De e mina ion o es sequence o in usi e measu emen o VTQoS …
47
DETERMINATION OF TEST SEQUENCE FOR INTRUSIVE MEASUREMENT OF
VTQOS IN ENVIRONMENT OF FIXED TELECOMMUNICATION NETWORK
Pe e Po a, Ma in Vaculík
Dep . o Telecommunica ions, FEE, Uni e si y o Žilina, Uni e zi ná 1, 010 26, Žilina, Slo akia
email: poc a@ el.u c.sk, aculik@ el.u c.sk
Summa y This pape desc ibes simula ions o es sequences ansmission o in usi e measu emen o VTQoS in
en i onmen o ixed elecommunica ion ne wo k. The aim o simula ions was a de ec ion on he in luence o his
en i onmen on he quali y o ansmission sequences. E alua ion he gene a ed sequences was based on he calcula ion o
mean squa e measu e and co ela ion coe icien . These measu es we e used as a c i e ion o sui able es sequences
selec ion. Reconside a ion o a con enience o he gi en es sequence, which is composed om simple signals, on in usi e
measu emen o VTQoS in he en i onmen o ixed elecommunica ion ne wo ks is he aim o his pape .
1.
INTRODUCTION
VTQoS is one o he impo an pa s o QoS
(Quali y o Se ice). I is e y impo an o he
p o ide s and o he use s. An inc ease in he
complica ion and he complexi y o ne wo ks is
isible, when communica ion ne wo ks inco po a e
mo e and mo e ansmission echnologies. The
measu emen o he oice ansmission quali y
becomes one pla o m o o he pla o ms, by help
o i we can compa e simul aneously di e en
ansmission echnologies and i is such close o
iew o he use s.
O cou se, i is possible o measu e and e alua e
he ansmission pa ame e s o he ne wo ks. Bu
only he e alua ion o end- o-end quali y p o ides
op imal esul s because o he complexi y o
ne wo k echnologies.
Thus, i is he e alua ion in he same way as
use s do. Since oice se ice is he mos wide-
sp ead se ice, in which a use uses il e and
p edica i e abili ies o human b ain, i is c ucial o
op imally e alua e a quali y o such se ice.
E alua ion o he Quali y o oice se ice may
be pe o med using in usi e and non-in usi e
me hod, objec i ely o subjec i ely.
Using non-in usi e me hod, we only moni o
exis ing dialogue. The d awback o his me hod is
ha he e alua ion algo i hm canno u ilize an
o iginal sample o he p ima y signal. Thus, i is
e y di icul o de ec some ypes o signal
dis o ion ha occu du ing ansmission. In he
in usi e me hods, only a es oice sample is
ansmi ed. These me hods ha e been known since
he beginning o he elecommunica ion
echnologies, when he special sequences o owels
(known as
loga homs) we e ansmi ed a e he
connec ion had been buil -up. A ecei e had o
ecognize hese loga homs. This way o subjec i e
e alua ion has been used ill nowadays (e.g. me hod
MoS).
Today’s echnical and so wa e acili ies
p o ide an objec i ica ion o his measu emen
me hod by ansmi ing he sound sample de ined
be o ehand, i s ecei ing on he des ina ion side,
and a compa ison o he ansmission sample and
he o iginal sample using he sui able algo i hm ha
imi a es he way o pe cep ion and e alua ion o he
quali y ansmission opinion by an a e age lis ene .
I is o example E-model de ined in ETR-250, o
algo i hm PSQM (Pe cep ual Speech Quali y
Measu emen ) de ined in P.861 ITU-T also PESQ
(Pe cep ual E alua ion o Speech Quali y) de ined
in P.862 ITU-T.
A choice o op imal es sequence is e y
impo an o all hese me hods.
The es sequence would consis o non-speech-
like ( ully a i icial) signals. These signals a e
close de ined in P.501 ITU-T and he
ecommenda ion di ides hem in o de e minis ic
and andom signals. An ad an age o using hese
signals is simplici y and possibili y o a compa ison
o he esul s measu ed in di e en language a eas.
The es sequence composed om hose signals
enables he compa ison o ne wo ks o indi idual
coun ies wi hin one co po a ion (e.g. Deu sche
Telecom, O ange, Voda one) om he poin o he
iew VTQoS.
Nowadays, he VTQoS in usi e measu emen s
a e pe o med by using samples o speech signal
bu he compa ison is possible only wi hin he
single-language a ea in his case.
He e we ocus o he in luence o BER and SNR
o ansmission o es sequences in en i onmen
o ixed elecommunica ion ne wo k.
2.
DESCRIPTION OF THE TEST
SEQUENCES
The leng h o each o he es sequences is se o
90 sec. This pe iod equals o he leng h o a phone
call o a e age use . The sequences we e c ea ed o
