255 Mrak Valcic Bonato

8/11/2019 255 Mrak Valcic Bonato

1/12

POMORSTVO ScienticJournal of Maritime Research 26/2(2012) str./pp. 351-362 351

Dr. sc. Zoran Mrak /Ph. D.Sanjin Vali, mag. ing.Dr. sc. Jasminka Bonato/ Ph. D.Sveuilite u Rijeci / University of RijekaPomorski fakultet u Rijeci /

Faculty of Maritime Studies RijekaStudentska 2, 51000 RijekaHrvatska / Croatia

Izvorni znanstveni radOriginal scientic paper

UDK / UDC:621.396.932

656.61

Primljeno /Received:12. listopada 2012. / 12thOctober 2012

Odobreno /Accepted:13. studenoga 2012. / 13thNovember 2012

USPOREDBA KVALITETE SIGNALA POSTOJEEG IALTERNATIVNOG POMORSKOG VHF KOMUNIKACIJSKOG

SUSTAVACOMPARISON OF SIGNAL QUALITY OF PRESENT AND

ALTERNATIVE MARITIME VHF COMMUNICATION SYSTEMS

SAETAK

U radu je provedena analiza postojeeg irokopojasnogi alternativnog uskopojasnog komunikacijskog sustavapomorskih VHF radiotelefonskih komunikacija,zasnovanih na simulacijskom modelu. Alternativnisustav predloen je od strane ITU organizacije kaojedno od moguih rjeenja problema nedostatkakomunikacijskih kanala u pomorskom VHF podruju.Pomou raunalnih simulacijskih modela ispitan jeutjecaj promjene modulacijskog postupka u po-morskim VHF radiotelefonskim komunikacijama nakvalitetu prenesene informacije, uvaavajui sverelevantne tehnike karakteristike pomorskih VHFradiotelefonskih ureaja. Postupak ocjenjivanja kvali-tete govornih signala proveden je uz pomo tehnologije

objektivnog ocjenjivanja kvalitete upotrebom raunal-nog programa koji provodi PESQ (PerceptualEvaluation of Speech Quality Measurement) algoritamime je osigurana usporedivost dobivenih rezultata stradicionalnim subjektivnim nainom ocjenjivanjakvalitete signala. Postupak ocjenjivanja kvaliteteproveden je upotrebom etiri standardizirana uzorkagovornih signala s razliitim utjecajem vanjskihsmetnji na komunikacijski kanal. Analiza dobivenihrezultata potvruje mogunost koritenja alternativnogkomunikacijskog sustava.

Kljune rijei: pomorske VHF komunikacije, usko-pojasna modulacija, kvaliteta signala, PESQ

SUMMARY

This paper presents an analysis, based on the simula-tion model, of the existing broadband and alternativenarrowband maritime VHF radiotelephone communi-cation systems. The alternative system proposed by theITU organization is one of the possible solutions to thelack of communication channels in the maritime VHFbands. The inuence of changes in the modulationprocess on the quality of information in the maritimeVHF radiotelephone communications is examined us-ing a computer simulation model taking into accountall the relevant technical characteristics of maritimeVHF radiotelephone equipment. The process of evalu-ating the quality of speech signals was carried out bymeans of an objective quality assessment technology

using a computer program that implements PESQ al-gorithm which ensures the comparability of obtainedresults with the traditional method of subjective evalu-ation of signal quality. The process of quality assess-ment was conducted using four standardized samplevoice signals with different inuences of external inter-ference on the communication channel. Analysis ofthe results conrms the possibility of using alternative

communication systems.

Key words: maritime VHF communications, narrow-band modulation, signal quality, PESQ


2/12

Z. Mrak, S. Vali, J. Bonato: USPOREDBA KVALITETE SIGNALA POSTOJEEG I ALTERNATIVNOG...

352 POMORSTVO ScienticJournal of Maritime Research 26/2(2012) str./pp. 351-362

1. UVOD

Porast pomorskog prometa potaknut globali-zacijom i posljedinim rastom trgovine roba iputnika, namee traganje za novim tehnikimrjeenjima kojima je mogue prilagoditi susta-

ve pomorskih komunikacija stalno rastuembroju korisnika, te porastu broja poruka i u nji-ma sadranih informacija.

Unato uvoenju satelitskog komunikacij-skog sustava INMARSAT, koji je na sebe preu-zeo znaajan dio komunikacijskih potreba bro-dova u obavljanju komercijalnih komunikacija,zbog ogranienja INMARSAT-a i dalje ne pre-staje potreba za upotrebom terestrikih komu-nikacijskih sustava, naroito VHF-a. Zemlje spoveanim pomorskim prometom ve zahtije-

vaju od meunarodnih organizacija IMO i ITUpoveanje broja komunikacijskih kanala naVHF podruju.

Dosadanja istraivanja vezana uz ekasnijeiskoritavanje frekvencijskog spektra, odnosnouz mogunosti poveavanja broja komunikacij-skih kanala mogu se razdijeliti u dvije skupine.Prva skupina istraivanja odnosi se na smanje-nje irine komunikacijskih kanala primjenomnovih metoda modulacije [8], dok se druga sku-

pina istraivanja bavi mogunostima uvoenjadigitalnog prijenosa analognih informacija radipoveanja broja komunikacijskih kanala [9].

U radu je analizirana i obraena mogua pri-mjena analogne uskopojasne fazne modulacije(engl. Narrow Band Phase Modulation NBPM) u frekvencijskom podruju pomorskihVHF komunikacija. Analiza je provedena krozsljedee faze: izrada simulacijskog modela, pri-

jenos govornog signala simulacijskim modelomi objektivno ocjenjivanje kvalitete govornih si-

gnala uz razliite odnose signala i uma (engl.Signal to Noise Ratio SNR) u komunikacij-skom kanalu.

