TELKOMNI
KA
ISSN:
1693-6930
■
Analysi
s, De
sign a
nd Im
plem
entation of An Em
bedded RealTim
e
Sound
…
…
(Arko Dj
ajadi
)
151
ANALYSIS, DESIGN AND IMPLEMENTATION
OF AN EMBEDDED REALTIME
SOUND SOURCE LOCALIZATION SYSTEM
BASED ON BEAMFORMING THEORY
Arko
Djajadi, Rusman
Ru
s
y
adi, Tommy
Handoko,
Maralo Sina
ga, Jürgen G
r
ueneb
e
rg
Mech
atro
nics Depa
rtment, Faculty of
Enginee
ring, Swiss G
e
rm
an University
BSD City, Indone
sia, Telp. 021-537
622
1
ext. 760
Email: arko
@sgu.a
c
.id, tommy.hando
ko.87@gmail.
com
A
b
st
r
a
k
Proyek ini b
e
r
tujuan u
n
tuk m
enganalisa
,
m
endesain
dan m
e
wuju
d
k
an
seb
uah
sistem
pen
cari ara
h
sum
ber sua
r
a
wa
ktu-nyata
yan
g
dip
a
sang di
seb
u
a
h
rob
o
t bero
da. Sistem
yang
dibuat m
e
m
anfaatka
n d
u
a
bua
h m
i
krof
on
seb
agai
s
ensor su
ara, sistem
m
i
kro
k
ontrolle
r Ard
u
ino
Duem
ilano
ve
denga
n AT
Mega3
28p
sebag
ai m
i
kro
p
ro
se
so
r, du
a buah m
o
tor DC m
a
g
net
perm
ane
n se
bagai a
k
tuato
r
peng
gerak
robot, satu m
o
tor se
rvo se
bagai pe
ngg
e
r
ak
web
c
am
ke
arah
sum
ber
sua
r
a, se
rta l
aptop/PC unt
uk tam
p
ilan d
an sim
u
lasi.
Untu
k pen
ent
uan arah sum
ber
sua
r
a, teo
r
i b
eam
form
ing diterap
k
a
n
. Sekali
ara
h
su
ara tel
ah di
d
e
teksi da
n di
tentuka
n
, ad
a 2
pilihan bai
k untuk m
engg
era
k
kan ro
b
o
t kearah su
ara atau
pun
hanya
web
c
am
yang a
k
a
n
berp
u
tar
ke a
r
ah sum
ber suara
seol
ah terjadi d
a
lam
ruang
konfe
r
e
n
si vi
deo. Se
cara terinteg
rasi
sistem
telah
diuji da
n h
a
si
lnya m
e
n
unju
k
kan b
a
h
w
a
sistem
m
a
m
pu seca
ra
ce
p
a
t dalam
wa
ktu
nyata
m
ene
m
u
kan
ara
h
sum
ber sua
r
a yang
p
o
si
si
nya
a
c
a
k
dal
am
bidan
g
setenga
h lin
gkaran
(0-180
0
) b
e
rj
ari-j
a
ri
0.3m
hingg
a 3m
,
deng
an
asu
m
si si
stem
sebag
ai titik p
u
satn
ya. K
a
rena
faktor
ke
cepa
tan ADC da
n pro
c
e
s
o
r
ya
n
g
rend
ah, ha
sil resol
u
si
sud
u
t terbaik b
e
rkisar 25
o
.
Kata kunci
:
beam
form
ing, correlation, em
bedded
system
, realtim
e
, sound source
A
b
st
r
a
ct
This p
r
oje
c
t is intend
ed to anal
yze, desig
n and im
plem
ent a re
altim
e
sound
sou
r
ce
locali
zation
system
b
y
u
s
i
ng a m
obile
robot
as th
e
m
edia. The i
m
plem
entated system
u
s
e
s
2
m
i
croph
one
s as the
sen
s
ors, Arduin
o
Duem
ilano
ve
m
i
crocontroll
er system
with ATMega
32
8p
as the m
i
crop
rocesso
r
, two
perm
ane
nt m
agnet DC
m
o
tors a
s
the
actuato
r
s for the m
obile robot
and a
se
rvo
m
o
tor as the
actuato
r
to ro
tate the
web
c
am
directing
to the locatio
n
of the so
un
d
sou
r
ce, an
d
a laptop/P
C
as the
sim
u
la
tion and
di
spl
a
y m
edia. In
orde
r to
achi
eve th
e o
b
je
ctive
of finding the
positio
n of a
spe
c
ific
so
u
nd source
, b
eam
form
ing theory i
s
ap
pli
ed to the
syst
em.
