11 Psychological measurement for sound description and ... · model have also been proposed...

227

Psychological measurement for sound description and evaluation

Patrick Susini,1 Guillaume Lemaitre,1 and Stephen McAdams2

1InstitutdeRechercheetdeCoordinationAcoustique/MusiqueParis,France2CIRMMT,SchulichSchoolofMusic,McGillUniversityMontréal,Québec,Canada

11.1 Introduction

Severaldomainsofapplicationrequireonetomeasurequantitiesthatarerepresenta-tiveofwhatahumanlistenerperceives.Sound quality evaluation,forinstance,stud-ieshowusersperceivethequalityofthesoundsofindustrialobjects(cars,electricalappliances,electronicdevices,etc.),andestablishesspecificationsforthedesignofthesesounds.Itreferstothefactthatthesoundsproducedbyanobjectorproductarenotonlyevaluatedintermsofannoyanceorpleasantness,butarealsoimportantin people’s interactions with the object. Practitioners of sound quality evaluationthereforeneedmethodstoassessexperimentally,orautomatictoolstopredict,whatusersperceiveandhowtheyevaluatethesounds.Thereareotherapplicationsrequir-ingsuchmeasurement:evaluationofthequalityofaudioalgorithms,management(organization,retrieval)ofsounddatabases,andsoon.Forexample,sound-databaseretrieval systemsoften requiremeasurementsof relevantperceptualqualities; thesearchingprocessisperformedautomaticallyusingsimilaritymetricsbasedonrel-evantdescriptorsstoredasmetadatawiththesoundsinthedatabase.

The“perceptual”qualitiesofthesoundsarecalledtheauditory attributes,whichare defined as percepts that can be ordered on a magnitude scale. Historically,thenotionofauditoryattribute isgrounded in the frameworkofpsychoacoustics.Psychoacoustical researchaims toestablishquantitativerelationshipsbetween thephysical properties of a sound (i.e., the properties measured by the methods andinstrumentsofthenaturalsciences)andtheperceivedpropertiesofthesounds,theauditoryattributes.Thephysicalpropertiesofasoundthatarerelatedtotheauditoryattributescanbecomputed from thesoundsignal.Thesevalues thereforepredicttheauditoryattributesfromthesoundsignalaloneandoncewellunderstoodcanbe substituted for experimental measurements. They are called psychoacoustical

11

Y116134.indb 227 10/10/11 2:12 PM

smc

Typewritten Text

In B. Berglund, G.B. Rossi, J.T. Townsend & L.R. Pendrill (Eds.), Measurement with Persons: Theory, Methods, and Implementation Areas, pp. 227-253. New York: Psychology Press.

228 Measurement with persons: Theory, methods, and implementation areas

descriptors.Psychoacousticalresearchhasisolatedseveralauditoryattributes:loud-ness,pitch,duration,andsharpness,amongothers.Methodshavebeendevelopedtomeasuretheseattributesexperimentally,andalgorithmshavebeendevisedtocom-putecorrespondingpsychoacousticaldescriptors.

Hereweuse the term“auditory attribute” in a slightlybroader sense than thepsychoacousticaldefinition. Indeed, listenerscan recovermanykindsof informa-tionfromasound.Notonlydotheyperceiveperceptsthatcanbedirectlymappedtothephysicalpropertiesofthesound,butmostofthetimetheyalsorecognizethesourcethatcausedthesoundandidentifyitsproperties.Gaver(1993a,1993b)ini-tiallyformalizedthisideabyintroducingtheconceptsofmusical listening(focusonthesounditself)andeveryday listening(focusonthepropertiesofthesource).Bymeasuring auditory attributes,wethereforemeanhere“providingquantitiesrepre-sentativeofwhatauserperceives.”

Thepurposeofthischapteristopresentthemeasurementoftheseauditoryattri-butes fromanappliedperspective.Someof theseattributesareeasilyunderstood(andhaveaname)andhavebeenstudiedindepth.Forinstance,loudness,pitch,andduration are auditory attributes forwhich experimentalmethods, and evenmath-ematicalpredictivemodels,areeasilyaccessible.Section11.1brieflysummarizessomeoftheresultsandmethodsassociatedwiththeseattributes.Otherattributesareless easily specified and often require metaphors from other sensory modalitiestobedescribed:brightness (or sharpness), roughness,fluctuationstrength,andsoon.InSection11.2,wepresentmorespecificallythemethodsusedtoexploretheseattributes.Becausetheycannotbeeasilyandunequivocallyspecifiedtoalistener,theseattributesrequireindirectandmultidimensionalmethodsthatallowexplora-tionofsoundperception.Section11.2presentsseveralfamiliesofmethods:semanticscales,similarityjudgmentsandmultidimensionalscaling,sortingtasks,andclusteranalyses.Section11.3presentsexamplesofapplications insoundquality.Finally,perspectivesintherealmofsonicinteractiondesignarebrieflyintroduced.

11.1 Basic knowledge and methods

11.1.1 Peripheral auditory system

Weprovidehereabroadoverviewof theperipheralauditorysystem.*Foramorecompletedescription,interestedreadersshouldrefertoMoore(2003).

11.1.1.1 Description

Thehumanperipheralauditorysystemiscomposedofthreeparts:theouterear,themiddleear,andtheinnerear.Theouter earismainlycomposedofthepinnaandtheauditorycanalbetweenthepinnaandtheeardrum.Theouterearamplifiesthesoundlevelattheeardrumforfrequenciesaround3kHz.Themiddle ear,composedofthree

* AnimationsbyProf.HerbertHudde fromBochumUniversitycanbe foundat the followingURL:http://www.ruhr-unibochum.de/ika/ika/forschung/gruppe_hudde/bohear_en.htm

Y116134.indb 228 10/10/11 2:12 PM

Psychological measurement for sound description and evaluation 229

verysmallossicles,matchesimpedancebetweentheairintheauditorycanal(outerear)andthefluidsinthecochlea(innerear).Italsoimprovessoundtransmissionforfrequenciesintherangeof0.5–4kHz.Fromapsychoacousticalpointofview,themostimportantpartoftheinner earisthebasilarmembrane(BM)thatcanbeconsideredasa“frequencyanalyzer.”AnincomingsoundsetsinmotiontheBMwithamaxi-mumdisplacementatacertainpositionthatdiffersaccordingtothefrequencyofthesound;thepositionofthemaximumvariesfromthebeginning(base)oftheBM(ovalwindow)forhighfrequenciestotheend(apex)oftheBMforlowfrequencies.ThefrequencyproducingamaximumofdisplacementontheBMisthecenterfrequencyofabandpassfilterforthatposition.Becausedifferentfibersoftheauditorynerveareconnectedtodifferentpositionsalongthebasilarmembrane,thefrequencyselectiv-ityofthebasilarmembraneresultsinafrequencydecompositionofthesoundsintheauditorynerve.Thefrequencyselectivityoftheauditorysystemhasveryimportantconsequencesforaudition.Particularly,the“masking”phenomenonhasintroducedtheconceptsofcriticalbands(CB)andauditoryfiltersandhasresultedinamodelthatisthebasisforthecomputationofpsychoacousticaldescriptors.

11.1.1.2 Masking, critical bands, and models

Fletcher(1940)introducedtheconceptofcriticalbandstoaccountformaskingphe-nomena.Forverynarrowbands,heshowedthatthethresholdofdetectionforapuretoneincreasesasthenoisebandwidthincreases.Afteracertainbandwidth,increas-ingthenoisebandwidthnolongerchangesthetonethreshold.Fletcherassumedthatonlyaneffectivepartofthenoisemasker,closetothefrequencyofthetone,hasthepowertomaskthetone.Thecorrespondingfrequencyregionisthecritical band.Furtherinvestigationsshowedthatamodelconsistingofabankofbandpassfilters,thebandwidthofwhichincreaseswiththecenterfrequency,couldaccountformask-ing (Zwicker, 1961;Zwicker&Fastl, 1972;Moore&Glasberg, 1983,1990).Theshapeofeachfilterisasymmetric:roll-offissharpforfrequenciesbelowthecenterfrequency(100dB/octave)andsmoothforfrequenciesabovethecenterfrequencies.Thesteepnessoftheroll-offdecreasesasthelevelofthestimulusincreases.

Thereareseveralmodelsofthesefilters.Third-octavebandpassfilterscanroughlymodeltheauditoryfilters.Fourth-octavebandpassfiltershavealsobeenproposedandshowntoapproximatefairlywelltheauditoryfiltersexceptforlowfrequencies(Hartmann,1997).Amorecomplexmodeluses theGammatonefilters(Patterson&Holdsworth,1991).Finally,basedonthisconceptofcriticalbands,severalscaleshavebeenproposed:theBark scale(Zwicker&Terhardt,1980)andtheEquivalent Rectangular Bandwidth (ERB) scale(Moore&Glasberg,1983).

11.1.1.3 Psychoacoustical descriptors

Modelsoftheauditorysystembasedoncriticalbandsareusedtocomputepsycho-acousticaldescriptors.TheclassicalpsychoacousticaldescriptorsaresummarizedinZwickerandFastl(1999)andMoore(2003).

Y116134.indb 229 10/10/11 2:12 PM


Thedescriptorofloudnessiswidespread.Modelshavebeenstandardized:ISO532-A (Stevens’ model); ISO 532-B for (Zwicker’s model). A BASIC program isalsoavailableinZwicker(1984).ANSIS3.4-2005isarevisionproposedbyMooreand Glasberg (1996) and Moore, Glasberg, and Baer (1997). Corrections of thismodelhavealsobeenproposedallowingabetteraccountof impulsivesounds inabackgroundmaskingnoise(Vos,1998)andoftime-varyingsounds(Glasberg&Moore,2002).Anotherdescriptorofloudness(Meunier,Boulet,&Rabau,2001)hasbeenproposedforenvironmentalandsynthesizedimpulsivesounds.Theloudnessiswellexplainedbyacombinationbetweenthelogarithmofthereleasetimeandtheenergy.

Psychoacousticaldescriptorscorrespondingtootherauditoryattributesarealsocommonly used: spectral centroid and sharpness, roughness (Daniel and Weber,1997),andsoon(seeZwicker&Fastl,1999,andFastl,1997,forsummaries).Theyhave also been implemented in several commercial software packages: BAS andArtemiSbyHeadAcoustics,dBSonicby01dB-Metravib,PULSEbyBrüel&Kjaer,and LEA by Genesis. The available descriptors that have been implemented arebased on experimental results using abstract sounds, thus these psychoacousticaldescriptorssometimesneedtobeadaptedforrealsounds(seetheworkbyMisdariiset al., 2010, on thisquestion).Only the loudnessdescriptorshavebeen standard-ized.Theyprovidereliableresultsforstationarysounds,butfurtherdevelopmentisneededfornonstationarysounds.