Za potrebe simulacije komunikacijskih susta-va, koriten je raunalni simulacijski programSystemViewkorporacije ELANIX, koji omogu-ava simuliranje analognih i digitalnih komuni-kacijskih ureaja, teoretskih modela kao i kom-pletnih komunikacijskih sustava. Za potrebeobjektivnog ocjenjivanja kvalitete govornih si-gnala koriten je algoritam (raunalni program)PESQ (Perceptual Evaluation of Speech Quality

Measurement), kojeg su za organizaciju ITUizradile kompanijePsytechnicsi Opticom[14].

1 INTRODUCTION

The increase in maritime trafc driven byglobalization and the resulting growth in tradeof goods and passengers imposes the search fornew technical solutions that can be adapted to

maritime communications systems having aconstant increase in the number of users, andthe number of messages and information theycontain.

Despite the introduction of INMARSAT sat-ellite communications system, which took uponitself a signicant part of the communicationneeds of ships in performing commercial com-munications, due to its limitations, there is stillthe need for the use of terrestrial communica-tion systems, particularly VHF. Countries with

increased vessel trafc have already requestedfor an increase in the number of communica-tion channels in the VHF band from interna-tional organizations IMO and ITU.

Previous research related to the efcient useof the frequency spectrum, i.e. the possibility ofincreasing the number of communication chan-nels can be divided into two groups. The rstgroup of studies relates to the reduction ofcommunications channels bandwidth of usingnew modulation methods [8], while the second

group deals with the possibilities of introducingdigital transmission of analog information inorder to increase the number of communica-tion channels [9].

The paper analyses the possibility of imple-menting the analog narrowband phase modula-tion (NBPM) in maritime VHF communica-tions. The analysis included the following:development of the simulation model, trans-mission of the speech signal through the simu-lation model and an objective evaluation of the

quality of speech signals with different values ofthe signal-to-noise ratio (SNR) in the commu-nication channel.

The computer simulation software System-View from ELANIX corporation was used forthe simulation of communication systems. Thischosen software allows the simulation of analogand digital communication devices, theoreticalmodels and complete communication systems.For the purposes of objective quality assess-ment of speech signals, the authors used the al-

gorithm (computer program) PESQ (Perceptu-al Evaluation of Speech Quality Measurement),developed for the ITU organization by Psytech-nics and Opticom companies[14].


3/12

Z. Mrak, S. Vali, J. Bonato: COMPARISON OF SIGNAL QUALITY OF PRESENT AND ALTERNATIVE...


2. RADIOTELEFONSKEKOMUNIKACIJE U

POMORSKOM VHF PODRUJU

Postojei i alternativni uskopojasni komuni-kacijski sustavi propisani su tehnikim karakte-

ristikama ureaja u preporukama ITU organi-zacije [8] i [10].

Postojee radiotelefonske komunikacije naVHF podruju koriste frekvencijsku modulaci-

ju s akcentuacijom signala informacije i karak-teristikom 6 dB po oktavi, odnosno faznu mo-dulaciju, koja je opisana s maksimalnomdevijacijom frekvencije prijenosnog signala od5 kHz i maksimalnom frekvencijom informacijeod 3 kHz. Ovim parametrima opisan je indeksmodulacije, odnosno maksimalna devijacijafaze u radijanima:

66667,1max

=

=

if

f . (1)

Potrebna irina komunikacijskog kanala izra-unana pomou Carssonovog pravila iznosi:

kHzfB iPM 16)1(2 max =+= . (2)

Alternativni komunikacijski sustav koristiuskopojasnu faznu modulaciju odreenu s

maksimalnom devijacijom frekvencije prijeno-snog signala od 2,5 kHz i maksimalnom fre-kvencijom informacije od 3 kHz. Indeks modu-lacije, odnosno maksimalna devijacija fazeprijenosnog signala u radijanima u ovom slua-

ju iznosi:

833334,0max

=

=

if

f . (3)

Potrebna irina komunikacijskog kanala pre-ma Carssonovom pravilu iznosi:

kHzfB iPM 11)1(2 max =+= . (4)

Kod kutnih modulacija dobivaju se odreenapoboljanja kod odnosa signal-um na prije-mnoj strani, koja su ovisna o indeksu modulaci-

je . Poboljanja su izraena kroz odnos signal-um nakon demodulacije SNR

O(engl. Signal to

Noise Ratio-Output) i odnos signal-um u ko-munikacijskom kanalu SNR

C (engl. Signal to

Noise Ratio-Channel).

Odnos signal-um nakon demodulacije je zafrekvencijski modulirane signale odreen jed-nadbom [1]:

2 RADIOTELEPHONYCOMMUNICATIONS IN THE

MARITIME VHF BAND

The technical characteristics of the equip-ment for the existing and alternative narrow-

band communication systems are regulated byITU through recommendations under [8] and[10].

Existing radiotelephone communications inVHF band use phase modulation, i.e. frequen-cy modulation with pre-emphasis of informa-tion signal by 6dB per octave. The modulationis described with a maximum deviation of carri-er frequency of 5 kHz, and maximum informa-tion frequency of 3 kHz. These parameters de-scribe modulation index, i.e. the maximum

phase deviation in radians:

66667,1max

=

=

if

f . (1)

The required bandwidth of the communica-tion channel is calculated using Carssons rule:

kHzfB iPM 16)1(2 max =+= . (2)

The alternative communication system usesnarrow-band phase modulation with a maxi-mum deviation of a carrier frequency of 2.5kHz and a maximum information frequency of3 kHz. The modulation index or the maximumdeviation of the transmission signal phase, inradians, in this case is:

833334,0max

=

=

if

f . (3)

The necessary bandwidth of the communica-tion channel according to Carssons rule is:

kHzfB iPM 11)1(2 max =+= . (4)

The angle modulation systems offer someimprovements in signal-to-noise ratio at the re-ceiver side, depending on the modulation index. Improvements are expressed as the relationof signal-to-noise ratio after demodulationSNR

O (Signal to Noise Ratio-Output) and the

signal-to-noise in the communication channelSNR

C(Signal to Noise Ratio-Channel).