Once the lo
cation of the sou
nd sou
r
ce is det
e
c
ted
and dete
r
m
i
ned, the choi
ce is
either t
he
m
obile robot
will adjust its positio
n according to
the dire
ction o
f
the sound
sou
r
ce or on
ly
web
c
am
will rotate in
the
d
i
rectio
n
of the
incom
i
ng
so
und
sim
u
latin
g
the
u
s
e
of t
h
is
system
in
a
vide
o confe
r
ence. The in
tegrat
ed
syst
em
has bee
n tested and
the result
s sho
w
the system
coul
d lo
cali
ze
in realtim
e
a
sou
nd
sou
r
ce
pla
c
ed
ra
ndo
m
l
y on a
half
circle
are
a
(0
- 1
8
0
0
) with a
radiu
s
of 0.3
m
- 3m, assu
m
i
ng the system
is the
cen
t
er point of th
e circle. Due
to low ADC a
nd
pro
c
e
s
sor
sp
eed, achie
v
a
b
le be
st angu
lar re
sol
u
tion
is still lim
ited
to 25
o
.
Key
w
ords
:
beam
form
ing,
correl
ation, em
bedded system
, realtim
e
, sound source
1. INTRODUCT
I
ON
Huma
n’s
ears could p
r
e
d
i
c
t the in
comi
ng sound f
r
o
m
all dire
ctio
ns (3 dime
nsi
ons). The
algorith
m
explaining this p
henom
eno
n is calle
d ce
te
ra algorithm. It involves the shado
w sou
n
d
cre
a
ted by th
e head a
nd the refle
c
tion
cau
s
e
d
by
the edge
s of the oute
r
ears. Ho
weve
r, to
s
i
mu
la
te th
e
c
o
mp
le
te
he
ar
in
g
s
y
s
t
e
m
on
ly b
y
us
ing
2 mi
cro
pho
ne
s, it is extre
m
ely difficult y
e
t
chall
engin
g
. With a pair o
f
microph
one
s, it is only p
o
ssible to de
tect an inco
ming so
und i
n
2
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 1
693-693
0
TELKOM
NIKA
Vol. 7, No. 3, Desem
b
e
r
2009 : 151
- 160
152
dimen
s
ion
s
,
without
being
able to
di
stin
guish whet
he
r the
so
und
is comi
ng from
the fro
n
t or the
back of the system.
This p
r
oje
c
t is intende
d to explain the h
u
man he
arin
g
system in 2 dimen
s
ion a
r
ea usin
g
a pai
r of mi
croph
one
s. Th
e appli
c
atio
n
s
which a
r
e
going to
be
simulated in
the p
r
oje
c
t are a
video co
nference syste
m
and a st
e
e
r
ing/navig
atio
n system o
n
a mobile robot ba
sed
on
incomi
ng
sou
nd directio
n
usin
g a lo
w-co
st micro
c
o
n
trolle
r a
s
th
e brai
n of th
e syste
m
. The
microcontroll
er is
chea
p a
nd available i
n
the local m
a
rket.
The st
rategy
of estimating
the dire
ction
of
the incom
i
ng soun
d is
based o
n
the
arrival
time differen
c
e of the so
und re
aching
the 2
micro
phon
es. To
achi
eve the obje
c
tive, cross
correl
ation m
e
thod [1,2,3] and positio
n
estimation tech
niqu
e mu
st be implem
ented [4,5]. The
real time
cont
rol sy
stem is i
m
pleme
n
ted
within a
readil
y
available mi
cro
c
o
n
troll
e
r.