11.1.2 Classical psychoacoustical methods

Thetraditionalpsychoacousticalapproachisunidimensional:itaimstoestablishaquantitativerelationshipbetweenasingleauditoryattributeandaphysicalpropertyofthesound.

11.1.2.1 Indirect methods

11.1.2.1.1 Thresholds. The indirect method is based on the measurement ofthresholds.Theabsolute thresholdistheminimumdetectablelevelofasound.Forinstance, for a pure tone it depends on the frequency of the tone. Under normalconditions,ayounglistenercanhearfrequenciesbetween20Hzand20kHz.Formostadults, the threshold increases rapidlyaboveabout15kHz.Thedifferential thresholdordifference limen(DL)isthesmallestchangeinasoundtoproduceajust-noticeable difference(jnd)intherelatedauditoryattribute.

11.1.2.1.2 Confusion scaling. ByvaryingaphysicalparameterandmeasuringtheDL for a given auditory attribute, a confusion scale for this attribute can be set.AssumingthatallDLscorrespondtoequalchangesoftheauditoryattribute(jnd),Fechner’slaw(1860,publishedinEnglishin1966)canbedetermined:

ψ=klog(ϕ)

Y116134.indb 230 10/10/11 2:12 PM


whereψisthemagnitudeoftheauditoryattribute,ϕisthephysicalparameter,andkisaconstantspecifictoeachauditoryattribute.

11.1.2.2 Direct methods

Ratio scalingisadirectmethodrelyingontheabilityofparticipantstomakenumer-ical judgmentsof theratiobetweenthemagnitudesof theirsensations.Theusualmethodsaremagnitude estimation andproduction.Formagnitudeestimation,theparticipants are required to assign a number proportional to their sensation (e.g.,loudness)oftheintensityforsoundspresentedatdifferentlevels.Forthemagnitudeproductionmethod, theparticipant is required in thiscase toadjust the levelofatestsoundtoaspecifiednumberproportionaltoitsloudness.Therelationbetweentheexpressedsensation(e.g.,loudness)usingsuchmethodsandthecorrespondingacousticalvalues(e.g.,soundpressurelevel)leadstothewell-knownpsychophysicallaw,Steven’slaw:

ψ=kϕα

whereψisthemagnitudeoftheauditoryattribute,ϕisthephysicalparameter,andkandαareconstantsspecifictoeachauditoryattribute. Forinstance,fortheloudnessofa1-kHztone,theexponentis0.6:a10-dBincreaseleadstoa2-soneincrease.Fora3-kHztone,theexponentis0.67.Steven’slawforloudnesshasledtothederivationofthesonescale.

The cross-modal matching methodwasproposedbyS.S.Stevens(1959).Thetaskconsists inmatching twosensations (e.g., loudnessandmuscular forcesensation),oneofwhichhasbeencalibratedbeforehandbyadirectestimationmethod(Stevens,1959).Thematchingfunctionbetween thesensations isknownorexperimentallyobtained.Thenratingsrelated to theothersensationaredirectlydeducedbywayofthematchingfunction.Thismethodcanbeusedtoscaletheloudnessoftime-varyingsounds(seethenextsection).

11.1.3 Perspectives: Loudness of time-varying sounds

Theclassicalpsychoacousticalmethodshavebeenbroadlyused to study theper-ceptionofshortandstationarysounds.Everydaysoundeventsandmusicalpieces,however,areusuallynonstationary.Thetemporalfluctuationsanddurations(upto20minutes)of suchnonstationarysoundsdonotallow theuseofclassicalmeth-ods,butrequirecontinuousratingsofthesounds.Theparticipantmustinthiscaserespondinstantaneouslytoanyvariationofthesound.Themethodsandthedevicesusuallyproposedcanbesortedintofivecategories.

1.The method of continuous judgment using categories was proposed byKuwanoandNamba(1978,1985)withtheaimofstudyingtemporalfluc-tuationsofthelevelofurbansounds.Inthisprocedure,participantsjudge

Y116134.indb 231 10/10/11 2:12 PM


theloudnessateachinstantusingaresponseboxwithsevenbuttonscor-respondingtosevencategories:very,veryloud–veryloud–loud–medium–soft–verysoft–very,verysoft.Thisprocessisapplicabletolong-durationstimuli,becausethetaskisnotdifficultandparticipantsexperiencelittlefatigue.Theparticipantsmodifytheirjudgmentassoonastheyperceiveachangeequivalenttothedistancebetweentwocategories.Themaindisad-vantageofthecontinuouscategoryjudgmentmethodisthatitdoesnotallowonetoobtainanalogicalresponsesasafunctionofthesignalcontour.

2.Theaudiovisual adjustment methodwasdevelopedbyKuwanoandNamba(1990).Inthismethod,participantsexpresstheirjudgmentbycontinuouslyadjusting the length of a line with a cursor so that the length is propor-tionaltotheauditorysensation.Themainproblemwiththismethodcomesfrom the clipping or ceiling effect at the top end of the judgment scale,becausethelengthofthelineislimited(computerscreen,sheetofpaper,etc.).Togetaroundthislimitation,KuwanoandNamba(1990)elaboratedadevicewithwhichthelinepresentedontheterminalscreenisprojectedon a large screenwith anoverheadprojector. In a similarmanner,Fastl(1989,1991)performedanexperimentinwhichtheparticipantjudgedtheinstantaneous loudnessbyassociating in real time thedisplacementof apotentiometeronamixingtable.However,thisdeviceprovideslittlefeed-back(asidefromhand/armposition)totheuser.

3.The continuous cross-modal matching method proposed by Susini,McAdams,andSmith(2002)isbasedonthecross-modalmatchingmethodwithaforce-feedbackdevice.Theparticipanthastoadjustamuscularforcesensationtotheperceivedloudness.Thisdevicewasusedtoassess1-kHzpure tones (Susini, McAdams, & Smith, 2002), urban sound sequences(Susini & Maffiolo, 1999), and sounds of accelerating cars (Susini &McAdams, 2008). The method has proved to be a flexible experimentalprocedureallowinganindividualcalibrationofthedeviceasafunctionofeachparticipant’sperceptualscale,withtheaimofavoidingcompressionorsaturationeffectsintheresponses.

4.Theanalog categorical scalingproposedbyWeber(1991)combinesthecategoricalandanalogicalmethods.Participantscanslideacursorcon-tinuouslyalongfivediscretecategorieslabeled(forexample):veryloud–loud–medium–soft–very soft. The distance between each category isconsideredasequivalent.Thismethodhasbeenwidelyused:forloudnessevaluationofvariable-amplitudesinusoidalsounds(Susini,McAdams,&Smith,2002,2007),forassessingspeechquality(Hansen&Kollmeier,1999; Gros & Chateau, 2001), for assessing the comfort of an urbansequenceofarunningbus(Parizet,Hamzaoui,Segaud,&Koch,2003),andforbrightnessratingsofvarioussounds(Hedberg&Jansson,1998).

5.The semantic scale used in real-time was introduced by several authorsto study more complex auditory attributes than loudness, and more spe-cifically to study real-timeemotional response tomusic.Thecontinuousresponse digital interface (CRDI) developed by Madsen (1996) allows a

Y116134.indb 232 10/10/11 2:12 PM


continuoustrackingoftemporalvariationsofmusicalworks,asdoesthetwo-dimensional emotional space (2DES) proposed by Schubert (1996)with which musically evoked emotions are evaluated in real time in atwo-dimensional semantic space. Several authors used continuous ratingtomeasureemotional force inmusicpieces (Sloboda&Lehmann,2001;McAdams,Vines,Vieillard,Smith,&Reynolds,2004).

11.2 Multidimensional and exploratory methods

Itisnotalwayseasytospecifyanauditoryattributeapriori.Apartfrompitchandloudness,veryfewwordsarespecifictosoundoreasilyunderstoodbynonspecial-ists.Therefore,unidimensionaltechniquessuchasdescribedabovecannotbeusedtomeasureauditoryattributesnoteasilycommunicatedtoparticipants,orthosethataresimplyunknowntotheexperimenter.Thissectionreportsmethodstoexploreormeasureunspecifiedauditoryattributesandmoregenerallytodeterminethepsycho-logicalaspectsofsoundperceivedbylisteners.

11.2.1 Judgments on multiple semantic scales

Theuseofmultiplesemanticscalesisafruitfultechniquetoassessdifferentpsycho-logicalaspectsofsounds:auditoryattributes (e.g., loudness, roughness),appraisal(e.g.,preference),emotionalresponse(e.g.,beauty,arousal),andconnotativedimen-sionsofthesoundsource(e.g.,thepowerofasportscar).

Semanticscalesarecategoryscalesdefinedeitherbyasinglesemanticdescrip-tor(unipolarscale)orbyapairofantonymicdescriptors(bipolarscale).Thescalesusuallyhavebetweenthreeandsevencategories.Itisusuallypreferredtouseanoddnumberofintervalstoincludethemiddlepointofthescale.

11.2.1.1 Method and analysis

Themostusedtechniqueisthesemanticdifferential(SD).Participantsareaskedtojudgeeachstimulusdirectlyalongasetofscaleslabeledwithtwoopposedsemanticdescriptors.Usuallytrueantonymlabelsareused(e.g.,good–bad,pure–rich,etc.),butalternativeshavebeenproposed(e.g.,good–notgood).

Thelabelsofthescalesarecalledsemantic descriptors.Theratingsofastimulusonthedifferentsemanticscalesyieldamultidimensionalrepresentationcalledthesemantic profile;anexampleispresentedinFigure 11.1.Afactoranalysiscancom-binesemanticscalesintomainfactors.Amultipleregressionanalysiscanhighlightrelationshipsbetween factors corresponding tocognitiveaspects (e.g.,preference)andfactorscorrespondingtoauditoryattributes(e.g.,loudness,roughness).Thelat-terfactorsareinterpretedbylookingforacousticalorpsychoacousticaldescriptorsthatarecorrelatedwiththem.Eachsemanticdescriptorishypothesizedtobepsy-chologically relevant to thewholesetof stimuliunderexamination.On theotherhand,ithastobeunderstoodbytheparticipantsoftheexperiment.

Y116134.indb 233 10/10/11 2:12 PM


11.2.1.2 Examples of semantic scales used to describe sounds

SinceSolomon(1958)andvonBismarck(1974),thesemanticdifferentialtechniqueproposed by Osgood (1952) has been widely used in the realm of sound percep-tiontodescribethemultidimensionalcharacterofthetimbreofmusicalinstruments(Wedin&Goude,1972;Pratt&Doak,1976;Kendall&Carterette,1992;Stepánek,2006),environmentalsounds(Björk,1985;Zeitler&Hellbrück,2001),andsoundproducts, such as cars, vacuum cleaners, air conditioning noises, or refrigerators(Chouard&Hempel;1999;Kyncl&Jiricek,2001;Siekiersky,Derquenne,&Martin,2001;Jeon,2006).

