]
Min
y
in
g Ji
ng,
Shumi
ng L
o
n
g
.
Collecti
ng
an
d Operati
on of
Sign
al us
ing
MAT
L
AB
. Informati
on
an
d
Co
mmun
icati
o
n
. 2012; (3): 10
~
11.
[9]
Aihu
a Z
han
g. Desig
n
an
d Implem
entatio
n o
f
Speech Si
gna
l Acquis
i
tion a
n
d
Processi
ng S
y
stem Bas
e
d
on DSP.
Journ
a
l of Z
hong Y
u
an Un
iversity o
f
T
e
chnolo
g
y
. 200
9; 20(4): 10
~
11.
[10]
EVC Sekhar
a Rao, Dr PVN Prasad. F
i
nite
Eleme
n
t Method Usin
g PDE
T
OOL of Matl
ab for H
y
br
i
d
Stepper
Motor
Desi
gn.
T
E
L
K
OMNIKA Ind
ones
ian
Jo
urn
a
l
of Electric
al
Eng
i
ne
eri
n
g
. 201
2;
1
0
(4)
:
680~
6
86.
[11]
Jun Li, Ha
ib
in
Z
hou, Shen
gl
in Qiu, Yanz
h
en Z
han
g, Xia
o
y
an C
a
i. Des
i
gn of Sp
eech
Recog
n
itio
n
S
y
stem Bas
ed
on DSP.
Scien
c
e Mosaic
. 20
1
1
; (07): 118~
1
2
2
.
[12]
Xu
eMi
n
Z
h
a
n
g
,
Jian
hon
g Z
h
ang, Y
ans
ong
Kang. T
he d
e
s
i
gn
an
d si
m
u
la
tion
of F
I
R filter b
a
sed
o
n
GUI.
Journal o
f
Changc
hu
n Institute of T
e
chnol
ogy
. 20
09; 10(0
4
): 33~
35.
[13]
Xi
ao
yo
ng
F
an,
Cha
ngl
in Z
h
o
u
, Xiao
ji
ng Ji
a
.
Real
izatio
n o
f
speec
h sig
n
a
l co
llecti
on
br
oadc
ast an
d
digit
a
l ech
o
ba
sed on DSP.
El
ectronic Me
asu
r
ement T
e
chn
o
l
ogy
. 20
07; 30(
8): 103~
10
6.
Evaluation Warning : The document was created with Spire.PDF for Python.