Re
sea
r
che
s
in
area
of th
e
re
altime em
bed
ded
syste
m
s
are
still
cu
rre
n
t and
p
r
og
re
ssi
ng
with th
e
aim
of cre
a
ting
sma
r
ter devi
c
e
s
or ap
plications, wh
ere
soci
al
robot
s are o
ne of the most co
mmon testb
e
d
s
[6,7,8]. This proje
c
t is gea
red towa
rd th
a
t
goal and to provide a
wo
rking
system.
The pa
per i
s
orga
nized a
s
follows: se
ction II
presents the prin
ciple
s
of cross
correlation
and position
estimation usi
ng beamformi
ng theory,
section III presents the short methodol
ogy to
achi
eve the
main pu
rp
ose
of the the
s
is proj
ect, se
cti
on IV pre
s
e
n
ts the exp
e
ri
mental results of
the system, a
nd se
ction V pre
s
ent
s t
he gene
ral con
c
l
u
sio
n
of the proje
c
t.
2.
CROSS CORRELATION
AND POSITION ESTIMATI
O
N
In this se
ctio
n, beamformi
ng theo
ry usi
ng
cro
ss
co
rrelation an
d p
o
sition e
s
tim
a
tion is
evaluated.
Figure 1. Example of cross correlatio
n
2.1. Cros
s-c
o
rrel
a
tion
Cro
s
s co
rrela
t
ion is a mea
s
ure of simila
ri
ty between t
w
o wavefo
rm
s as a fun
c
tion of a
time-lag
appl
ied to
one
of them [1,2
,3]. Figur
e
1
is
an
exam
ple of
cross co
rrelation
for
measuri
ng time-la
g
between to simila
r sign
als.
Th
e expre
ssi
on
to find the
delay time of two
simila
r sig
nal
s is given by
Equation 1:
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOMNI
KA
ISSN:
1693-6930
■
Analysi
s, De
sign a
nd Im
plem
entation of An Em
bedded RealTim
e
Sound
…
…
(Arko Dj
ajadi
)
153
(1)
whe
r
e x
i
is the sign
al re
cei
v
ed by micro
phon
e i and x
j
is the signal
received by microph
one j.
is the delay time of the two sign
als. Th
e point
wh
ere the pea
k of the cro
s
s co
rrel
a
tion re
su
lts
shows the del
ay time of the two signal
s,
as illustrated
below in Figure 1.
There are two method
s of
correlatin
g si
gnal
s. The first one i
s
non
-ci
r
cular m
e
thod (no
data wrappi
n
g
). This m
e
th
od is u
s
ed if
the signal
s
are n
o
t conti
nuou
s o
r
in other
words,
the
sampl
ed si
gn
al rep
r
esents
the whole fo
rm of the si
gn
al. The se
con
d
one is ci
rcu
l
ar metho
d
. It is
use
d
if the
si
gnal
s are
con
t
inuou
s (fo
r
e
x
ample
sinu
soid o
r
re
ctan
gular wave
) o
r
in oth
e
r
wo
rds,
the sam
p
led
sign
al only re
pre
s
ent
s som
e
part
s
of
the whole fo
rm
of the signal.
It is said to be
circula
r
b
e
ca
use
wh
en th
e data i
ndex
is o
u
t of ra
nge, it is "
w
rappe
d" ba
ck within
ran
g
e
,
for
example: x(-1
) = x(N-1
)
, x(N+5) = x(5).
2.2. Position
Esti
mation
Once the
cro
s
s correlatio
n
is pe
rform
e
d
to find
the
d
e
lay time
of t
he
sign
als, th
en it i
s
possibl
e to predi
ct or e
s
timate the dire
ct
ion of the incomi
ng
sound b
a
sed on ce
rtai
n
assumptio
n
s
to simplify modeling [4,5]. They are:
The so
und
sp
eed in air i
s
c
The so
und
so
urce is lo
cate
d far from the
micro
pho
ne
s
There is no e
c
ho a
nd bo
un
cing effe
ct
Figure 2. Geo
m
etry of posit
ion estimatio
n
[4].