Typically,resultsfromthedifferentstudieshaveshownthatthesetofsemanticdifferentialscanbecombinedintothreeorfourmainfactorsthataccountforagreatdealofthevarianceinthejudgments.Forinstance,invonBismarck’sstudyonthetimbreofsyntheticsounds, results revealedfour independentscales:“dull–sharp”(44% of the variance explained), “compact–scattered” (26%), “full–empty” (9%),and“colorful–colorless”(2%).Onlythescalesreferringtosharpnesswereconsid-eredascandidates foragenerallyusablescale for themeasurementof timbre. InPrattandDoak(1976),threescales(“dull–bright”,“pure–rich”,“cold–warm”)wereselectedtobethemoresignificantdescriptorsforinstrumentaltimbres.Insummary,resultsfromdifferentstudiesrevealedthatdescriptorsrelatedtosharpness(“sharp,”

Intrusive

Dully

Bright

Speed

Rough

Hard

Vol

Fluct

Hum_L

Hum_P

Whisp_L

Whisp_P

Figure 11.1 (See color insert.) Sensory profiles obtained for three air-conditioningnoisesfromSiekierski,Derquenne,andMartin(2001).Labelsof thesemantic descriptorsare Intrusive, Dully, Brightness, Speed, Roughness, Hardness, Voluminous, Fluctuation,Humming,Whispering.ThelettersLandPcorrespondtotheLevelandPitch,respectively,ofthewhispering(noise)partandthehumming(motor)part.

Y116134.indb 234 10/10/11 2:12 PM


“bright,”“metallic,”orattheopposite,“dull,”“muffled,”“round”)areappropriatetodescribethemostsalientaspectoftimbre.

Kuwano and Namba (2001) report three main factors (“powerful,” “metallic,”and“pleasant”) thathadconsistentlybeenextracted inmostof their formerstud-iesofsoundquality.Furthermore,thesemanticdescriptor“powerful”wasusuallywell correlated with computed loudness (Zwicker’s loudness level based on ISO532B),and thesemanticdescriptor“metallic”waswellcorrelatedwithcomputedsharpness. The “pleasant” factor was related to cognitive and cultural factors aswellastophysicalpropertiesofsounds.InZeitlerandHellbrück’sstudy(2001)onenvironmental sounds, four factorswere linked, respectively, to a hedonic aspect(“ugly”–“beautiful”),timbre(“dark–light”),power(“weak–strong”),andrapidtem-poral variations (“unstable–stable”). The three latter factors were well correlatedwiththreecalculateddescriptors,sharpness,loudness,androughness,respectively.Resultsfromotherstudiesonsounds(speechorsonarsounds)arequitesimilar:themostimportantfactorsweregenerallyinterpretedasrepresentingloudness,timbre(sharpness)orpitch,andanoverallsubjectiveimpression.

11.2.1.3 Prerequisites to use semantic scales

11.2.1.3.1 Controlling loudness, pitch, and duration. First, acoustical param-eterssuchasloudnessandpitch,aswellasvariationsovertime,stronglyaffecttheperceptionoftimbre.Tostudyauditoryattributesindependentlyofthoseobviousparameters, it is thereforerecommendedtocontrol themandtousesteady-statesounds,equalizedinloudness,pitch*andduration.Thisstatementisinagreementwith the current ANSI definition and summarized by Krumhansl (1989, p. 44):timbre is“thewayinwhichmusicalsoundsdifferonce theyhavebeenequatedfor pitch, loudness and duration.” Otherwise, it is recommended to ask partici-pantstoignoretheseparameters,followingtheproposalbyPrattandDoak(1976),whodefinetimbreas“thatattributeofauditorysensationwherebyalistenercanjudgethattwosoundsaredissimilarusinganycriteriaotherthanpitch,loudnessorduration.”

11.2.1.3.2 Selecting an appropriate number of semantic scales. Second,arestrictednumberofsemanticpairssuitablefordescribingtimbrehavetobeselected.Indeed,the preselection of semantic descriptors by the experimenter may strongly affecttheresults,forthesedescriptorsmaynotnecessarilyconformwiththoseapartici-pantwouldusespontaneously.Forinstance,PrattandDoak(1976)investigated(bya questionnaire) what were the most appropriate adjectives for describing timbreofinstrumentsamongalistof19commonlyusedterms.Sevenwordsemergedasfavorites:rich,mellow,colorful,brilliant,penetrating,bright,andwarm.Similarly,

* Forpitch,thiscanbedoneasforloudnessbyanadjustmentprocedureusingthereal-timeSuperVPsoftwareprogrambasedonthephasevocodertechnique;itisthenpossibletotranspose,stretch,orshortensoundsinreal-time.

Y116134.indb 235 10/10/11 2:12 PM


invonBismark(1974),participantswereaskedtorateeachof69semanticscalesintermsoftheirsuitabilityfordescribingtimbre.Finally,28scaleswereconsideredasrepresentative.However,itshouldbenotedthatinvonBismarck’sstudy,the69scaleswereratedindependentlyofthesoundselectedforthestudyandthusmaynotberelevantfordescribingtheperceptualdimensionsofthesesounds.

11.2.1.3.3 Selecting relevant descriptors. Thethirdprerequisiteconsistsinaskingparticipantstojudgetherelevanceofthesemanticdescriptorsconcerningthesoundsusedinthestudy.Faure(2010)gatheredasetofsemanticdescriptorsfromafreever-balizationexperimenton12musicalsounds.Theywereusedtobuildsemanticscales.Therelevanceofthesescaleswasthenjudgedforthesamesetofsounds.Comparisonbetweentherelevancejudgmentsofthescalesandthevocabularyproducedsponta-neouslyshowedthatseveralsemanticdescriptorsthatwerespontaneouslyproduced(suchas“strong”,“loud”,etc.)werenotconsideredasrelevantwhenpresentedwiththescales,evenbytheparticipantswhoproducedthem.Inversely,severalsemanticdescriptorsthatwererarelyusedspontaneouslywerejudgedtobegloballyrelevantby the majority of participants (e.g., “soft,” “muffled/dull sounding,” “metallic,”“nasal”).Inanotherstudy(Kyncl&Jiricek,2001),participantsfreelydescribedsixvacuumcleanersounds.Amongthe33pairsofsemanticoppositionsobtainedfromthevocabularyspontaneouslyproduced,only5wereconsistentlyjudgedasrelevantfordescribingthesounds(“fuzziness,”“atypicality,”“inefficiency,”“loudness,”and“pleasantness”).ThesestudieshighlighttheimportanceofjudgingtherelevanceofdescriptorsusedinSD-scalesforaspecificcorpusofsounds.

11.2.1.3.4 Defining the meaning of the scales. The fourthprerequisiteconcernsthedefinitionofthescales.Indeed,itiscrucialthattheparticipantscorrectlyunder-standthemeaningofthelabels.Forinstance,inFaure(2010),thestimuliwereequal-izedinloudness.Surprisingly,theparticipantsspontaneouslyusedtheword“loud”to describe the sounds. Actually, the participants’ comments revealed that theyused“loud”todescribedifferentperceptions:“stronginsonicpresence,theattack,”“evokesthepowerandpersistenceofthesound.”Similarly,invonBismarck(1974),althoughthesoundswereequalizedinloudness,participantsusedthescale“soft–loud”todescribeattributesotherthanloudness,suchas“unpleasant.”Therefore,theexperimentmustclearlydefinethemeaningofthesemanticscalestoeliminateanyriskofsemanticambiguity.Presentingtheminasentencecanhelpdefinethemean-ingofthedescriptors.ForinstanceParizetandNosulenko(1999)showedthatratingsofinternalnoisesofvehiclesweremorereliablewhenthesemanticdescriptorswerepresentedinasentencethanwhenpresentedinisolation.Susini,Houix,Misdariis,Smith,andLanglois(2009)introducedthesemanticdescriptor“loud”bythesen-tence “The TV is too loud, we can’t have a discussion.” This sentence aimed atclearlyindicatingthat“loud”referredtothesoundlevelandnottheunpleasantness.

Inadditiontotheseveralprerequisitespresentedabove,otherrecommendationsshouldbetakenintoconsiderationwhenusingtheSD-scaletechniquetorateacor-pusofsounds.

Y116134.indb 236 10/10/11 2:12 PM


• Several studieshave shown that subjects feel uncertain ingiving ratingsunlesstheycanreferthemtothewholesampleofsounds.Thustheentirerangeofsoundshastobepresentedbeforethemainexperiment,andpar-ticipantsmustbeinstructedtousethefullrangeofthescale.Inaddition,itisrecommendedthattherangeofsensitivitycorrespondingtoeachseman-ticdescriptoroftheselectedsetofsoundsbebroadenough.

• Manystudiesontimbrehaveusedthetraditionalsemanticdifferentialpara-digm (e.g., dull–sharp).Bipolar adjectivepairs raise thequestionofpre-sentingtherightantonymlabels(isdulltheoppositeofsharpwhenusedtodescribesounds?).InChouardandHempel(1999),clearantonymswerefoundinabout23%ofthecasesforalistof242adjectivesproducedbytheparticipantstodescribeinteriorcarsounds.Thusanimportantproblemintheuseofbipolaroppositesisthatthe“opposite”issometimesunobtainableornotalwaysagoodantipode.Tosolvethisproblem,KendallandCarterette(1992)proposedusingascaleboundedbyanattributeanditsnegation(e.g.,sharp–notsharp)toratethetimbreofmusicalsounds.Theauthorstermedthismethodverbal attribute magnitude estimation(VAME),becausethetaskfor theparticipant is torate thedegree towhichanattribute ispos-sessedbyastimulus.

• Finally,werecommendpresentingthewholesetofsoundsforeachseman-tic descriptor instead of the classical way consisting in presenting onesoundtotheparticipant,whohastoevaluateitonthewholesetofsemanticdescriptors.InastudybyParizetandcolleagues(1999,2005),thecompari-sonofthetwomethodsshowedthattheformerprovedtobemoreaccurateandwithashorterdurationthantheclassicalone,becauselistenerswerefocusedononesemanticdescriptoratatimewhilehearinganewstimulus.Inaddition,tomeasuresubjectreliabilityoraccuracy,randompresentationofthestimulicanberepeated.Cross-correlationcoefficientsarecalculatedbetweenthedatafrombothpresentationsoftherepeatedstimulitocom-putesubjectreliability.

11.2.2 Dissimilarity judgments and multidimensional scaling technique

Semanticscalescomparestimulialongdimensionsdirectlydescribedsemantically.Itisthereforepossibletoassessvariouspsychologicalaspectsofacorpusofsounds,rangingfromelementaryauditoryattributestocognitiveandemotionalaspects.Thedisadvantageisthatthenumberofscalesisoftentoohighand,withtheexceptionofafewstudiesmentionedintheprevioussection,someoftheselectedsemanticdescriptorsarenotperceptuallyrelevant to thecorpusstudiedandaresometimesredundantinrelationtoeachother.However,thisapproachisappropriatetostudytheperceptionofvariousenvironmentalsounds,aslongasseveralprerequisitesaretakenintoaccount.