he signals, whose con enience as e i ied in [1].
Th ee ypes o es sequences wi h di e en
pa ame e s o signals we e o med. The c ea ion o
es sequences was based on supe posing Sinusoidal
signal and Gaussian whi e noise on Squa e bipola
signal. We mus hold on he condi ion o
Ad ances in Elec ical and Elec onic Enginee ing
48
o hogonali y. This ule only ela es o pe iodic
signals.
These signals wi h compe en pa ame e s we e used
o c ea e he sequence 21:
• Sinusoidal signal wi h equencies 300,
600, 900, 1200, 1500, 1800 Hz ,
• Gaussian whi e noise wi h
µ
= 0 and
δ
 =
0,0001; 0,005; 0,001; 0,05; 0,025; 0,01.
Squa e bipola signal wi h equency 300 Hz as
used as a ca ie signal.
These signals wi h compe en pa ame e s we e used
o c ea e he sequence 22:
• Sinusoidal signal wi h equencies 400,
800, 1200, 1600, 2000, 2400 Hz ,
• Gaussian whi e noise wi h
µ
= 0 and
δ
=
0,0001; 0,005; 0,001; 0,05; 0,025; 0,01.
Squa e bipola signal wi h equency 400 Hz as
used as a ca ie signal.
These signals wi h compe en pa ame e s we e used
o c ea e he sequence 23:
• Sinusoidal signal wi h equencies 500,
1000, 1500, 2000, 2500, 3000 Hz ,
• Gaussian whi e noise wi h
µ
= 0 and
δ
=
0,0001; 0,005; 0,001; 0,05; 0,025; 0,01.
Squa e bipola signal wi h equency 500 Hz as
used as a ca ie signal.
Fig. 1 Ini ial pa o es sequence 21 (seq21)
Fig. 2 Ini ial pa o es sequence 22 (seq22)
Fig. 3 Ini ial pa o es sequence 23 (seq23)
The p inciple o he c ea ion o he inal es
sequences is based on an a angemen o ini ial
pa s o ele an es sequences, which a e shown in
Figu es 1 – 3. The a angemen s shown in Figu e 1-
3 a e used six imes o o m he inal es sequences.
Thus, each inal es sequence consis s o six pa s.
The signals s ep-by-s ep ha e go he alues
de ined abo e. Tha means, in he second pa o
he es sequence 21 ( om 15 sec. o 30 sec.), he
signals ha e he ollowing alues: Sinusoidal signal
= 600 Hz, Gaussian whi e noise
δ
= 0,005 , he
pa ame e o ca ie signal is no changed. The
alues o he signals in he i s pa s o he es
sequences ( om 0 sec. o 15 sec.) a e he same as
hose in Figu es 1-3.
3.
SIMULATION DESCRIPTION
Tes signals a e modeled in Ma lab as he
sequence o digi al samples, which pass h ough he
compe en ype o communica ion channel (Noise
Channel AWGN and Bina y Syme ic Channel
BSC). The simula ions o si ua ions o ansmission
sequence in gi en ansmission chain a e he ask o
his model. The simula ions a e ealized especially
om he poin o iew o he pa ame e s o he
channels. Main pa ame e s a e E o P obabili y o
BSC channel and pa ame e SNR o AWGN
channel. The model is made o occu many e o s
o ansmission. The e o e I did no use any
channel coding. We need maximum numbe o
e o s, because we wan o ge he mos sensi i e
sequence o hese e o s.
O iginal sou ce and des ina ion ile a e eco ded
by sound ca d in o wa ile. The bo h iles a e
compa ed simul aneously. This compa ison is based
on calcula ion o mean squa e measu e and
co ela ion coe icien .
3.1 De ailed desc ip ion o he simula ion model
This model comes ou om he model desc ibed
in [1]. The model was wide sp ead abou he block
„F om Wa e File“. Simula ions we e done o wo
models. We only p esen one simula ion model o
AWGN channel. The blocks „Ze o-O de Hold 2“,
„Sa u a ion 2“, „Quan ize 2“ a e no used in he
simula ion model o BSC. Ins ead o AWGN
channel is uses BSC channel.
Fig. 4 Simula ion model o AWGN channel
De e mina ion o es sequence o in usi e measu emen o VTQoS …
49
3.2 P inciple o simula ion
The sou ce sequences a e c ea ed using
Sound o ge so wa e. The p inciple o he c ea ion
o sou ce sequences is desc ibed in Chap e 2. The
des ina ion sequences a e c ea ed by simula ion.
The leng h o he simula ion is 90 sec. This leng h
is he same as he leng h o he sou ce sequence.
The leng h o he des ina ion sequence is also 90
sec. wha esul s om he leng h o he simula ion.
The sou ce and he des ina ion sequences a e
compa ed a e inishing he simula ion. The
p inciple o he compa ison is desc ibed by he
ollowing s eps:
1. Reading in sou ce and des ina ion es
sequences
2. Segmen a ion o he es sequences in o n
in e als, each wi h 8000 samples. The
ollowing pa ame e s a e calcula ed in each
in e al:
• Co ela ion coe icien
i
,
• Coe icien s FFT,
• Mean squa e measu e d
i
.
3. Calcula ion o he a e age alues:
n
n
ii