Signal-to-noise ratio after demodulation forfrequency-modulated signal is determined by

the equation [1]:


4/12



=2

3 2

maxi

FMCO

f

BSNRSNR

+++

.....

321

1

1

32

CCC SNRSNRSNR

. (5)

Navedena jednadba opisuje krivulju, kojaima karakteristian nelinearni dio pri malim

vrijednostima SNRC (do priblino 10 dB) te

linearni dio koji predstavlja prihvatljivi odnosSNR

C.

Temeljem Carsonovog pravila za izraun iri-ne frekvencijskog spektra za kutno moduliranesignale i uz zanemarivanje izraza u zagradi iz

prethodne jednadbe pri velikim vrijednostimaSNR

C, poboljanje za faznomodulirane signale

iznosi:

CPMPMO SNRSNR += )1(

2 , (6)

tj. odnos SNRO/ SNR

Cizraen u decibelima:

[ ])1(log10 2 +=PMPM

C

O

SNR

SNR . (7)

=2

3 2

maxi

FMCO

f

BSNRSNR

+++

.....

321

1

1

32

CCC SNRSNRSNR

. (5)

The above equation describes a curve, whichhas a distinctive non-linear part at low valuesSNR

C (approximately 10 dB), and the linear

part representing an acceptable ratio of SNRC.

Based on the Carssons rule for calculatingthe bandwidth of the frequency spectrum of an-gle-modulated signals and by ignoring the termin brackets in the previous equation at large

values of SNRC, the improvement of phasemodulated signals is:

CPMPMO SNRSNR += )1(

2 , (6)

and the relation SNRO / SNR

C expressed in

decibels:

[ ])1(log10 2 +=PMPM

C

O

SNR

SNR . (7)

FM =2

10 20 30 40 50

10

20

30

40

50

60

70

80

SNRO(dB)

SNRC(dB)

Slika 1.Ovisnost SNROi SNRC uz= 2Figure 1SNROi SNR

Cdependency with = 2

Izvor / Source: Autori /Authors


5/12



Kako je poboljanje u odnosu SNRO/ SNR

C

ovisno o indeksu modulacije i izraunava se

prema izrazu ( ) CO

SNRSNR += 12 , za po-

stojei sustav uz =1,66667 iznosi 7,4 dB, dokza alternativni sustav uz =0,833334 iznosi sve-ga 1,27 dB. Budui da je kod analognih modu-

lacija glavni uzrok degradacije kvalitete komu-nikacija odnos signal/um, kod predloeneuskopojasne modulacije, kvaliteta signala naprijemu bit e loija u odnosu na postojei ko-munikacijski sustav.

Postojei radiotelefonski sustav koristi ko-munikacijske kanale irine 25 kHz, dok bi upo-treba alternativnog sustava koji bi koristio ko-munikacijske kanale irine 12,5 kHz omoguilaudvostruenje broja komunikacijskih kanala.

Kvaliteta signala na prijemu ovisi takoer oselektivnosti prijemnika koja za oba sustavaiznosi 70 dB u odnosu na susjedne kanale.

3. METODE ISPITIVANJAKVALITETE GOVORNIH

KOMUNIKACIJA

U analognim govornim radiokomunikacija-ma do gubitka kvalitete govornog signala (spee-

ch quality) koja se odraava na razumljivost po-ruke (speech intelligibility) dolazi uglavnomutjecajem umova na elektromagnetske valoveu komunikacijskom kanalu kao i u primopre-dajnim ureajima. Navedene pojave mogue jeublaiti odabirom odgovarujue vrste modula-cije te poveanjem izlazne snage predajnika.

Kvaliteta komunikacija je u posljednjih dese-tak godina postala kljuan faktor kod dizajnira-nja komunikacijskih sustava, ali i u njihovoj ek-sploataciji. Za potrebe testiranja kvalitete

govornih (telefonskih) komunikacija ITU je ra-zradio dva naina ispitivanja: subjektivno [13] iobjektivno testiranje [14].

Subjektivno testiranje provodi se anketira-njem 20 do 50 osoba koje ocjenjuju kvalitetuprimljene informacije bez uvida u originalni si-gnal. Postoji niz subjektivnih testova, od kojih

je testiranje jednosmjernih ili naizmjeninodvosmjernih veza (engl. Absolute Category Ra-ting ACR) najee upotrebljavan test, a pro-

vodi se tako da svaka osoba daje svoje miljenje

o kvaliteti signala ocjenom od 1 do 5 (Tablica1), te se statistikom obradom rezultata generi-ra ukupna srednja ocjena kvalitete MOS (engl.

As the improvement in the relation SNRO/

SNRCdepends on the modulation index and it

is calculated according to the formula ( )

CO SNRSNR += 1

2 . The improvementfor the existing system is 7.4 dB, while only1.27 dB for the alternative system. Since in ana-

log modulation, the major cause of degradationof communication quality is the signal/noise ra-tio, the signal quality at the reception for theproposed narrowband modulation will be infe-rior to the existing communication system.

The current system uses a radiotelephonecommunication channel of 25 kHz bandwidth.On the other hand, the use of an alternativesystem with a 12.5 kHz bandwidth would dou-ble the number of communication channels.

Signal quality at the reception also dependson the selectivity of the receiver, which is 70 dBrelative to adjacent channels for both systems.