Incontrast,themultidimensionalscalingtechnique(MDS)isbasedondissimi-larityratingsandthusdoesnotrequireaprioriassumptionsconcerningthenumber

Y116134.indb 237 10/10/11 2:12 PM


ofperceptualdimensionsortheirnature,unlikethemethodsthatuseratingsalongspecifieddimensions.

11.2.2.1 MDS and auditory perception

The multidimensional scaling technique is a fruitful tool for studying perceptualrelationsamongstimuliand foranalyzing theunderlyingauditoryattributesusedbytheparticipantstoratetheperceivedsimilaritybetweentwosounds.MDSrep-resents theperceivedsimilarities ina low-dimensionalEuclideanspace(so-calledperceptual space), so that the distances among the stimuli reflect the perceivedsimilarities (seeMcAdams,Winsberg,Donnadieu,Soete,&Krimphoff,1995, forareviewofthedifferentMDSalgorithms).Eachdimensionofthespace(so-calledperceptual dimension)isassumedtocorrespondtoaperceptualcontinuumthatiscommontothewholesetofsounds.Itisalsoassumedthateachdimensioncanbewellexplainedbyanacousticparameterorapsychoacousticaldescriptor.Inotherwords,theMDStechniqueisappropriatefordescribingsoundsthatarecomparablealongcontinuousauditoryattributes,whichmeansthatitisappropriateforstudyinghomogeneouscorporaofsounds,thatis,thosemadeofsoundsproducedbythesametypeofsource.


Participantsratetheperceiveddissimilaritybetweeneachpairofsoundsundercon-sideration, that is,N(N–1)/2 ratings forN stimuli,onacontinuousscale labeled“VerySimilar”attheleftendand“VeryDissimilar”attherightend.Then,thedis-similaritiesaremodeledasdistancesinaEuclideanspaceofRdimensionsexpectedtobethemostrelevantperceptualdimensionssharedbythesounds.Inthepercep-tual space, a large dissimilarity is represented by a large distance. The final andthemostdifficultpartof thisapproach lies inmatchingperceptualdimensions toacousticalorpsychoacousticaldescriptors.

11.2.2.3 Example of MDS studies to describe timbre of musical sounds

ManyresearchershaveappliedtheMDStechniquetocharacterize theperceptualdimensionsofsounds,sincetheseminalstudiesbyPeters(1960)andPlomp(1970).Peters(1960)startedtoapplytheMDStechniquetoacorpusofsoundswithaknowndimensionality(16puretonescomposedof4frequenciesat4soundpressurelevels:theacousticaldimensionalityistherefore2).Theanalysisofthedissimilarityjudg-mentsfrom39participantssuccessfullyhighlightedthetwoexpectedauditoryattri-butes:pitchandloudness.HethereforeconcludedthattheMDStechniquemightbeusefultoexploresetsofsoundstheauditoryattributesofwhichwouldbeunknown.To test this idea, he applied the technique to other corpora of sounds, for whichthesalientauditoryattributeswereunknown(syntheticcomplexsoundsandspeechsounds).Theresultswerelesseasilyinterpretable(hefoundbetweenthreeandsixdimensionsforthecomplexsounds).Butcomparedtowhatheobtainedwithmore

Y116134.indb 238 10/10/11 2:12 PM


traditionalapproaches(freeverbaldescription,partitionscaling,magnitudeestima-tion), he concluded that “the most promising approach for the isolation and defi-nitionofperceptualdimensionsofcomplexsoundswas theMDSmodel” (p.52).Plomp(1970)appliedMDStosetsofmusicalsounds,whichyieldedthreeorthogo-naldimensions.

Since then, several psychoacoustical studies using MDS have shown clearlythat musical timbre is a multidimensional attribute. Grey (1977) identified threesalient dimensions shared by a corpus of musical sounds. Using a refinement oftheclassicalMDStechnique(EXSCAL,developedbyWinsberg&Carroll,1989),Krumhansl(1989)alsofoundaspacewiththreedimensionssharedbyacorpusofsynthesizedmusical sounds (winds, bowed string, plucked strings,mallet percus-sion).ThesamesetofsoundswasanalyzedbyMcAdams,Winsberg,Donnadieu,Soete,andKrimphoff(1995),whoalsofounda3-Dspace.Thefirstdimensionoftheperceptualspacewascorrelatedwiththecentroidoftheamplitudespectrum.Ithasgenerallybeen reported tocorrespond to the semanticdescriptors“metallic,”“sharp,”or“brilliant.”Theseconddimensionwascorrelatedwiththelogarithmoftheattacktimeoftheamplitudeenvelope,andcorrespondstothesemanticdescrip-tors“fast-slowattack,”“resonant,”or“dry.”Thethirddimensionwascorrelatedwiththespectral irregularity(logarithmof thespectraldeviationofcomponentampli-tudesfromaglobalspectralenvelopederivedfromarunningmeanoftheamplitudesofthreeadjacentharmonics)orthespectralflux(averageofthecorrelationsbetweenamplitudespectrainadjacenttimewindows).

11.2.2.4 Prerequisites for using MDS to study auditory perception

11.2.2.4.1 Controlling loudness, pitch, and duration. Itisimportanttoemphasizethatthemusicalsoundsusedinthestudiespreviouslymentionedwereequalizedinpitch,subjectiveduration,andloudness,sothatratingswouldonlyconcernthedif-ferencesintimbre.Indeed,certainauditoryattributes,suchasloudness,mightdomi-nateandoverpowerlesssalientones,asmentionedinSection11.2.1.3forsemanticscales.Twosoundsthatdiffermainlyintermsofloudnesswillbejudgedobviouslydifferentaccordingtothisdimension,withlittlecontributionfromotherdimensionsofvariationbeingtakenintoaccount.

11.2.2.4.2 Selecting a homogeneous corpus of sounds. Asmentionedearlier,MDSishypothesizedtorepresentacorpusofsoundsbyalimitednumberofcontinuousauditorydimensionsthatarecommontoallthesounds.Thatmeansthecorpushastobecomposedofhomogeneoussoundobjects(soundsproducedbythesametypeofobjectorstimulithatsoundrathersimilar,e.g.,aclassofcarsounds)inordertoavoidaperceptualstructurethatisstronglycategoricalforwhichtheMDSapproachisnotadapted(seenextsection).Aclusteranalysisonthesimilarityratingscanrevealthedegreeofhomogeneityofthesoundcorpus.Ifthetreestructureobtainedrevealsastrongcategorizationofthecorpus,itisadvisabletodeterminewhichcategoriesbestrepresenttheobjectivesofthestudyinordertoobtainappropriatestimuli.

Y116134.indb 239 10/10/11 2:12 PM


11.2.2.4.3 Limiting the number of sounds. Asparticipantsmaybecomefatiguedor losemotivationover time, applicationof theMDS technique is restricted to arathersmallnumberofsounds(moreor less20well-chosensounds),because thenumberofpairs(N(N–1)/2)growsrapidlywiththenumberofsounds(N).Thusapreliminarycategorizationexperimentmaybeadvisableinordertoselectthemostrepresentativesounds(seeSusini,McAdams,Winsberg,Perry,Vieillard,&Rodet,2004).Anotherpossibilitytoavoidbeingconfinedtoasmallnumberofstimuliistousesortingtasks.Indeed,thevalidityofusingsortingtasksforsoundsinsteadofpairedcomparisonshasbeentestedandshowntobeeffectivewithtwodifferentsetsofauditorystimuli(Bonebright,1996).However,furthertestshavetobeperformedinordertoconfirmthevalidityforcollectingdatausingsortingtasks.

11.2.2.4.4 Collecting information from participants. Once the perceptual con-figuration is obtained, it is important to identify the perceptual meaning of eachdimensionoreventolabelthedimensionsusingsemanticdescriptors,andalso,togiveaphysicalinterpretationbyestablishingsystematicrelationsbetweenthestimu-luscharacteristicsandtheirlocationsinthespace.Knowledgeandfamiliaritywiththesoundcorpusandperceptuallyrelevantacousticparametersarethusnecessaryinordertocharacterizethedimensionsofthespaceobjectively.Anotheroptionistodirectlyasktheparticipantstodescribewhichsensationtheyattendedtowhilejudgingthedissimilarities.

11.2.3 Sorting tasks

TheMDS technique isnot appropriate for setsof sounds causedbyverydiffer-ent and obviously identified sources. For instance, Susini, Misdariis, McAdams,& Winsberg (1998) applied an MDS analysis to an extremely heterogeneous setofenvironmentalsounds(trains,cars,andplanes).Theanalysisyieldedastronglycategoricalperceptualstructure:listenersidentifiedthesoundsourcesratherthancomparingthemalongcontinuousdimensions.Therefore,thispredominantcogni-tivefactor—recognition,classification,andidentificationofthesoundsource(seeMcAdams,1993)—violated theassumptionofunderlyingcontinuousdimensionsrequiredby theMDStechnique. In thiscase,otherexperimentalapproachesareneededand,particularly,thesortingtasks.

11.2.3.1 Sorting task, categorization, and auditory cognition

Sortingtasksareverycommonlyusedincognitivepsychologytoaddresstheques-tions of identification and categorization of sound sources. These questions aretightlybound: identifying thesourceofsoundcanbeviewedasconnectingaudi-toryperceptiontoconcepts,andconceptstolanguage,inabidirectionalrelationship(McAdams,1993;Goldstone&Kersten,2003).Severalapproachestotheorganiza-tionandprocessingofconceptsandcategorieshavebeendeveloped(seeGoldstone&Kersten,2003,orKomatsu,1992forareview).Beforepresentingthetechnicalprocedureofsortingtasks,webrieflyrecallthegeneralprinciplesoftheprototypical

Y116134.indb 240 10/10/11 2:12 PM


approachtocategorizationdevelopedbyRosch(1978),whichisveryoftenusedasanunderlyingframeworkinsortingtasks.Thisapproachisbasedonthenotionofsimilarityandisthereforewelladaptedtoaccountforperceptualconceptssuchasthoseusedtodescribesounds.

Rosch’sapproachtocategorizationreliesontwoprinciples.Firstcategorizationisbasedonthecognitive economyprinciple:categoriesalloworganismstohandletheinfinitenumberofstimulibytreatingthemasequivalentwhenthedifferentia-tionisirrelevantforthepurposeathand.Thesecondprincipleisthatthe world has structure.Categorizationoftheworldisthusnotarbitrary,butreliesonitsperceivedstructure(Rosch,Mervis,Gray,Johnson,&Boyes-Braem,1976).