Referrin
g to
the a
s
sumpti
ons and
Fig
u
re
2, t
he mathemati
c
al
expre
s
sion
to
cal
c
ulate
the
incomi
ng an
g
l
e of the soun
d can b
e
deri
v
ed by
simple
trigonom
etry comp
utation
as follo
ws.
(2)
(3)
So,
(4)
Φ
is the angle of so
und
source di
re
ct
ion
c is the sound
spe
ed i
n
the air (me
a
su
red
)
s is the di
stan
ce bet
wee
n
the microp
hon
es
∆
T
ij
is the del
ay time of the signal
s from
the mi
croph
o
nes (cal
cul
a
ted from cro
ss correlatio
n)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 1
693-693
0
TELKOM
NIKA
Vol. 7, No. 3, Desem
b
e
r
2009 : 151
- 160
154
3. SHORT
MET
H
O
D
OL
OGY
Figure 3
sh
o
w
s wh
at the
environ
ment
of the
sy
stem
looks li
ke. T
he
soun
d
sou
r
ce
could
be pla
c
e
d
an
ywhere in f
r
o
n
t of the sy
stem (ran
ge of
0
o
to 180
o
).
T
he
incoming sound signal will
be re
ceive
d
b
y
microp
hon
e
1 and 2, an
d
then it is
pro
c
e
s
sed in the
system. The
system h
a
s
3
outputs
and t
hey are
se
nt to the laptop
or PC, to
the
web
c
am’
s
se
rvo, and to th
e mobile
rob
o
t.
Either the webcam or the mobile
robot will re
sponse
by moving toward the direction
of the
incomi
ng so
und. The m
obile ro
bot’s enco
der
will
give an an
gular fee
dba
ck b
a
sed on
the
numbe
r of pu
lse
s
to the m
a
in syste
m
. In addition,
th
e laptop o
r
P
C
will pl
ot the sou
nd
sign
als
received by
microph
one
1 and 2, the
cro
s
s correlat
i
on re
sult fro
m
both sig
nal
s, and it will
also
sho
w
the pi
cture taken fro
m
the web
c
a
m
.
Figure 3. Glo
bal system ov
erview
The Fig
u
re
4
sho
w
s the m
o
re d
e
tailed
block
dia
g
ra
m of the syst
em. The sen
s
ors u
s
e
d
in the
system
are
co
nde
nser mi
cr
opho
n
e
s. Th
ey are
che
ap a
nd av
ailable i
n
the
local
market.
In
addition, the
y
also have a good sen
s
i
t
ivity whic
h
make
s them
suitable for
the system. The
microph
one
s are co
nne
cte
d
to
thei
r re
spective pre
-
a
m
plifier
circui
t to pro
d
u
c
e
a wea
k
el
ectric
sign
al rep
r
e
s
enting the so
und wave.
Still within Figure
4, the next circuit is
anti-alia
sing filter. The purpose
of the filter is to
limit the hi
gh
est frequ
en
cy co
mpo
nent
s allo
wed
to e
n
ter th
e
syst
em. It sh
ould
empl
oy an
a
n
ti-
aliasi
ng filter
to fulfill the Nyqui
st sampli
ng theorem that the ma
ximum frequency of the signal
s
h
ould be half of the s
a
mpling rate of the A/D c
onverter [1,3].
Next stage is band-pa
ss filter and
main
amplifier. The
band-pa
ss filter here i
s
to rest
rict
the band
widt
h of the sign
al from 200 to 3000 Hz.
T
h
is is the ba
n
d
width of the
average h
u
man
voice. Ho
wev
e
r, in a norm
a
l con
d
ition, the voice fre
quen
cy is fro
m
200 to 80
0 Hz ba
se
d on
experim
ent u
s
ing
sou
nd spectrum an
al
yser (se
e
Fig
u
re 5
)
.