3 METHODS FOR ASSESSINGVOICE COMMUNICATION

QUALITY

In the analog voice radio communications aloss of quality of speech signals (speech quali-ty), which is reected on the intelligibility of

messages (speech intelligibility) are mostly in-uenced by the noise in the communicationchannel as well as the transceiver device. Thisphenomenon can be mitigated by choosing ap-propriate type of modulation and increasing ofthe transmitter output power.

The quality of communication in the pastdecade has become a key factor in the design ofcommunication systems and their exploitation.For the purpose of testing the quality of voice(telephone) communication, the ITU has de-

veloped two assessment methods: subjective[13] and objective testing [14].

Subjective testing is conducted by surveying20 to 50 persons to evaluate the quality of re-ceived information without access to the origi-nal signal.

There are a number of methods for subjec-tive tests, one of which is ACR (Absolute Cate-gory Rating), the most commonly used for test-ing simplex or half-duplex links. The test iscarried out so that each person gives their opin-

ion on the quality of the signal in scale from 1to 5 (Table 1.). The statistical analysis of the re-sults generate the overall mean score of quality


6/12



Mean Opinion Score MOS). Problemi subjek-tivnog naina testiranja su u potrebnom veli-kom broju ispitanika kako bi se postigla pouz-danost ispitivanja, posebnim prostorijama zatestiranje, neponovljivosti rezultata testiranja isporom odzivu na promjene (poboljanja) u ko-

munikacijskom sustavu.Objektivno se testiranje moe provoditi na

dva naina. Jedan nain, intruzivno testiranje,podrazumijeva pristup i predajnoj i prijemnojstrani veze kako bi se mogao usporediti odasla-ni i primljeni signal. Posljedica intruzivnog te-stiranja je da kanal za vrijeme testiranja nijedostupan korisnicima. Ovakva testiranja, ukoli-ko se provode u vrijeme najveih optereenjamree, mogu dovesti do pogrene (loije) pro-cjene. Intruzivno testiranje koristi se u labora-

torijskim uvjetima.U redovnim uvjetima, za potrebe stalnog

nadzora kvalitete, koriste se stoga neintruzivneparametarske i neparametarske metode kojemjere kvalitetu signala u redovnoj upotrebi. Pa-rametarske metode ne mjere kvalitetu prekogovornog signala nego nadziru kvalitetu prije-nosnog puta, dok metode koje mjere govornisignal predviaju kvalitetu signala na prijemupolazei od poznate kvalitete odaslanog signa-la. U usporedbi s intruzivnim metodama, ove su

metode manje precizne.U ovome radu proveden je postupak objek-

tivnog intruzivnog testiranja kvalitete govornogsignala.

4. PESQ ALGORITAM

Ispitivanje kvalitete govornih signala za po-trebe telefonskih komunikacija s limitiranimfrekvencijskim opsegom signala informacije do

3,1 kHz kao i za ispitivanje razliitih naina ko-diranja informacije za digitalne komunikacijskesustave opisano je od organizacije ITU [14].

MOS (Mean Opinion Score). Problems in sub-jective testing methods are in the required largenumber of participants in order to achieve reli-able results, special facilities for conductingtests, unrepeatable test results and a slow re-sponse to changes (improvements) in commu-

nication systems.Objective testing can be conducted in twoways. The rst way, intrusive testing, includesaccess to both the transmitting and the receiv-ing side of the connection to be able to com-pare the transmitted and received signal. As aconsequence of intrusive testing the channelduring the testing period is not available to us-ers. Such tests, if performed at the largest net-

work load time, can lead to incorrect (poor) es-timates. Intrusive testing is conducted inlaboratories.

Therefore, in the working conditions of thecommunication system, in order to have con-stant quality supervision, nonintrusive paramet-ric and nonparametric methods which measurethe signal quality in regular use are applied.Parametric methods do not measure the qualitythrough speech signal but monitor the qualityof the transmission path, while the method

which measures the speech signal predict thequality of the received signal based on theknown quality of the transmitted signal. These

methods are less accurate if compared to intru-sive methods.

In this work an intrusive objective quality testof voice signal is implemented.

4 PESQ ALGORITHM

The testing of voice signal quality for tele-phone communication with limited frequencyrange of information signals up to 3.1 kHz as

well as the testing of different ways of encodinginformation in digital communication systemsis described by the ITU [14]. The PESQ algo-

Tablica 1.ITU skala za ocjenu kvalitete govornih signalaTable 1The ITU speech quality ratings

Osjet kvalitete /Perception of quality Ocjena /RatingIzvrsna /Excellent 5Dobra / Good 4

Zadovoljavajua /Fair 3Slaba /Poor 2Loa /Bad 1

Izvor / Source: ITU-T Recommendation P.800 [13]


7/12



PESQ algoritam (engl.Perceptual Evaluation ofSpeech Quality PESQ) je objektivna metodakoja se koristi za predvianje subjektivne kvali-tete ispitivanja govornih signala. U provedenimdugogodinjim istraivanjima ustanovljeno jeda ova metoda ispitivanja uspjeno mjeri uin-

ke izoblienja signala i utjecaj uma na infor-maciju kod jednosmjernih komunikacija (sim-pleks). Kod istovremeno dvosmjernihkomunikacija (dupleks) gdje se javlja utjecajparametara koji takoer sudjeluju u formiranjuukupne ocjene kvalitete (gubitak glasnoe, ka-njenje, jeka...) PESQ nije dao zadovoljavajuerezultate. Rezultati PESQ algoritma korelirajuu vrlo visokoj mjeri s ACR testom koji se pro-

vodi subjektivnom metodom.