TwoconceptsareoftenborrowedfromRosch’swork:first,Roschandhercol-leagueshaveexperimentallyidentifiedthreelevelsintaxonomiesofobjects:

• Thebaselevel:itemsinthesecategoriessharemanyelementsincommon.• Thesuperordinatelevel:thislevelismoreinclusivethanthebaselevel,but

itemsinthecategoriesatthislevelsharefewerelementsincommon.• Thesubordinatelevel:itemsinthecategoriesatthislevelsharemanyele-

mentsincommon,buttheclassesarelessinclusive.

Second,Roschhasintroducedthenotionofanalogcategorymembership:catego-riesareinternallystructuredintoaprototypeandnonprototypemembers.Fortheselattermembers,thereisagradientofcategory membership(Roschetal.,1978).


Inasorting task,listenersarerequiredtosortasetofsoundsandtogroupthemintoclasses.Whentheexperimenterdoesnotspecifyanyspecificcriteriathatthelisten-ershavetouse,thetaskiscalleda free-sorting task.Usually,thelistenersarealsorequiredtoindicatethemeaningofeachclass.*Sometimes,thelistenersalsohavetoselectaprototypeineachcategory(themostrepresentativemember).

Technically,becausepersonalcomputersarewidespreadinthelab,† theproce-dureamountstoprovidingthelistenerswithaninterfaceallowingthemtolistentothesoundsbyclickingoniconsandmovingtheiconssoastoformgroups.

Toanalyzetheresults,thepartitionofthesoundscreatedbyeachlisteneriscodedinanincidence matrix(inthematrix,0indicatesthattwosoundswereinseparategroupsand1 that theywere in the samegroup).Aco-occurrencematrix is thenobtainedbysummingtheincidencematrices,whichcanbeinterpretedasaproxim-itymatrix(Kruskal&Wish,1978).Therefore,aswithdissimilarityratings,sortingtasks result inestimating similaritiesbetween the sounds.However, the structureofthesedatamightbedifferentdependingontheprocedure.Forinstance,Aldrich,

* Aclassificationofthesoundsistheresultofasortingtask.“Categoriesareequivalenceclassesofdifferent(i.e.,discriminable)entitiesandcategorizationistheabilitytoformsuchcategoriesandtreatdiscriminableentitiesasmembersofanequivalenceclass”(Sloutsky,2003,p.246).

† Thingswererathermorecomplicatedwithoutcomputers;seeVanderveer(1979).

Y116134.indb 241 10/10/11 2:12 PM


Hellier,andEdworthy(2009)showedthatdissimilarityratingsencouragedpartici-pants touseacoustical information,whereasafree-sortingprocedureemphasizedcategorical information. Different techniques are available to visualize the prox-imitydata.Whenthedatafollowthetriangularinequality,butnottheultrametricinequality,*theyarebestrepresentedinalow-dimensionalgeometricalspace(e.g.,byusingMDS).Whentheyalsofollowtheultrametricinequality,theyarebestrep-resentedinatreerepresentation(Legendre&Legendre,1998).Clusteranalysescre-atesuchrepresentations.Themostpopulartreerepresentationis thedendrogram.Itconsistsinrepresentingthedatainahierarchicaltree.Insuchatree,theleavesrepresentthesounds,andtheheightofthenodethatlinkstwoleavesrepresentsthedistancebetweenthetwosounds.Therepresentationishierarchical,wellsuitedtorepresentclassinclusion,andthereforefitswellwithRosch’sframework.

11.2.3.3 Examples of urban soundscape categorization

Sortingtaskshavebeenlargelyusedtostudythecategorizationofeverydaysoundsand soundscapes† (Guyot, 1996;Guyot,Castellengo,&Fabre, 1997;Vogel, 1999;Maffiolo, Dubois, David, Castellengo, & Polack, 1998; Guastavino, 2007; seeSchulte-Fortkamp&Dubois,2006,forareviewofrecentadvances).

Morerecently,Tardieu,Susini,Poisson,Lazareff,andMcAdams(2008)conductedanexperiment thataimed tohighlight thedifferent typesofauditory informationthatareperceivedinthesoundscapesoftrainstations.Thegoalwasalsotodeter-minetheinformationthatparticipantsusedintherecognitionofthespacetypology.Sixty-sixsoundscapesampleswerepresentedtoparticipantsinafree-categorizationtaskwithverbalization.Theresultsshowedthatthelistenersgroupedtogetherthesamplesintoeightglobalcategories.Furtheranalysisaimedtoexplainthecategoriesonthebasisofthefreeverbalizations.Eachverbalizationwasreducedtothewordsthatcontainedadescriptivemeaning.Forexample, the text“Ihavegroupedherethesequencesthattookplaceinaticketoffice.Weclearlyhearpeopletalkingaboutpriceand ticket” is reduced to thewords“ticketoffice,clearlyhear,people, talk-ingaboutprice.”ThisreductionwasmadewiththehelpofthesoftwareLEXICO(2003),whichautomaticallycountseverywordinatext.Then,wordsaregroupedinto semantic fields that are deduced from theverbal descriptions. Five semanticfields were deduced (Figure 11.2): sound sources (e.g., trains, departure boards,ticket-punchingmachines,whistle,etc.),humanactivities(e.g.,conversations,steps,

* ThetriangularinequalitystatesthatforanythreepointsA,B,andC,d(A,C)≥d(A,B)+d(B,C),wheredisthedistancebetweenthetwopoints.InaEuclideanspace,thelengthofanysideofatrianglecan-notbegreaterthanthesumoftheothertwosides.Inanultrametricspace,thisinequalityisreplacedbyd(A,C)≤max{d(A,B),d(B,C)}.Inthiskindofspace,anygivensidemustbelessthanorequaltothelongeroftheothertwosides.NotethatthisislessconstrainingthantheEuclideancase.Theultra-metricinequalityistomostformsofhierarchicalclusteringwhatthetriangleinequalityistotwo-waymultidimensionalscaling.

† Theterm“soundscape”wasintroducedinthelate1970sbytheCanadiancomposerR.MurraySchafer(1977),whodefinedsoundscapeastheauditoryequivalenttolandscape.BesideSchafer’sproject,thetermsoundscapeperceptionisusedinascientificcontexttocharacterizehowinhabitantsperceive,experience,andappraisetheirsonicenvironment.

Y116134.indb 242 10/10/11 2:12 PM


Part

icip

ants

’ Cat

egor

ies:

Part

icip

ants

’ Ver

baliz

atio

ns:

Soun

d so

urce

sTi

cket

disp

ense

r12

34

67

85

Hum

an ac

tiviti

es

Room

effec

tSm

all s

pace

Conv

ersa

tion,

wai

ting,

step

s

Type

of s

pace

Judg

men

tsQ

uiet

spac

e

Wai

ting

room

,tic

ket o

ffice

Ann

ounc

emen

t mus

ic

Reve

rber

atio

n,co

nfus

ed

Wai

ting

conv

ersa

tions

Unp

leas

ant

Bar,

hall

Whe

eled

suitc

ases

Departure board

Echo

es, r

eson

ance

sEx

tern

al

Plat

form

sCo

rrid

or, h

all

Noi

sy, r

hyth

mic

Step

s, vo

ice

Trai

ns

–

Jingl

es

Arr

ival

s,de

part

ures

Voic

e,ar

rival

s,de

part

ures

Reso

nanc

es,

reve

rbs.

Hal

l

–

Mus

ic, t

rain

sVo

ices

,co

nver

satio

ns

Larg

ere

verb

erat

ion

Hal

l

Noi

sy,

unpl

easa

nt

– Clos

ed sp

ace,

smal

l

Voic

e, pe

ople

Qui

et, i

ntim

ate

Bar,

wai

ting

room

Sem

antic

Fie

lds:

Fig

ure

11.2

(S

ee c

olor

inse

rt.)

Den

drog

ram

repr

esen

ting

the

cate

gori

eso

ftra

ins

tati

ons

ound

scap

esfo

und

inT

ardi

eue

tal.

(200

8).T

hed

escr

ipti

ons

oft

hec

ateg

orie

spr

ovid

edb

yth

epa

rtic

ipan

tsa

rer

epor

ted

ont

het

able

bel

ow,g

roup

edi

nto

five

sem

anti

cfie

lds.

The

mos

tci

ted

verb

aliz

atio

nsa

re

show

nin

ala

rger

siz

efo

nt.

Y116134.indb 243 10/10/11 2:12 PM


transaction,departure,etc.),roomeffect(e.g.,reverberation,confined,exterior/inte-rior,etc.),typeofspace(e.g.,waitingroom,platforms,halls,etc.),andpersonaljudg-ment(e.g.,annoying,pleasant,beautiful,musical,etc.).

11.2.3.4 Prerequisites for using sorting tasks

Thesortingtaskisveryintuitiveforthelisteners,and,inthecaseofthefree-sortingtask,hasthegreatadvantageofleavingthelistenersfreetoarrangethesoundsastheywish.Contrary todissimilarityratings,a largenumberof thesoundscanbehandledbythelistenersinasession.

11.2.3.4.1 Considering a large number of existing sounds. It is possible withsortingtaskstotestmanyexistingsoundsthatarerepresentativeofthevarietyofsoundsunderconsideration.Forinstance74environmentalsoundswerepresentedinBonebright’sstudy(2001),150recordedsoundeffectsinScavone,Lakatos,andHarbke’sstudy(2002),48alarmsoundsinSusini,Gaudibert,Deruty,andDandrel’sstudy(2003),and66trainstationsoundscapesinTardieuetal.’sstudy(2008).

11.2.3.4.2 Collecting information on the type of similarities used for each cate-gory. Fromapracticalpointofview,contrarytotheMDSapproach,categorizationtasksarewelladaptedtodescribeperceptuallyheterogeneouscorporaofsoundsandtorevealdifferentlevelsofsimilaritiesbetweenthesounds.However,greatcarehastobetakenwhenanalyzingthecategoriesbecausethetypeofsimilaritiesusedbytheparticipantsmayvaryfromonecategorytoanother,dependingonthedifficultyinidentifyingthesoundsandontheexpertiseoftheparticipants(moreorlessskillwithsoundevaluation).Indeed,threetypesofsimilaritieshavebeenidentified(Lemaitreetal.,2010),basedonacousticalproperties(loudness,roughness,intensityfluctua-tions,etc.),identifiedphysicalinteractionscausingthesound(impactsoundonglass,rattlesoundonmetal,soundeffect,etc.)andmeaningsassociatedwiththeidentifiedsoundsources(soundsofbreakfast,soundsthatremindoneofchildhood,etc.).