Again ref
e
rring to Figu
re
4, the main
am
plifier
circuit is u
s
in
g
non-i
n
vertin
g op-amp
config
uratio
n. Whe
never the si
gnal l
e
vel is
too l
o
w, the potenti
o
meter
co
uld
be adj
uste
d
to
amplify the
si
gnal l
e
vel. Th
e level
shifter is u
s
ed
to
ad
d a
DC
co
mp
onent to
the
sign
al. Th
e A
/
D
conve
r
ter can
only
co
nvert
the si
gnal
wit
h
in the
r
ang
e
of 0
-
5
V. Th
erefo
r
e, the
A
C
sign
al
sho
u
l
d
be add
ed wit
h
DC voltag
e
and shifted u
p
to meet
the device re
qui
rement. Then
it will be limited
by voltage limiter to avoid the sig
nal
exceedin
g
the limit of 0V or 5 V.
Next, the sig
nal ente
r
s th
e digital processi
ng afte
r being
conve
r
t
ed by the A/D device.
Here all
the
calcul
ation ta
kes
pla
c
e. T
h
e
cro
s
s correla
t
ion an
d p
o
sit
i
on e
s
timatio
n
techniq
ue
a
r
e
impleme
n
ted
here. ATMe
g
a328
p is emp
l
oyed as
the
microprocessor of the syst
em [9].
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOMNI
KA
ISSN:
1693-6930
■
Analysi
s, De
sign a
nd Im
plem
entation of An Em
bedded RealTim
e
Sound
…
…
(Arko Dj
ajadi
)
155
Figure 4. Main system blo
c
k diag
ram
Figure 5. FFT
of human voice (m
ale)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 1
693-693
0
TELKOM
NIKA
Vol. 7, No. 3, Desem
b
e
r
2009 : 151
- 160
156
In the last p
a
rt of Figure
6, in ord
e
r to
sh
ow h
o
w the
raw
sign
als a
r
e cal
c
ul
ated
by cro
s
s
correl
ation m
e
thod within t
he microcontroller to get
the delay time of two signal
s, PC software
is
made. Th
e software i
s
d
e
v
eloped u
s
in
g C p
r
og
ram
m
ing lan
gua
ge. The mai
n
purpo
se of
the
softwa
r
e i
s
mainly to plo
t
the data se
nt by t
he microcontroller vi
a se
rial to the PC. The pl
ots
show the si
gnals’ form
s an
d the correl
a
tion results, as
illustrated in
Figure 1.
Figure 6 sho
w
s the flowch
art of the logic pr
o
g
ram inside the micro
c
ontrolle
r. The value
s
of int MODE and int PLOT
are obtain
e
d
from PC soft
ware via se
ri
al comm
uni
cation. After being
processed, the final output
will be
sent to the actuators.
The servo
motor will
act
as the actuat
o
r
of the video confe
r
en
ce
system simul
a
tion and the
DC m
o
tors will perform as the actuato
r
s of
the mobile ro
bot.
(a). Main p
r
o
g
ram
(b). sub p
r
ocess: Process
Figure 6. The
logic flow of the pro
g
ram in microcontro
ller
4. RESULTS
In order to
re
cord th
e p
e
rf
orma
nce of
th
e
whole
sy
ste
m
, seve
ral
ex
perim
ents a
r
e
ca
rri
ed
out as
sh
own
in the Fig
u
re
7 as the b
a
si
c me
as
urem
e
n
t setup,
whe
r
e the
po
sitio
n
of the
soun
d
sou
r
ce i
s
ch
ange
d a
r
ou
n
d
a
half
circumfe
ren
c
e
with the
micropho
ne
s a
s
the center.
T
h
e
con
c
e
p
t is already illustrated in Figu
re 2
and Figu
re 3
above.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOMNI
KA
ISSN:
1693-6930
■
Analysi
s, De
sign a
nd Im
plem
entation of An Em
bedded RealTim
e
Sound
…
…
(Arko Dj
ajadi
)
157
Figure 7. Experime
n
t Layo
u
t
The first a
s
sumptio
n
of
position
est
i
mation tech
nique i
s
the
soun
d spe
ed in air.
Ho
wever, the
sound
spe
e
d
in air is different from o
ne place to anothe
r dep
e
nding on the
air
comp
ositio
ns.
The
r
efo
r
e,
an exp
e
rim
e
nt to me
asure the
actu
al
so
und
spee
d in th
e a
c
t
ual
environ
ment sho
u
ld be pe
rforme
d.