PESQ algoritam usporeuje originalni signal

s degradiranim signalom koji nastaje prolaskomizvornog signala komunikacijskim sustavom.PESQ algoritam neovisan je od komunikacij-skog sustava te je stoga pogodan za testiranjerazliitih vrsta telefonskih komunikacija. Testi-ranje se provodi s unaprijed pripremljenim re-ferentnim signalom u trajanju od 8 sekundi kojise i koristio za procjenu uspjenosti algoritma.Poruke su na razliitim jezicima i izgovaraju ihrazliiti govornici, a formirane su od reenica smeusobnim pauzama. Preporuka je da se mje-

renje obavlja u vie navrata s razliitim poruka-ma.

Algoritam koristi modeliranje slunog susta-va ovjeka (perceptualno modeliranje) i mode-liranje raspoznavanja kvalitete govornog signa-la (kognitivno modeliranje) te usporeuje

rithm (Perceptual Evaluation of Speech Quality),is an objective method used to predict the sub-

jective quality of voice signals under tests. Inpreviously conducted extensive research it isfound that this test method successfully meas-ures the effects of signal distortion and noise

impact on the information in simplex commu-nication systems. In duplex communication,where an impact of parameters that also con-tribute to total quality scores appears (volumeloss, delay, echo ...) the PESQ has not givensatisfactory results. The results obtained by thePESQ algorithm correlate, to a great extent,

with the results of ACR test carried out by sub-jective method.

The PESQ algorithm compares the originalsignal with the degraded signal generated bypassing the original signal through a communi-cation system. The PESQ algorithm is inde-pendent from the communication system and istherefore suitable for testing various types oftelephone communication. Testing is conduct

with reference signals in duration of 8 secondsprepared in advance, which are also used to es-timate the performance of the algorithm. Mes-sages are in different languages with variousspeakers, and are formed of sentences withpauses between them. It is recommended thatthe tests are performed repeatedly using differ-

ent messages.The algorithm uses the modeling of the hu-

man auditory system (perceptual modeling)and the modeling of the recognition of thespeech signal quality (cognitive modeling), andthen compares the original and degraded sig-

Inputfilter

Time

align andequalise

Auditorytransform

Inputfilter

Disturbanceprocessing

Cognitivemodeling

Identify badintervals

Levelalign

Levelalign

PESQMOS

Systemunder test

Auditorytransform

Referencesignal

Degradedsignal

Re-align bad intervals

Slika 2.Struktura PESQ algoritmaFigure 2The PESQ algoritham structure

Izvor / Source: PESQ: An Introduction [16]


8/12



izvorni i degradirani signal. Rezultat usporedbeje ocjena kvalitete govornog signala odaslanogkomunikacijskim sustavom. Izlazni rezultat al-goritma PESQ-MOS je numerika vrijednost ugranicama od 0,5 do 4,5. Kako je ITU kodocjenjivanja kvalitete informacije predvidio

granice za MOS od 1 do 5, da bi se moglo uspo-reivati rezultate subjektivnog i objektivnogocjenjivanja, razraena je i funkcija za prila-godbu PESQ-MOS u MOS-LQO (engl. MeanOpinion Score Listening Quality Objective)[15]:

6607,4MOS-PESQ4945,11

999,0999,4999,0LQO-MOS

+

+

+=

e

. (8)

Prilagoavanje referentnog i degradiranogsignala obavlja se podeavanjem amplituda si-

gnala (Level alignment) i ltriranjem (Input l-ter). Komunikacijski sustavi unose razliita ka-njenja signala, a za usporeivanje potrebno jereferentni i degradirani signal vremenski uskla-diti (Time align and equalise). Takoer postoji ipovratna veza za dodatno usklaivanje vremen-skih odsjeaka (Re-align bad intervals). Percep-tualno modeliranje algoritam obavlja kroz dvijerazine, i to posebno za referentni, a posebno zadegradirani signal. Prva razina predstavlja opo-naanje reagiranja slunog aparata na zvune

signale i transformiranje ulaznih signala iz vre-menske u frekvencijsku domenu (Auditory tran-sform) te mjerenje jaine pojedinih frekvencij-skih komponenata signala. Druga razinamodeliranja eliminira signale koji nisu znaajniza subjektivni dojam zvunih signala (Time ali-

gn and equalise).

Sljedea funkcija koju algoritam izvodi je us-poredba izlaznih rezultata modeliranja slunogaparata za referentni i degradirani signal, u ko-

joj se mjere razlike izmeu amplituda referen-

tnog signala i degradiranog signala na podaci-ma iz algoritma za perceptualno modeliranje.Razlike izmeu dvaju signala izraavaju se kaosrednja vrijednost smetnje i srednja asimetrika

vrijednost smetnje. Linearnim kombinacijamakoje su optimizirane na velikom broju subjek-tivnih eksperimenata (Cognitive modeling), izovih dvaju podatka se izraunava izlazna vrijed-nost algoritma koja daje predvianje subjektiv-ne kvalitete govornog signala PESQ-MOS.

Raunalni program razvijen od strane kom-

panija Psytechnics i Opticom, koji provodiPESQ algoritam, pokazao je vrlo visoku korela-ciju s paralelno izvedenim subjektivnim ocjenji-

nal. The result of the comparison is the evalua-tion of the quality of the speech signal transmit-ted through a communication system. Anoutput of the algorithm PESQ-MOS is a nu-merical value ranging from -0.5 to 4.5. In orderto be able to compare the results of subjective

and objective assessment of voice quality, theITU has developed a function for adjusting thePESQ-MOS results to t the MOS LQO (MeanOpinion Score Listening Quality Objective) val-ues [15]:

6607,4MOS-PESQ4945,11

999,0999,4999,0LQO-MOS

+

+

+=

e

. (8)

Alignment of reference and degraded signalsis done by adjusting the amplitude of signals(Level alignment) and ltering (Input lter).Due to different signal delay in communica-tions systems, it is necessary to align the refer-ence and degraded signal in time in order tocompare them. (Time align and equalise). Thereis also a feedback for further adjustment intime intervals (Re-align bad intervals). The algo-rithm performs perceptual modeling at two lev-els separately, one for the reference signal andthe other for the degraded signal. The rst levelconsists of the emulation of the hearing systemresponse to sound signals and the transform ofinput signals from the time domain to the fre-

quency domain (Auditory Transform), as well asthe measurement of the strength of the individ-ual frequency components of the signal. Thesecond level of modeling eliminates signals thatare not relevant to the subjective impression ofsound signals (Time align and equalise).