11.2.3.4.3 Selecting the type of similarities. Semanticanalysesoftheverbaldescrip-tionsofthecategoriesproviderichinsightsthatreveal,ontheonehand,thestrategyusedbytheparticipantstoformthecategories,andontheotherhand,thetypeofinformationused.However,semanticanalysesareoftentimeconsumingandhavetobedonerigorouslybyexperts.Lemaitreetal.(2010)proposedanalternative,whichconsistsofaskingtheparticipantstorateforeachcategorywhichtypeofsimilarity(acoustical,causal,semantic)theyhadused.Theresultsmayhelptheexperimentertounderstandtheleveloftheperceptualstructuresunderlyingeachcategory.

11.3 Application: Sound quality of environmental sounds

Thequalityoftheacousticenvironmentiscurrentlyanimportantissue.Effortsarebeingmade toaccount for theannoyancecausedbynoises (Guski,1997).At thesametime,designersareseekingtoimprovethesoundqualityofindustrialproducts.

Y116134.indb 244 10/10/11 2:12 PM


Theideaofsoundqualityhasemergedrelativelyrecently.Itreferstothefactthatthesoundsproducedbyanobjectorproductarenotonlyannoyingorunpleasant,butarealsoawayforpeopletointeractwithanobject.Inthecaseofindustrialproducts,itisthereforeofmajorimportancetodesignsoundstomeetconsumerexpectations.

Sincethebeginningofthe1990s,soundqualityhasbeenconceivedofmainlyintheparadigmofpsychoacoustics.Thishasledtothedesignofexperimentalmeth-odsandauditorydescriptors relevant to soundquality.For instance,ZwickerandFastl(1999)askedparticipantstoratepleasantnessonaunidimensionalscale(e.g.,ratio scale). Then the pleasantness scores were correlated with psychoacousticaldescriptors. Ellermeier, Mader, and Daniel (2004) gathered preference judgmentsof environmental soundsusinga2AFC (twoalternative forcedchoice)procedureandanalyzedthemusingtheBTLtechnique(Bradley–Terry–Luce).Thistechniquerepresentedtheperceivedunpleasantnessonaratioscale.Theunpleasantnessscoreswerethenpredictedbyalinearcombinationofpsychoacousticdescriptors(rough-ness and sharpness). The semantic differential technique is also used to evaluatesoundquality.Ithasbeenlargelyusedforcars(Bisping,1997;Chouard&Hemepl,1999),vacuumcleaners(Ihetal.,2002),andrefrigerators(Jeon,2006).However,asnotedinSection11.2.3.1,definingtheappropriatesemanticdescriptorsofthescalesmustbedonecarefully.

Mostofthestudiesusepsychoacousticaldescriptors(loudness,roughness,etc.)toexplainunpleasantnessscoresorsemanticratings.Thesedescriptorsarecurrentlyincludedinmostsoundqualitysoftwarepackages,yettheyarenotalwaysadaptedtodescribingallkindsofeverydaysounds.Indeed,itappearsthatrelevantperceptualdimensionsaredifferentfromonestudytoanotheraccordingtothecorpusofsoundsunderconsideration.Therefore,thereareno“universal”acousticalorpsychoacousti-caldescriptorsthatcanbeusedtomeasurerelevantauditoryattributesforallcatego-riesofenvironmentalsounds,andwhichwouldthusprovidethesameeffectonthesoundqualityofanyproduct.

11.3.1 Application of the MDS technique to describe environmental sounds

Acrucialaspectfortheresearchinsoundqualityistodeterminetherelevantaudi-tory attributes related to a specific family of environmental sounds. The MDStechniquehasbeenshowntobeafruitfultoolforrevealingandcharacterizingtheunknownperceptualdimensionsunderlyingthetimbreofmusicalsounds.Duringthelastdecade,theMDStechniquehasbeensuccessfullyappliedtodifferentkindsofenvironmentalsounds:everydaysounds(Bonebright,2001),interiorcarsounds(Susini,McAdams,andSmith,1997),air-conditioningnoises(Susinietal.,2004),car door closing sounds (Parizet, Guyader, and Nosulenko, 2006), and car hornsounds(Lemaitre,Susini,Winsberg,McAdams,andLetinturier,2007).Forallthementionedstudies,MDSanalysesledto3-Dperceptualspaces(Figure 11.3pres-entsthe3-Dspaceobtainedforcarsounds)andallthedimensionsexceptoneweredescribedbydifferentacousticalparameters.

Y116134.indb 245 10/10/11 2:12 PM


The spectral centroid* is the acoustical descriptor shared by all the percep-tualspaces related toenvironmentalsounds.Therefore, thisdescriptorappears todescribemusicalsoundsaswellasenvironmentalsoundsandisrelatedtotheseman-ticdescriptors“metallic,”“sharp,”or“brilliant.”Asidefromthespectralcentroid,nouniversalauditoryattributesexisttocharacterizethetimbreofanysound,andaninventoryofthedifferentsalientauditoryattributestodescribethedifferentfam-ilyofsoundsisneeded.Ameta-analysisof10publishedtimbrespacesconductedbyMcAdams,Giordano,Susini,Peeters,andRioux(2006)usingmultidimensionalscaling analyses (CLASCAL) of dissimilarity ratings on recorded, resynthesizedorsynthesizedmusicalinstrumenttones,revealedfourprimaryclassesofdescrip-tors: spectral centroid, spectral spread, spectral deviation, and temporal envelope(effectiveduration/attacktime).

* Thespectralcentroidistheweightedmeanfrequencyofthespectrumofthesignal;eachpartialtoneisweightedbyitscorrespondingamplitude.Thecalculationofthisfeaturecanbemoreorlesscomplex(seetheworkbyMisdariisetal.,2010),butthebasicexpressionis:

SC

A f

A

i ii

ii

=×∑

∑

whereAiandfiaretheamplitudeandfrequencyofthecorrespondingpartial.

0.3

0.2

0.1

0

Dim

. III

0.1

0.2

0.3

0.20

0.20Dim. IDim. II

0.20.2

0.4 0.4

0.4

Figure 11.3 (See color insert.) Three-dimensional space for car sounds: dimension I isexplainedby theenergy ratiobetween theharmonicandnoisyparts,dimension IIby thespectralcentroid,anddimensionIIIbythedecreaseinthespectralenvelope.

Y116134.indb 246 10/10/11 2:12 PM


11.3.2 A general framework for sound quality

Inamoregeneralframework,theMDStechniquemaybecombinedwithanotherapproachbasedonasemanticstudyof thecorpusofsoundsunderconsideration,inordertomappreferencejudgmentsontobothrelevantobjectivedescriptorsandappropriate semantic descriptors. Figure 11.4 presents the framework of the dif-ferent stages of these related approaches. This general framework was appliedusingair-conditioningnoisesasanexampleinathree-partstudybySusini,Perry,Winsberg,Vieillard,McAdams,andWinsberg(2001),Siekierskietal.(2001),andJunker,Susini,andCellard(2001).

ThefirststepconsistsindeterminingtheperceptualspaceusingtheMDStech-nique. Then, in a second step, the acoustical descriptors that are correlated withthepositionsof thesoundsalong theperceptualdimensionsaredetermined. Inaparallelthirdstep,thesoundsareverballydescribedthroughadescriptiveanalysisthatinvolvesasmallnumberoftrainedlisteners.Thisstepprovidesalistofselectedsemanticdescriptors—whichwillbeusedtodefinerelevantsemanticscales—andaverbaldescriptionof theauditorycuesusedby theparticipants tocompare thesounds inorder toguide the researchof theobjectivedescriptorscorrelatedwiththeauditorydimensionsobtainedinthepreviousstage.Inthelaststep,participantsrate their preference (or annoyance) of the sounds. The degree of preference (or,inversely,annoyance)associatedwitheachsoundisrelatedtoafunctionofthesig-nificantobjectivedescriptorsontheonehand,andthesemanticdescriptorsontheother.Theadvantageofthisglobalapproachisthatitdoesnotlimittheexplorationandcharacterizationofthecomponentsofsoundqualitytoacousticalandsemanticdescriptorsthatarealreadyknown.Itprovidesamethodforfindingnewobjective

Semantic description

Relevant objective descriptors ( 1, 2, …)

Acoustical descriptionMultidimensional description

Relevant auditory dimensions

Pref

Preference judgments ( 1, 2, …), (S1, S2, …) S2

S1

2

1

Description of the auditorycues used to compare the

sounds

Relevant semanticdescriptors (S1, S2, …)

Figure 11.4 Frameworkforaglobalsoundqualityapproach,involvingmultidimensional,acoustical,andsemanticdescriptionscombinedwithpreferencejudgments,basedonSusinietal.(2001)andSiekierskyetal.(2001).

Y116134.indb 247 10/10/11 2:12 PM


andsemanticdescriptorsthatareperceptuallyrelevantfordescribingandevaluatingasoundobjectinthedesignprocess.

11.4 Perspectives: Sounds in continuous interactions

Themethodsreportedinthischapteralladdressthemeasurementofquantitiesrep-resentativeofwhatahumanlistenerperceives.Theevaluationoftheperceivedsound qualityofindustrialobjectsisaveryimportantdomaininwhichthesemethodsareapplied.Traditionally,theparadigmofsoundqualityevaluationconsidersalistenerpassively receiving information from the soundsof theproduct.Suchevaluationswould,forinstance,studytheacousticalpropertiesofacarengineroarthatauserprefers(aesthetics)andthatarerepresentativeofasportscar(functionality).

Newtechnologiesforsensingandembeddedcomputation,however,havemadeitpossiblefordesignerstoconsidersonicaugmentationsofamuchwiderarrayofeverydayobjectsthatincorporateelectronicsensingandcomputationalcapabilities.Wherecontinuousauditoryfeedbackisconcerned,thesoundisnolongerproducedinastaticorisolatedway,butisrathercoupledtohumanactioninrealtime.Thisnewdomainofapplicationsiscalledsonic interaction design.

Fromthestandpointofperception,thelevelofdynamicalinteractivityembodiedbysuchartifactsisverydifferentfromthesituationofpassivelisteninginwhichmostofthemethodsreportedarecarriedout.Insonicinteractions,participantsarenotlis-teningtosequencesofstaticsoundsselectedbyanexperimenter,butinsteaddynami-callyexplorethesoundsofaninteractiveobject.Thiscontextmaybethoughttobemorecloselyalliedwithenactiveviewsofperception(e.g.,Bruner,1966)thanwithsomeofthemoretraditionalapproachesfoundinexperimentalauditorypsychology.

Thestudyofsonicinteractionentailsanunderstandingofperceptual–motorbehav-ior,because theseprocessesunderlieanyformofhuman interaction.Newmethodsmay therefore be required. Such methods experimentally study how users performwhenrequiredtodoataskinvolvingsonicinteraction.Aninterestingexampleispro-videdinworkbyRath(Rath&Rocchesso,2005;Rath,2006,2007;Rath&Schleicher,2008).TheydescribetheBallancer,atangibleinterfaceconsistingofawoodenplankthatmaybetiltedbyitsuserinordertodriveavirtualballrollingalongtheplank.Theauthorsusedthisinterfacetostudyparticipants’abilitiestousethisauditoryfeedbackinataskinvolvingguidingtheballtoatargetregionalongthelengthoftheplank,dependingon thekindof soundused.Lemaitreet al. (2009)usedanother tangibleinterface(theSpinotron)implementingthemetaphorofachild’sspinningtoptostudyhowcontinuoussonicinteractionsguidetheuserinmakingaprecisegesture.