=
=
1
, (1) and n
d
d
n
ii

=
=
1
, (2)
whe e n is he numbe o in e als,
i
is he
co ela ion coe icien o he i- h in e al, d
i
is he
mean squa e measu e o he i- h in e al.
The segmen a ion o he es sequences in n
in e als and calcula ion o ele an pa ame e s in
hese in e als enables o ob ain mo e p ecise
esul s. The calcula ion o he co ela ion
coe icien is ealized by he ollowing ela ion:





−





−
−−
=



m n
mn
m n
mn
m n
mnmn
i
BBAA
BBAA
22
)()(
))((
, (3)
whe e
A
= mean2(A) , and
B
=mean2(B). Ma lab
unc ion mean2 ealize he calcula ion o he mean
alue.
The mean squa e measu e is based on he
spec al compa ison o he es mic osegmen wi h
he e e ence mic osegmen . The mos common
no m is L
2
-no m, which is de ined like as:
( )
2/1
1
2
),( 




−=

=
N
j j ji
yy d
, (4)
whe e N is he numbe o FFT poin s in gi en
mic osegmen , y
j
is he absolu e alue o he j- h
FFT coe icien o he es mic osegmen , y
j
is he
absolu e alue o he j- h FFT coe icien o he
e e ence mic osegmen .
Nowadays, he compa ison is usually done
using he co ela ion coe icien . I we wan o
disco e di e ences in he spec um a ea, ano he
pa ame e is added. Using he co ela ion
coe icien and he mean squa e measu e we may
ind he es sequence ha is he mos sensi i e on
noise in luences, which a ise in communica ion
channels.
4.
PRESENTATION OF RESULTS
These simula ions we e ealized by he usage o
he sequences displayed in uppe pa o documen
o comp essed cha ac e is ics acco ding o he
cu es A and o he bo h ypes o he channels
(AWGN, BSC).
1000
2000
3000
4000
5000
5 10 15 25 50
SNR
d
seq21
seq22
seq23
Fig. 5 G aphical p esen a ion o he esul s o he
simula ions o mean alue o mean squa e measu e
(AWGN channel)
0,7
0,8
0,9
1
5 10 15 25 50
SNR
seq21
seq22
seq23
Fig. 6 G aphical p esen a ion o he esul s o he
simula ions o mean alue o co ela ion coe icien
(AWGN channel)
0
1000
2000
3000
4000
1,E-05 1,E-04 1,E-03 1,E-02 1,E-01
BER
d
seq21
seq22
seq23
Fig. 7 G aphical p esen a ion o he esul s o he
simula ions o mean alue o mean squa e measu e
(BSC)
Ad ances in Elec ical and Elec onic Enginee ing
50
0,85
0,9
0,95
1
1,05
1,E-05 1,E-04 1,E-03 1,E-02 1,E-01
BER
seq21
seq22
seq23
Fig. 8 G aphical p esen a ion o he esul s o he
simula ions o mean alue o co ela ion coe icien
(BSC)
5.
CONCLUSION
The p inciple o es sequences selec ion o
in usi e measu emen o VTQoS is based on a
simple ule. The ule is ha such g oup o es
sequences is used o measu emen which ob ains
he maximum di e ence o alues o he ele an
bounda y coe icien s.
The calcula ions a e ealized
by o mulas 5-8.
I is o BSC channel:
)1,0()00001,0( BERBER
−=
δ
(5)
)00001,0()1,0( BERBERd
dd −=
δ
(6)
I is o AWGN channel:
)5()50( SNRSNR
−=
δ
(7)
)50()5( SNRSNRd
dd −=
δ
(8)
whe e
δ