The next function that the algorithm per-forms is a comparison of output results of themodeling of the hearing system for referenceand degraded signals, which measures the dif-ference between the amplitude of the respec-

tive signals on the data from the perceptualmodeling algorithm. The differences betweenthe two signals are expressed as the mean signal

value and the mean value of the asymmetricdisturbances. By using linear combinations op-timized on a large number of subjective experi-ments (Cognitive Modeling), the algorithm out-put value giving the prediction of the subjectivequality of speech signals PESQ-MOS is calcu-lated from these two data.

A computer program developed by the com-

panies Psytechnics and Opticom, which con-Opticom, which con-duct the PESQ algorithm, showed a very highcorrelation with subjective assessments done inparallel. The source code of this computer pro-


9/12



vanjima. Izvorni kd navedenog raunalnogprograma nalazi se u dodatku preporuke ITUorganizacije [14].

5. SIMULACIJSKI MODEL I

ANALIZA REZULTATAUvaavajui tehnike karakteristike ureaja

izraen je simulacijski model komunikacijskogsustava koji se razlikuje za irokopojasnu (zapostojei sustav) i uskopojasnu (za alternativnisustav) modulaciju po karakteristikama modu-latora i irini frekvencijskog spektra kod odre-ivanja selektivnosti prijemnika. U svim osta-lim karakteristikama simulacijski model jeidentian za obje vrste modulacije.

Na postavljenom simulacijskom modelu jenajprije obavljeno testiranje sa sinusnim izvo-rom signala, a potom je uveden govorni signalza potrebe testiranja kvalitete komunikacija. Usimulacijski model nije ukljuena potrebnaosjetljivost prijemnika, jer je zanemaren gubi-tak signala u prijenosnom putu. Kod ispitivanjakvalitete govornih komunikacija uzet je ukupanutjecaj smetnji koje se pojavljuju u komunika-cijskom sustavu, a smetnja je uvedena u modu-lirani signal s razliitim amplitudama signala

uma.Govorni signali kojima je izvedena simulacija i

ispitivanje kvalitete signala na prijemu, odabranisu iz raspoloivih uzoraka koji se nalaze u dodat-ku preporuke ITU organizacije [14]. Odabrana

gram is in the Annex of the ITU recommenda-tion [14].

5 SIMULATION MODEL ANDANALYSIS OF RESULTS

Taking into account the technical character-istics of the equipment the authors developedsimulation model of the communication sys-tem, with different characteristics of modulatorand channel bandwidth for the selectivity of thereceiver for wide-band (the existing system)and narrow-band (for an alternative system)modulation. In all other characteristics the sim-ulation model is identical for both types ofmodulation.

The function of the simulation model wasrst tested with sine signal and then the qualityof communication system was tested with the

voice signal. The simulation model does not in-clude the required receiver sensitivity, becausethe signal loss in the transmission path is ig-nored. When testing the quality of voice com-munications, the total impact of interferenceoccurring in the communication system is tak-en, while the interference is introduced into themodulated signals with different amplitudes ofthe noise signal.

Speech signals for the simulation and testingof the signal quality at the reception were se-lected from the available samples provided bythe Annex of ITU recommendation [14]. Foursamples were chosen, two male and two female

Slika 3.Simulacijski model VHF komunikacijskog sustavaFigure 3Simulation model of the VHF communication system



10/12



su etiri uzorka koji predstavljaju dva muka idva enska glasa na engleskom jeziku. Provede-na je simulacija u kojoj je za svaki odabrani uzo-rak dobiven izlazni govorni signal. Za potrebeispitivanja kvalitete govornih signala provedeno

je est simulacija na svakom odabranom uzorku,

i to posebno za svaku vrstu modulacije. Ukupnoje dakle dobiveno 48 izlaznih signala.

Prvi izlazni signal dobiven je simulacijom ko-munikacijskog sustava u idealnim uvjetima, tj.bez utjecaja signala uma na modulirani signal.Ostalih pet izlaznih signala generirano je uzutjecaj smetnji u komunikacijskom sustavu.Odabrane su vrijednosti jaine signala smetnje,zadane kroz odnos signal/um u komunikacij-skom kanalu SNR

Cod 10, 15, 20, 25 i 30 dB.

Ocjena kvalitete svakog od 48 primljenih si-gnala provedena je pomou raunalnog progra-ma za objektivno ocjenjivanje kvalitete govor-nih signala. Obavljena je usporedba odaslanog(originalnog) i primljenog (degradiranog) go-

vornog signala za svaki uzorak i dobiveni su re-zultati za PESQ-MOS, koji su pretvoreni uMOS-LQO.

Iz rezultata dobivenih simulacijom komuni-kacijskih sustava i metodom objektivnog ocje-njivanja kvalitete govornih signala vidljivo je da

je ocjena kvalitete za pojedini uzorak u ideal-nim uvjetima gotovo identina za obje vrstemodulacije, a neznatno se razlikuje za pojedinigovorni signal.

Ocjena MOS-LQO kree se za idealne uvjeteod 3,2 do 3,7. Razlike u ocjeni nastale su zbograzliitih karakteristika glasova pojedinih go-

vornika (visina glasa...), a takoer i zbog razlii-tog odnosa suglasnika i samoglasnika u uzorci-ma. Prilikom ocjenjivanja kvalitete govornihsignala u idealnim uvjetima u komunikacijskom

kanalu na ocjenu utjeu elementi za ltriranjesignala koji proputaju samo raspon frekvencijaod 300 do 3000 Hz, kao i odreeni gubitak si-gnala koji nastaje zbog dozvoljenog izoblienjasignala kod odreivanja potrebne irine kanalanakon modulacije.