F
r
o
m
Tabl
e
1,
th
e ave
r
age
so
und
sp
eed
in
air is c=3
1
7
m
/s.
This data
will be used in the pos
ition esti
mation cal
c
ul
ation.
Table 1. Experime
n
t result of the sound
spe
ed in air
(c)
Dista
n
c
e
(c
m
)
Time Del
a
y (ms)
Sound Spe
e
d
in Air (m/s)
13
0.42
309.52
20
0.66
303.03
22
0.68
323.53
24
0.7
342.86
30
0.98
306.12
Next impo
rta
n
t experim
ent
is to g
e
t the
calib
ration va
lue of the
system. The
cali
bration
of the delay
time betwe
en the 2 si
g
nals
sho
u
ld
be pe
rform
e
d in the digi
tal pro
c
e
ssi
n
g
to
comp
en
sate t
he multiplexe
r delay in the
A/D conve
r
ter an
d the error d
ue to th
e com
pon
ent
s’
toleran
c
e u
s
e
d
in the syste
m
which will l
ead to the ph
ase
shifting o
f
the signal.
The time del
ay calibration
is calculate
d
by finding the differen
c
e value between the
actual a
nd t
he theo
retica
l calculat
ion.
From exp
e
riments at an
gle 180
o
position with fixed
distan
ce
s
be
tween t
w
o mi
cro
pho
ne
s, the a
c
tual
d
e
l
a
y time cal
c
u
l
ated u
s
ing
cross
co
rrelati
on
techni
que by
the system’s pro
g
ra
m is 16 sample
s.
The multiple
xer delay of the internal
A/D
conve
r
ter of the ATMega
3
28p a
c
cordi
n
g to the
datasheet of Ardui
no Du
emilan
o
ve is 100 µs per
cha
nnel. In t
h
is
system,
2
cha
nnel
s
are used. T
herefore, the
del
ay time of th
e sa
mple
s i
s
200
µs. With this information, the theoreti
c
al val
ue of the delay time of the sound
reaching the
2
microph
one
s
can b
e
cal
c
ul
ated. The calculatio
n is sh
own a
s
follo
ws:
(5)
(6)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 1
693-693
0
TELKOM
NIKA
Vol. 7, No. 3, Desem
b
e
r
2009 : 151
- 160
158
By substitutin
g
Eq. (5) to Eq. (6) an
d put
ting t
he value
of s=20
cm a
nd c=31
7 m/s, then: n=16.
85
sampl
e
s.
So, by substracting the th
eoreti
c
al an
d
the act
ual value, it is found that the calibratio
n
value is 0.85
sampl
e
s. In
this exp
e
rim
e
nt, all
data
are in inte
ger value
s
, whi
c
h
mean th
ere
is no
fraction
ed
nu
mber be
cau
s
e all valu
es a
r
e
rou
nded
d
o
wn. T
h
u
s
, in
cludi
ng the
calibratio
n
val
ue in
the calculati
on, 92.8
6
%
results of th
e po
siti
on
e
s
timation
ex
perim
ent a
r
e
su
cce
ssful.
The
sampli
ng freq
uen
cy of the
A/D co
nverte
r dete
r
mine
s
the minimum
mea
s
urable
delay time. T
he
faster the
sampling frequency,
the more
accurate the m
e
asured
delay
time
will be. Since the
delay time
of the 2
sig
nal
s i
s
u
s
ed
to
cal
c
ul
ate
the
ang
ular po
si
tion of the i
n
comin
g
sou
n
d
dire
ction, th
e
preci
s
ion
of
the
dete
r
min
ed a
ngle
al
so
ha
s
a limitat
i
on.
Th
e syst
em coul
d
d
e
tect
the dire
ction
of the incomi
ng so
und
with pre
c
isi
on of
25
o
from ang
ular rang
e of 0
o
-180
o
.