References

Aldrich,K.M.,Hellier,E.J.,&Edworthy,J.(2009),Whatdeterminesauditorysimilarity?The effect of stimulus group and methodology, Quarterly Journal of Experimental Psychology, 62(1),63–83.

Bisping, R. (1997). Car interior sound quality: Experimental analysis by synthesis, Acta Acustica United with Acustica, 83,813–818.

Y116134.indb 248 10/10/11 2:12 PM


Björk,E.A.(1985).Theperceivedqualityofnaturalsounds,Acustica, 57,185–188.Bonebright,T.L. (1996).Vocal affect expression: A comparison of multidimensional scal-

ing solutions for paired comparisons and computer sorting tasks using perceptual and acoustic measures,Doctoraldissertation,UniversityofNebraska.

Bonebright,T.L.(2001).Perceptualstructureofeverydaysounds:Amultidimensionalscalingapproach.InICAD,Helsinki,Finland.

Brunner,J.(1966).Toward a theory of instruction.Cambridge,MA:HarvardUniversityPress.Chouard,N.,&Hempel,T.(1999).Asemanticdifferentialdesignespeciallydevelopedfor

theevaluationofinteriorcarsounds.InJoint Meeting: ASA/EAA/DEGA(JASA(105),Berlin),p.1280.

Daniel,P.,&Weber,R.(1997).Psychoacousticalroughness:Implementationofanoptimizedmodel,”Acta Acustica United with Acustica,83,113–123.

Ellermeier,W.,Mader,M.,&Daniel,P.(2004).Scalingtheunpleasantnessofsoundsaccord-ingtotheBTLmodel:Ratio-scalerepresentationandpsychoacousticalanalysis,Acta Acustica United with Acustica, 90,101–107.

Fastl,H.(1989).Averageloudnessofroadtrafficnoise.In InterNoise 89,NewportBeach,CA.Fastl,H.(1991).Evaluationandmeasurementofperceivedaverageloudness.InA.Schick,

J. Hellbrück, and R. Weber (Eds.), Fifth Oldenburg Symposium on Psychological Acoustics(BIS,Oldenburg),205–216.

Fastl,H.(1997).Thepsychoacousticsofsound-qualityevaluation,Acta Acustica United with Acustica, 83,754–764.

Faure, A. (2010). Des sons aux mots, comment parle-t-on du timbre musical? EditionsEdilivre—APARIS,Paris.

Fechner, G.T. (1966). Elements of psychophysics. NewYork: Holt, Rinehart andWinston(Original1860).

Fletcher,H.(1940).Auditorypatterns,Review of Modern Physics,12,47–65.Gaver,W.W.(1993a).Whatdowehearintheworld?Anecologicalapproachtoauditoryevent

perception.Ecological Psychology, 5,1–29.Gaver,W.W.(1993b).Howdowehearintheworld?Explorationsinecologicalacoustics.

Ecological Psychology, 5,285–313.Glasberg,B.R.,&Moore,B.C.J.(2002).Amodelofloudnessapplicabletotime-varying

sounds,Journal of Audio Engineering Society,50,331–342.Goldstone,R.L.,&Kersten,A.(2003).Conceptsandcategorization.InA.F.Healy&R.W.

Proctor(Eds.),Comprehensive handbook of psychology, Vol. 4: Experimental psychol-ogy,chapter22,pp.599–621.NewYork:Wiley.

Grey,J.M.(1977).Multidimensionalperceptualscalingofmusical timbres,Journal of the Acoustical Society of America, 61,1270–1277.

Gros, L., & Chateau, N. (2001). Instantaneous and overall judgments for time-varyingspeechquality:Assessmentsandrelationships,Acta Acustica United with Acustica, 87,367–377.

Guastavino, C. (2007). Catagorisation of environmental sounds, Canadian Journal of Experimental Psychology, 61,54–63.

Guski,R.(1997).Psychophysicalmethodsforevaluatingsoundqualityandassessingacous-ticinformation,Acta Acustica United with Acustica, 83,765–774.

Guyot,F.(1996).Etudedelaperceptionsonoreentermesdereconnaissanceetd’appréciationqualitative: Une approche par la catégorisation. Doctoral dissertation, Université duMaine.

Guyot,F.,Castellengo,M.,&Fabre,B. (1997).Etudede lacatégorisationd’uncorpusdebruitsdomestiques.InD.Dubois(Ed.),Catégorisation et cognition,Paris:Kimé.

Y116134.indb 249 10/10/11 2:12 PM


Hansen,M.,&Kollmeier,B.(1999).Continuousassessmentoftime-varyingspeechquality.Journal of the Acoustical Society of America 106,2888–2899.

Hartmann,W.H.(1997).Signals, sound, and sensation.NewYork:AIPPress.Hedberg,D.,&Jansson,C.(1998).Continuous rating of sound quality.Stockholm:Karolinska

Institutet.Ih,J.-G.,LimD.-H.,JeongH.,&Shin,S.-H.(2002)Investigationonthecorrelationbetween

soundqualityandspectralcompositionvacuumcleanersoundbyusingtheorthogonalarray.InSound Quality Symposium,Dearborn,MI.

Jeon,J.Y.(2006).Soundradiationandsoundqualitycharacteristicsofrefrigeratornoiseinreallivingenvironments,Applied Acoustics, 68,1118–1134.

Junker, F., Susini, P., & Cellard, P. (2001). Sensory evaluation of air-conditioning noise:Comparativeanalysisoftwomethods.InICA,Rome,Italy,p.342.

Kendall,R.A.,&Carterette,E.C.(1992).Verbalattributesofsimultaneouswindinstrumenttimbres.I.vonBismarck’sadjectives.Music Perception, 4,185–214.

Komatsu,L.K.(1992).Recentviewsofconceptualstructure.Psychological Bulletin,112(3),500–526.

Krumhansl,C.L.(1989).Whyismusicaltimbresohardtounderstand?InS.Nielsen,andO.Olsson(Eds.),Structure and perception of electrostatic sound and music(pp.43–53).Amsterdam:Elsevier.

Kruskal,J.B.,&Wish,M.(1978).Multidimensional scaling.BeverlyHills,CA:Sage.Kuwano,S.,&Namba,S. (1978).On the loudnessof road trafficnoiseof longerduration

(20min)inrelationtoinstantaneousjudgment.InASA&ASJjointmeeting.(Journal of the Acoustical Society of America),Berlin, 64,127–128.

Kuwano,S.,&Namba,S.(1985).Continuousjudgementoflevel-fluctuatingsoundsandtherelationshipbetweenoverallloudnessandinstantaneousloudness.Psychology Research, 47,27–37.

Kuwano, S., & Namba, S. (1990). Continuous judgement of loudness and annoyance. InF.Müller(Ed.),Fechner Day, Proceedings of the 6th Annual Meeting of the International Society for Psychophysics,Würzburg,Germany,pp.129–134.

Kuwano,S.,&Namba,S. (2001).Dimensionsofsoundqualityand theirmeasurement. InICA, Rome,Italy,p.56.

Kyncl, L., & Jiricek, O. (2001). Psychoacoustic product sound quality evaluation. In ICA,Rome,Italy,p.90.

Lamalle,C.,Martinez,W.,Fleury,S.,Salem,A.,Fracchiolla,B.,Kuncova,A.,&Maisondieu,A. (2003).LEXICO 3, Outils de statistique textuelle—Manuel d’utilisation,SYLED-CLA2T(Paris3).

Legendre,P.,&Legendre,L.(1998).Numerical ecology.Amsterdam:Elsevier.Lemaitre,G.,Houix,O.,Misdariis,M.,&Susini,P.(2010).Listenerexpertiseandsoundiden-

tificationinfluencethecategorizationofenvironmentalsounds,Journal of Experimental Psychology: Applied,16, 16–32.

Lemaitre,G.,Houix,O.,Visell,Y.,Franinovic,K.,Misdariis,N.,&Susini,P.(2009)Towardthe design and evaluation of continuous sound in tangible interfaces:The spinotron.International Journal of Human Computer Studies,67,976–993.

Lemaitre,G.,Susini,P.,Winsberg,S.,McAdams,S.,&Leteinturier,B. (2007).Thesoundquality of car horns:Apsychoacoustical studyof timbre,Acta Acustica United with Acustica, 93,457–468.

Madsen,C.K.(1996).Empiricalinvestigationofthe“aestheticresponse”tomusic:Musiciansandnonmusicians.In4th International Conference on Music Perception and Cognition,Montreal,Canada,pp.103–110.

Y116134.indb 250 10/10/11 2:12 PM


Maffiolo,V.,Dubois,D.,David,S.,Castellengo,M.,&Polack,J.D.(1998).Loudnessandpleasantness instructurationofurbansoundscapes. InInterNoise,Christchurch,NewZealand.

McAdams,S.,&Bigand,E.(Eds.).(1993).Thinking in sound—The cognitive psychology of human audition,Oxford:UniversityPress.

McAdams,S.,Giordano,B.,Susini,P.,Peeters,G.,&Rioux,V.(2006).Ameta-analysisofacousticcorrelatesoftimbredimensions.InASA,Hawaii.

McAdams,S.,Vines,B.W.,Vieillard,S.,Smith,B.,&Reynolds,S. (2004). Influencesoflarge-scale formoncontinuous ratings in response toacontemporarypiece ina liveconcertsetting.Music Perception, 22,297–350.

McAdams,S.,Winsberg,S.,Donnadieu,S.,Soete,G.D.,&Krimphoff,J.(1995).Perceptualscalingofsynthesizedmusicaltimbres:Commondimensions,specificities,andlatentsubjectclasses.Psychological Research, 58,177–192.

Meunier,S.,Boullet,I.,&Rabau,G.(2001).Loudnessofimpulsiveenvironmentalsounds.InICA,Rome,Italy,p.91.

Misdariis,N.,Minard,A.,Susini,P.,Lemaitre,G.,McAdams,S.,&Parizet,E.,(2010).Environ-mental sound perception: Metadescription and modeling based on independent primarystudies.Journal on Audio, Speech, and Music Processing. doi:10.1155/2010/362013.

Moore,B.C.J. (2003).An introduction to the psychology of hearing (5thed.).NewYork:AcademicPress.

Moore,B.C.J.,&Glasberg,B.R.(1983).Suggestedformulaeforcalculatingauditory-filterbandwidthsandexcitationpatterns.Journal of the Acoustical Society of America,74 (3),750–753.