is he di e ence o a e age alues o
bounda y pa ame e s o co ela ion coe icien ,
δ

d
is
he di e ence o a e age alues o bounda y
pa ame e s o mean squa e measu e,
)50(SNR
is
a e age alue o co ela ion coe icien o ele an
alue o bounda y pa ame e o AWGN channel,
)50( SNR
d
is a e age alue o mean squa e measu e
o ele an alue o bounda y pa ame e o
AWGN channel,
)1,0( BER
is a e age alue o
co ela ion coe icien o compe en alue o
bounda y pa ame e o BSC,
)1,0( BER
d
is a e age
alue o mean squa e measu e o compe en alue
o bounda y pa ame e o BSC. Bounda y
pa ame e s a e SNR o he alue o 50 dB and o
he alue o 5 dB on AWGN channel. I is
pa ame e BER o he alues 0,00001 and 0,1,
when we use BSC channel.
This ac was a basis o de i a ion o his ule
because we need he es sequence which is he
mos sensi i e on all in luences, which can a ise in
eal ne wo ks. We can hypo hesize ha i he
maximum di e ence is eached o same
condi ions, he gi en es sequence is mo e sensi i e
on in e e ence in luences han he o he s. Thus he
es sequence wi h such p ope y is mo e sui able
o measu ing o VTQoS. The
bes es sequence
in
he sense o he mean squa e measu e and he
co ela ion coe icien is chosen om he esul s.
Table 1 Di e ence
δ

o co ela ion coe icien
Table 2 Di e ence
δ

d o mean squa e measu e
δ
d Seq21 Seq22 Seq23
AWGN 2912,60
3014,60
3019,90
BSC 2685,58
2801,15
2800,60
We can see om he esul s ha he in luence o
he en i onmen o ixed elecommunica ion
ne wo k on hese es sequences is app oxima ely
he same. Hence we can use any o hese es
sequences o he in usi e measu emen o VTQoS
in en i onmen o ixed elecommunica ion
ne wo k. We decided ha we used he es sequence
o en i onmen o ixed elecommunica ion
ne wo k ha eached he bes esul s o he
simula ions in he mobile en i onmen . This
decision was in luenced by s onge in luence o
he mobile en i onmen on hese es sequences.
Tes sequence 23 ga e he bes esul s in he mobile
en i onmen in he sense o he mean squa e
measu e and he co ela ion coe icien . The
selec ion o sui able es sequence in mobile
en i onmen is desc ibed in [8]. In he u u e,
con enience o his es sequence o in usi e
measu emen o VTQoS will be e i ied p ac ically
by eal measu emen s in con e gen ne wo k o he
Uni e si y o Žilina.
REFERENCES
[1
] Po a, P., Vaculík M. Me hod o choice o es
signals o au oma ic in usi e measu emen VTQoS,
In
P oceedings o Con e ence MESAQIN 2005,
P ague (Czech epublic), 2005, ISBN 80-01-
03262.
[2] Kenek J., Holub, J. Meení k ali y hlaso ého
p enosu elekomunikaních sí ích, ST
5/2004, 1996, pp.6-8.
[3] Kenek J., Holub, J. Hodnocení hlaso ých
penos elekomunikaních sí ích, ST
6/2001, 1996, pp.3-5.
[4] Ma lab help
[5] Kon i , M. Teó ia oznamo ania, ALFA, 1989,
274 p, ISBN 80-05-00191-6.
[6] Psu ka, J:. Komunikace s poí aem mlu enou
eí, ACADEMIA, 1995, 287 p, ISBN 80-200-
0203-0.
[7] F aneko á, M. Modelo ania komunikaných
sys émo p os edí Ma lab, Simulink
a Communica ions Toolbox, EDIS, 2003,
129 p, ISBN 80-8070-027-3.
[8] Po a, P. Vaculík, M. De e mina ion o es
sequence o in usi e measu emen o VTQoS
in mobile en i onmen , A icle send as an
con ibu ion o he con e ence Pos e 2006.
δ
 Seq21 Seq22 Seq23
AWGN 0,2704
0,2706
0,2705
BSC 0,1406
0,1406
0,1406