Uvoenjem smetnji u komunikacijski kanalprimjeuje se pad ocjene kvalitete govornih si-gnala uz smanjenje odnosa SNR

C,odnosno po-

veanjem utjecaja smetnji. Postojea modulaci-ja MOS-LQO za pojedine uzorke kree se u

granicama od 2,0 do 3,5 to bi se moglo izrazitikroz osjet kvalitete kao slaba do zadovoljavaju-a/dobra kvaliteta.

voices in the English language. A simulationwas conducted in which, for each selected sam-ple, an output voice signal is obtained. For thepurposes of testing the quality of speech sig-nals, six simulations on each of the selectedsample was carried out, separately for each

type of modulation. A total of 48 output signalswere obtained.

The rst output signal is obtained by the sim-ulation of communication systems in ideal con-ditions, that is, without the inuence of noiseon the modulated signal. The other ve outputsignals are generated with the impact of inter-ference on the communication system. The se-lected values of interference signals are givenby signal/noise ratio in the communicationchannel SNR

Cof 10, 15, 20, 25 and 30 dB.

The rating of the quality for each of the 48received signals was performed using a compu-ter program for an objective evaluation of thequality of voice signal. The comparison of thetransmitted (original) and received (degraded)speech signal is performed for each sample andthe obtained results for the PESQ-MOS areconverted to MOS-LQO.

From the results obtained by the simulationof communication systems and by using objec-tive methods to assess quality of speech signals,

it is evident that the result for each sample inideal conditions is almost identical for bothtypes of modulation and slightly different foreach individual voice signal.

In ideal conditions the MOS-LQO scoreranges from 3.2 to 3.7. Differences in the as-sessment arise from the different characteris-tics of the speaker (pitch ...), and due to differ-ent relationships of consonants and vowels inthe samples. In evaluating the quality of speechsignals in ideal conditions in the communica-

tion channel, the assessment is affected by theelements for ltering signals which attenuatefrequencies outside the 300-3000 Hz range, as

well as the signal loss that occurs due to signaldegradation allowed in determining the re-quired bandwidth of the channel after modula-tion.

With the introduction of noise in the com-munication channel, a decline in the quality as-sessment of speech signals is observed when re-ducing SNR

Ci.e. increasing interference. In the

existing modulation, the MOS-LQO for indi-vidual samples ranges form 2.0 to 3.5 whichcould be expressed as a sense of quality, frompoor to satisfactory/good quality.


11/12



Kod alternativnog komunikacijskog sustavaprimjeuje se pad kvalitete u odnosu na posto-

jei komunikacijski sustav, a ocjena MOS-LQOkree se u granicama od 1,8 do 3,1 to bi sekroz osjet kvalitete moglo ocijeniti kao slaba dozadovoljavajua kvaliteta.

Provedene analize kvalitete govornih signala iusporedba dobivenih rezultata potvruju teoret-ske pretpostavke da smanjenjem indeksa modu-lacije kod kutnih modulacija dolazi do gubitkakvalitete govornih signala. No, alternativni ko-munikacijski sustav kojim bi se omoguilo udvo-struavanje broja kanala na pomorskom VHFpodruju ne iskazuje znaajan gubitak u uspo-redbi s postojeim komunikacijskim sustavom.

6. ZAKLJUAK

U ovome radu provedeno je istraivanje im-plikacija eventualnog suavanja komunikacijskihkanala s obzirom na uinak koji bi takva mjeraimala na kvalitetu govornih komunikacija.

Ispitivanja kvalitete signala provode se pre-ma strogim standardiziranim postupcima ITU-a kako bi se dobili relevantni rezultati koje jemogue meusobno usporeivati. U ovom raduodabrana je PESQ tehnologija, koja predstavljaobjektivnu raunalnu tehnologiju najnovije ge-neracije. Uz pomo raunalnog programa za

objektivno ocjenjivanje kvalitete govornih si-gnala utvreno je da prosjeni gubitak kvaliteteprimljenog signala alternativnog sustava iznosi

In an alternative communication system adecrease in quality compared to the existingcommunication system was observed, and theMOS-LQO score ranges from 1.8 to 3.1 whichcould be expressed as a sense of quality, frompoor to satisfactory quality.

The conducted analysis of voice signal qualityand comparison of obtained results conrm the

theoretical assumption that reducing the modula-tion index for angular modulation leads to a lossin voice signal quality. However, an alternativecommunication system that would allow doublingthe number of channels in the maritime VHFband does not show a signicant loss in compari-son to the existing communication system.

6 CONCLUSION

This paper analyses the implications of apossible narrowing of communication channels

with regard to the impact that such a step wouldhave on the quality of voice communications.