Figure 8. Picture of the sy
stem
Figure 8 sh
o
w
s the pi
ctu
r
e of the syst
em. T
he we
b
c
am i
s
pla
c
e
d
in the front
of th
e
mobile
ro
bot
to sim
u
late th
e video
confe
r
en
ce
sy
stem
. The l
aptop
i
s
con
n
e
c
ted t
o
the
we
bca
m
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOMNI
KA
ISSN:
1693-6930
■
Analysi
s, De
sign a
nd Im
plem
entation of An Em
bedded RealTim
e
Sound
…
…
(Arko Dj
ajadi
)
159
to show the i
m
age from it. Two DC mo
tors a
r
e set
unde
r the ba
se plate to move the mobile
robot. An en
coder i
s
empl
o
y
ed to give the positio
n fee
dba
ck of the
mobile robot to the system.
Table 2. Re
sults of positio
n estimation
Inc
o
ming
sou
nd
dire
ction
Theo
retical
cal
c
ulatio
n
(
s
amp
l
es
)
Experiment
data of delay
(
s
amp
l
es
)
Delay after
calibration
(
s
amp
l
es
)
Floor
roun
ding
(
s
amp
l
es
)
Statement
of result
0
o
3.15
2
2.3
2
TRUE
30
o
2.73
1
1.88
1
TRUE
45
o
2.23
1
1.38
1
TRUE
50
o
2.02
0
1.17
1
FALSE
60
o
1.57
0
0.72
0
TRUE
70
o
1.07
0
0.22
0
TRUE
80
o
0.54
19
19.69
19
TRUE
90
o
0
19
19.15
19
TRUE
100
o
19.45
18
18.6
18
TRUE
110
o
18.92
18
18.07
18
TRUE
120
o
18.42
17
17.57
17
TRUE
130
o
17.97
17
17.12
17
TRUE
150
o
17.26
16
16.41
16
TRUE
180
o
16.85
16
16
16
TRUE
Table 2 sho
w
s the result of the position
estimatio
n
experiment. The so
und
source is
placed at sev
e
ral po
sition
s rega
rdin
g to the posit
io
n o
f
the system and the results are re
co
rde
d
.
For exam
ple,
the sou
nd
source i
s
pla
c
ed at 110
o
re
spe
c
ted to th
e system
(se
e
Figu
re 7
)
. The
theoreti
c
al
value of t
he
del
ay time i
s
1
8
.
92 samp
l
e
s.
Next, the th
e
o
retical valu
e
is sub
s
tra
c
te
d
with the
calib
ration valu
e (0.85 sample
s) an
d the
re
sult is 18.0
7
sample
s. Thi
s
value is
ro
un
ded
down b
e
cau
s
e this is
calculated i
n
a
n
i
n
teger vari
abl
e, and
the
re
sult i
s
1
8
sa
mples.
The
result
of the calcula
t
ion is the sa
me with the e
x
perime
n
t
data. Therefo
r
e,
the stat
ement
of the result for
this
experiment is
true.
5. CO
NCL
USIO
N
By
usin
g
2 mi
cro
pho
ne
s as t
he sensors, it has been succe
ssf
ully si
mulating a sy
stem to
locali
ze the di
rectio
n of an i
n
comi
ng sou
nd. The preci
s
ion of the sy
stem is 2
5
o
. There i
s
still lack
of accu
ra
cy o
f
the p
o
sitio
n
estimation
re
sult
be
ca
us
e o
f
th
e
s
l
ow
sa
mp
lin
g
fr
eq
ue
n
c
y o
f
th
e
A/D
conve
r
ter. Ho
wever, this li
mitation coul
d be over
com
e
relatively easy by using
an A/D conve
r
ter
whi
c
h ha
s a faster
sam
p
lin
g rate in com
b
ination
with a faster mi
croco
n
trolle
r wi
th large
r
on
chip
RAM. Theref
ore, the
precision
of the
position estimation
will be
com
e
much better. Also the
system still react
s
to every soun
d, even t
he unde
sire
d one (n
oise
). The p
o
ssible furth
e
r
improvem
ent is to add a voice re
co
gnitio
n
system
. So, the system will be able to rea
c
t only to a
s
p
es
ific
s
o
und
o
n
l
y.