Moore,B.C.J.,&Glasberg,B.R.(1990).Derivationofauditoryfiltershapesfromnotched-noisedata.Hearing Research, 47,103–138.

Moore,B.C. J.,&Glasberg,B.R. (1996).A revisionofZwicker’s loudnessmodel.Acta Acustica United with Acustica, 82,335–345.

Moore,B.C.J.,Glasberg,B.R.,&Baer,T.(1997).Amodelforthepredictionofthresholds,loudnessandpartialloudness.Journal of the Audio Engineering Society, 45,224–240.

Osgood,C.E.(1952).Thenatureandmeasurementofmeaning.Psychological Bulletin, 49,197–237.

Parizet,E.(1999).Comparisonoftwosemantic-differentialtestmethods.InEAA & ASA joint meeting.Journal of the Acoustical Society of America,Berlin,p.1189.

Parizet,E.,&Nosulenko,V.N.(1999).Multi-dimensionallisteningtest:Selectionofsounddescriptorsanddesignoftheexperiment.Noise Control Engineering Journal, 47,1–6.

Parizet,E.,Guyader,E.,&Nosulenko,V.(2006).Analysisofcardoorclosingsoundquality.Applied Acoustics,69,12–22.

Parizet,E.,Hamzaoui,N.,andSabatié,G.(2005).Comparisonofsomelisteningtestmethods:Acasestudy.Acta Acustica United with Acustica, 91,356–364.

Parizet,E.,Hamzaoui,N.,Segaud,L.,&KochJ.-R.(2003).Continuousevaluationofnoisediscomfortinabus.Acta Acustica United with Acustica,89,900–907.

Patterson,R.D.,&Holdsworth,J.(1991).Afunctionalmodelofneuralactivitypatternsandauditoryimages.Advances in speech, hearing and language processing,Vol. 3.London:JAIPress.

Peters, R. W. (1960). Research on psychological parameters of sound, (pp. 60–249).Hattiesburg:MississippiSouthernCollege.

Plomp, R. (1970). Timbre as a multidimensional attribute of complex tones. In R. Plomp&G.P.Smoorenburg(Eds.),Frequency anaysis and periodicity detection in hearing(pp.397–414).Leiden:A.W.Sijthoff.

Y116134.indb 251 10/10/11 2:12 PM


Pratt,R.L.,&Doak,P.E.(1976).Asubjectiveratingscalefortimbre.Journal of Sound and Vibration, 45,317–328.

Rath,M. (2006).On the relevanceof auditory feedback for subjective andobjectivequal-ityofcontrolinabalancingtask.InProceedings of the 2nd ISCA/DEGA Tutorial and Research Workshop on Perceptual Quality of Systems,Berlin,pp.85–88.

Rath,M.(2007).Auditoryvelocityinformationinabalancingtask.InICAD,Montréal,Canada,pp.372–379.

Rath, M., & Rocchesso, D. (2005). Continuous sonic feedback from a rolling ball. IEEE MultiMedia, 12,60–69.

Rath,M.,&Schleicher,D.R.(2008).Ontherelevanceofauditoryfeedbackforqualityofcontrolinabalancingtask.Acta Acustica United with Acustica, 94,12–20.

Robinson,D.W.,&Dadson,R.S.(1956).Are-determinationoftheequal-loudnessrelationsforpuretones.British Journal of Applied Physics, 7, 166–181.

Rosch,E.(1978).Principlesofcategorisation.InE.RoschandB.B.Lloyd(Eds.),Cognition and categorization,(pp.27–47).Hillsdale,NJ:LawrenceErlbaumAssociates.

Rosch, E., Mervis, C. B., Gray,W. D, Johnson, D. M., & Boyes-Braem, P. (1976). Basicobjectsinnaturalcategories.Cognitive Psychology, 8,382–443.

Scavone,G.P.,Lakatos,S.,&Harbke,C.R.(2002).Thesonicmapper:Aninteractivepro-gramforobtainingsimilarityratingswithauditorystimuli.InICAD, Kyoto.

Schafer,R.M.(1977).The tuning of the world.NewYork:RandomHouse.Schubert,E.(1996).Continuousresponsetomusicusingthetwodimensionalemotionspace.

InICMPC, Montreal,Canada,pp.263–268.Schulte-Fortkamp,B.,&Dubois,D.(2006).Recentadvancesinsoundscaperesearch.Acta

Acustica United with Acustica, 92(6),5–6.Siekierski,E.,Derquenne,C.,&Martin,N. (2001).Sensoryevaluationofair-conditioning

noise:Sensoryprofilesandhedonictests.InICA,Rome,Italy,326.Sloboda,J.A.,&Lehmann,A.C.(2001).Trackingperformancecorrelatesofchangesinper-

ceivedintensityofemotionduringdifferentinterpretationsofaChopinpianoprelude.Music Perception, 19,87–120.

Sloutsky,V.M.(2003)Theroleofsimilarityinthedevelopmentofcategorization.Trends in Cognitive Science, 7,246–251.

Solomon,L.N.(1958),Semanticapproachtotheperceptionofcomplexsounds.Journal of the Acoustical Society of America, 30(5),421–425.

Stepánek,J.(2006).Musicalsoundtimbre:Verbaldescriptionanddimensions.InDAFx-06,Montréal.

Stevens,J.C.,&Tulving,E.(1957).Estimationsofloudnessbyagroupofuntrainedobserv-ers,American Journal of Psychology, 70,600–605.

Stevens,S.S.(1959).Cross-modalityvalidationsofsubjectivescalesforloudness,vibrations,andelectricshock.Journal of Experimental Psychology, 57, 201–209.

Susini,P.,Gaudibert,P.,Deruty,E.,&Dandrel,L.(2003).Perceptivestudyandrecommenda-tionforsonificationcategories.InICAD,Boston.

Susini,P.,Houix,O.,Misdariis,N.,Smith,B.,&Langlois,S.(2009.)Instruction’seffectonsemanticscaleratingsofinteriorcarsounds.Applied Acoustics, 70(3),389–403.

Susini,P.,&Maffiolo,V.(1999).Loudnessevaluationofurbansoundscapesbyacross-modalmatchingmethod. InEAA & ASA joint meeting,Journal of the Acoustical Society of America,Berlin,p.943.

Susini,P.,&McAdams,S.(1998).Globalandcontinuousjudgementsofsoundswithtime-varyingintensity:Cross-modalmatchingwithaproprioceptiveinputdevice.InICA & ASA joint meeting,Journal of the Acoustical Society of America,Seattle,p.2812.

Y116134.indb 252 10/10/11 2:12 PM


Susini,P.,&McAdams,S. (2008).Loudness asymmetry ratingsbetween accelerating anddeceleratingcarsounds,inAcoustics’08, ASA-EAA Joint Conference,Paris.

Susini,P.,McAdams,S.,&Smith,B.(2002).Globalandcontinuousloudnessestimationoftime-varyinglevels.Acta Acustica United with Acustica, 88,536–548.

Susini,P.,McAdams,S.,&Smith,B.(2007).Loudnessasymmetriesfortoneswithincreasinganddecreasinglevelsusingcontinuousandglobalratings.Acta Acustica United with Acustica, 93,623–631.

Susini, P., McAdams, S., and Winsberg, S. (1997). Perceptual characterisation of vehiclenoises. InEEA Symposium: Psychoacoustic, in industry and universities,Eindhoven,TheNetherlands.

Susini,P.,McAdams,S.,Winsberg,S.,Perry,I.,Vieillard,S.,&Rodet,X.(2004).Characteriz-ingthesoundqualityofair-conditioningnoise.Applied Acoustics, 65(8),763–790.

Susini,P.,Misdariis,N.,McAdams,S.,&Winsberg,S.(1998).Caractérisationperceptivedebruits.Acoustique et Techniques,13,11–15.

Susini,P.,Perry,I.,Winsberg,S.,Vieillard,S.,McAdams,S.,&Winsberg,S.(2001).Sensoryevaluation of air-conditioning noise: Sound design and psychoacoutic evaluation. InICA,Rome,Italy,p.342.

Tardieu,J.,Susini,P.,Poisson,F.,Lazareff,P.,&McAdams,S.(2008).Perceptualstudyofsoundscapesintrainstations,Applied Acoustics, 69,1224–1239.

Vanderveer,N.J.(1979)Ecological acoustics: Human perception of environmental sounds.Unpublisheddoctoraldissertation,CornellUniversity,Ithaca,NY.

Vogel, C. (1999). Etude sémiotique et acoustique de l’identification des signaux sonores d’avertissement en contexte urbain.Doctoraldissertation,UniversitéParis6.

vonBismarck,G.(1974).Timbreofsteadysounds:Afactorialinvestigationofitsverbalattri-butes,Acustica, 30,146–158.

Vos,J.(1998).Theloudnessofimpulseandroad-trafficsoundsinthepresenceofmaskingbackgroundnoise,Acta Acustica United with Acustica, 84,1119–1130.

Wedin,L.,&Goude,G.(1972).Dimensionanalysisoftheperceptionoftimbres.Scandinavian Journal of Psychology, 13,228–240.

Weber,R.(1991).Thecontinuousloudnessjudgementoftemporallyvariablesoundswithan “analog” category procedure. InA. Schick, J. Hellbrück, and R.Weber (Eds.),Fifth Oldenburg Symposium on Psychological Acoustics,(pp.267–294)Oldenburg:BIS.

Winsberg,S.,&Carroll,D.(1989).Aquasinon-metricmethodformultidimalscalingviaanextendedEuclidianmodel.Psychometrika, 54,217–229.

Zeitler,A., & Hellbrück, J. (2001). Semantic attributes of environmental sounds and theircorrelationswithpsychoacousticmagnitude.InICA,Rome,Italy,p.114.

Zwicker,E.(1961).Subdivisionoftheaudiblefrequencyrangeintocriticalbands,Journal of the Acoustical Society of America, 33,248.

Zwicker,E.(1984).BASIC-Programforcalculatingtheloudnessofsoundsfromtheir1/3-octbandspectraaccordingtoISO532-B,Acustica, 55,63–67.

Zwicker, E., & Fastl, H. (1972). On the development of the critical band, Journal of the Acoustical Society of America, 52,699–702.

Zwicker,E.,&Fastl,H.(1999).Psychoacoustics: Facts and models (2nded.).Heidelberg:Spinger-Verlag.

Zwicker,R.,andTerhardt,E.(1980).Analyticalexpressionsforcriticalbandrateandcriticalbandwidthasafunctionoffrequency,Journal of the Acoustical Society of America, 68,1523–1525.

Y116134.indb 253 10/10/11 2:12 PM

11 Psychological measurement for sound description and ... · model have also been proposed...

Documents

Transcript of 11 Psychological measurement for sound description and ... · model have also been proposed...