Signal quality tests are conducted understrict standardized ITU procedures, in order toobtain relevant comparable results. In this re-search, the latest generation of computing tech-nology for objective evaluation of speech sig-nals quality, the PESQ technology was selected.By this computer program for the objectiveevaluation of the quality of speech signals, it

had been established that the average loss ofthe receiving signal quality for the alternativesystem is about 0.5 MOS-LQO, which repre-

Tablica 2.Rezultati analize za postojei komunikacijski sustavTable 2Analysis results for the existing communication system

Uzorak /Sample MOS-LQO Osjet kvalitete /Perception of qualitymuki glas 1 /male voice 1 2,3 3,2 slaba do zadovoljavajua/dobra /poor to fair/goodmuki glas 2 /male voice 2 2,0 3,2 slaba do zadovoljavajua/dobra /poor to fair/good

enski glas 1 /female voice 1 2,1 3,5 slaba do zadovoljavajua/dobra /poor to fair/goodenski glas 2 /female voice 2 2,0 3,5 slaba do zadovoljavajua/dobra /poor to fair/good


Tablica 3.Rezultati analize za alternativni komunikacijski sustavTable 3Analysis results for the alternative communication system

Uzorak /Sample MOS-LQO Osjet kvalitete /Perception of qualitymuki glas 1 /male voice 1 2,0 3,1 slaba do zadovoljavajua /poor to fairmuki glas 2 /male voice 2 1,9 2,9 slaba do zadovoljavajua /poor to fair

enski glas 1 /female voice 1 1,8 3,0 slaba do zadovoljavajua /poor to fair

enski glas 2 /female voice 2 1,8 3,0 slaba do zadovoljavajua /poor to fairIzvor / Source: Autori /Authors


12/12


362 POMORSTVO Scientic Journal of Maritime Research 26/2(2012) str /pp 351-362

oko 0,5 MOS-LQO, to predstavlja relativnomali gubitak kvalitete u usporedbi s razinomkvalitete postojeeg sustava. Razlika u ocjena-ma izmeu modela postojeeg sustava i u radupredloenog alternativnog modela u normal-nim uvjetima prostiranja, uz najnii ispitani

iznos odnosa signal/um u komunikacijskomkanalu (SNRC= 10 dB), varira kod modela po-

stojeeg stanja izmeu 2,0 i 2,3, dok se za isturazinu smetnji kod ispitivanja alternativnogmodela dobivaju vrijednosti od 1,8 do 2,0. Unajpovoljnijim uvjetima, s najviim ispitanimiznosom signal/um, koji je blizak najeimstvarnim uvjetima (SNR

C= 30 dB), vrijednosti

za model postojeeg stanja kreu se izmeu 3,2do 3,5, dok se za alternativni model kreu izme-u 2,9 do 3,1. Izraeno kroz osjet kvalitete to je

unutar granica zadovoljavajue kvalitete.Iako je ispitivanjem utvreno da kvaliteta si-gnala na prijemu opada, ona jo uvijek ostajena zadovoljavajuoj razini, te se stoga uskopo-

jasna fazna modulacija moe podrati kao pri-hvatljiv nain ublaavanja problema nedostatkakomunikacijskih kanala na VHF podruju.

sents a relatively small loss in quality if com-pared to the quality of the existing system. Thedifference in scores between the existing systemmodel and the proposed alternative model un-der normal conditions of propagation, with thelowest tested amount of signal/noise ratio in

the communication channel (SNRC = 10 dB),varies for the existing system model between2.0 and 2.3, while the scores for the same levelof noise in the testing of alternative models arebetween 1.8 and 2.0. In the most favorable con-ditions, with the highest tested amount of sig-nal/noise ratio, which is close to the most com-mon real-world conditions (SNR

C= 30 dB), the

values for the model of the existing system isbetween 3.2 and 3.5, while for the alternativemodel, between 2.9 and 3.1. If these values areexpressed by means of the sense of quality, they

are ranked as satisfactory quality.Although the analysis revealed that the qual-

ity of the received signal decreases, it still re-mains at a satisfactory level, and therefore, thenarrowband phase modulation can be support-ed as an acceptable way to mitigate problems ofthe lack of communication channels in theVHF band.

LITERATURA /REFERENCES

[1] Duncan, J.,FM Demodulator Threshold Reduction,Final Report N65-19880 Prepared for NASA, 1964. [2] ITU-R Recomendation BS.1116-1, Methods for the Subjective Assessment of Small Impairments in Audio

Systems Including Multichannel Sound Systems,1997. [3] ITU-R Recomendation BS.1283, Subjective Assessment of Sound Quality, a Guide to Existing Recommendati-

ons, 1997. [4] ITU-R Recomendation BS.1284, Methods for the Subjective Assessment of Sound Quality, General Require-

ments, 1997. [5] ITU-R Recomendation BS.1285,Pre-selection Methods for the Subjective Assessmentof Small Impairments in

Audio Systems, 1997. [6] ITU-R Recomendation BS.562-3, Subjective Assessment of Sound Quality, 1990. [7] ITU-R Recomendation BS.644-1,Audio Quality Parameters for the Performance of a High-quality Sound-pro-

gramme Transmission Chain, 1990.

[8] ITU-R Recomendation M.1084, Interim Solutions for Improved Efciency in the Use of the Band 156-174MHz by Stations in the Maritime Mobile Service, 2001.

[9] ITU-R Recomendation M.1312,A Long Term Solution for Improved Efciency in the Use of the Band 154-174MHz by Stations in the Maritime Mobile Service, 1997.

[10] ITU-R Recomendation M.489, Technical Characteristics of VHF Radiotelephone Equipment Operating in theMaritime Mobile Service in Channels Spaced by 25 kHz, 1995.

[11] ITU-R Recomendation P.341, The Concept of Transmission Loss for Radio Links, 1999.[12] ITU-R Recomendation SM 332, Selectivity of Receivers, 1978.[13] ITU-T Recommendation P.800,Methods for Subjective Determination of Transmission Quality, 1996.[14] ITU-T Recommendation P.862, Perceptual Evaluation of Speech Quality (PESQ), an Objective Method for

End-to-end Speech Quality Assessment of Narrow-band Telephone Networks and Speech codecs, 2001.

[15] ITU-T Recommendation P.862.1,Mapping Function for Transforming P.862 Raw Result Scores to MOS-LQO,2003.[16] PESQ: An Introduction, http://www.sageinst.com/downloads/960B/wp_pesq_introduction.pdf, 3.10.2012.

255 Mrak Valcic Bonato

Documents

Transcript of 255 Mrak Valcic Bonato