REFERE
NC
ES
[1].
Ifeacho
r, E.C. and Je
rvis, B.W., “
Digital Signal Processing”
, Pre
n
tice Hall, 2002
.
[2].
Ark
o
A,
Wat
e
rfall
R. C, B
e
ck
M. S, Dy
ak
ows
k
i
T, Sutc
liffe P, Byars M, “
Dev
e
lopment o
f
Electrical Ca
pacita
nce T
o
mography
for
So
lids M
ass
Flo
w
M
e
asureme
nt a
nd
Contr
o
l
of Pneuma
tic Conv
e
y
ing Sy
stems
”, 1
s
t Wo
rld Co
ngress on Indu
strial Pro
c
e
ss
Tomog
r
ap
hy, Buxton, Greater Man
c
h
e
ster, April 14-1
7
, 1999
.
[3].
Hayki
n
, S. and Veen, B.V., “
Signal and Sy
stems 2n
d Edition
”, John Wil
e
y &
Sons, 200
3.
[4].
Valin, J.M., Micha
ud, F.
and
Rouat,
J., “
Robus
t S
ound Sourc
e
Localizati
on Using
a
Microphon
e Array
on a Mobile Robo
t
”,
Université de
Sherbroo
ke
2005
[5].
Valin, J.M., Micha
ud,
F., Rouat, J., and Leto
u
rne
au, D.,
"
Robus
t S
ound Sour
c
e
Localiza
t
ion
Using
a
Microphon
e
Array
on
a Mobile
Robo
t
",
Proc
. IEEE/RSJ
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 1
693-693
0
TELKOM
NIKA
Vol. 7, No. 3, Desem
b
e
r
2009 : 151
- 160
160
Internation
a
l Confe
r
en
ce
o
n
Intelligent Rob
o
ts
a
nd System
s (IROS),
Vol.2. p
p
. 1228
-1
233
,
2003.
[6].
S. Briere, D. Letourn
eau
, M.
Frech
e
tte, J.-M. Valin, F. Michaud, “
Emb
e
dded a
nd
Integra
t
ion
Audition
for
a Mobile Robot
”,
Proc
eedings
AAAI Fall
Sy
mpos
ium Work
shop
Aurall
y Inform
ed Perform
ance: Integra
t
ing Ma
chi
n
e
Listening a
n
d
Auditory Prese
n
tation in
Rob
o
t
i
c S
y
st
em
s
, FS-06-0
1
, 6-10, 200
6
.
[7].
S. Yamamoto, K.
Nak
a
dai,
J
.
-M. Valin, J
.
Roua
t, F.
Micha
ud, K.
Komatani, T. Ogata, H. G.
Oku
no, “
Ma
king a rob
o
t reco
gnize thre
e simultaneou
s
sente
nce
s
in real-time
”,
Proceedi
ngs of IEEE/RSJ Inte
rnational
Conference on In
telligent Robot
s and System
s
(IROS
)
, 2005
.
[8].
F. Mich
aud,
D. Léto
u
rnea
u, P. Lep
age
, Y.
Morin,
F. Gag
non, P.
Giguè
re, É.
Beaud
ry, Y.
Brosse
au, C.
Côté, A. Du
quette, F.-F.
Lapla
n
te
, M.-A. Legault, P. Moisan, A.
Ponch
on, C.
Raïevsky, M.-A. Roux, T. Salter, J.
-M. Valin, S.
Ca
ron
,
P. Frenette,
P. Masson, F
.
Kaban
za,
M. Lauri
a
, “
Socially
interactiv
e robots for re
al life use
”, Proce
edin
g
s
Worksh
op o
n
Mobile
Ro
bo
t Comp
etition, Ameri
c
an
Asso
ciat
ion for A
r
tific
i
al Intelligenc
e
Conference
(AAAI), 2005.
[9].
Atmel.c
o
m,
“
Da
ta
shee
t
ATm
e
ga
48P
A-8
8
PA
-16
8
P
A-3
28P
”,
448 p
age
s,
revisio
n
D,
update
d
10-2
009.
Evaluation Warning : The document was created with Spire.PDF for Python.