Downloaded by: Purdue University Libraries. Copyrighted material. ... - PDF document

Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. Downloaded by: Purdue University Libraries. Copyrighted material. ndividualswithhigh-frequencysensori-neuralhearingloss(SNHL)aredeniedaccesstopotentiallyimportantspeechinformation.Formilderlosses,thiscanoccuriftheminiatureelectronicsinhearingaidsareunabletoprovidesufÞcienthigh-frequencyampliÞcationorcan-notdosowithoutaudiblewhistlingandover-tonescausedbyfeedback.Formoreseverelosses,theinnerhairscellsthatcodethesefrequenciesmaysimplybeÒdead,Ópossiblyrenderingampli-Þcationinthisregionlessuseful.1Ð5Astheseverityoflossincreasestoincludemorelow-frequencycontent,theamountofÒlostÓspeechinformationincreasesalongwiththechallengesforthelistener.Mostofthesoundswewillbeconcernedaboutinthisdiscussioninvolvethefricatives,affricates,andtheinitialsegmentsofthestopconsonants,whichareprimarilychar-acterizedbyaperiodicmid-tohigh-frequencyspectralinformation.Fig.1showsaspectrogramofthesentenceÒchildrenlikestrawberriesÓwiththepreviouslymentionedsoundclassesdenotedbyarrows.Ascanbeseen,theacousticenergyinthesesoundscanbequitehighinfrequency,withpeakenergysometimesaround9,000Hz,espe-ciallyforwomenandchildrentalkers.6Ð8contrast,thevowelsandotherconsonantsareprimarilycharacterizedbybandsofenergy,for-mants,atrelativelylowerfrequencies.Unlikethediffusespreadofenergythatischaracteristicofthefricatives,affricates,andinitialstopconso-nantsegments,changingthefrequencyrelation-shipoftheformantscanhaveseriousconsequencesfortheidentityofthecorrespond-ingspeechsound.Speechislinguisticallyandacousticallyredundantand,withvaryingdegreesofsuccess,listenerscanidentifyhigh-frequencyphonemesusingonlythetransitionsfromthelower-frequencyformantsofthecoarticulatedpho-nemesthatprecedeandfollowthem.spitethis,thereisevidencethatspeechperceptionimprovesforbothadultsandchil-drenwhenaneffortismadetopreservethehigh-frequencynoiseenergyassociatedwithfrication,especiallywhenidentifying7,11Ð21TheseÞndingsandthefactthathearingaidshavelimitedusablebandwidthhavebeenusedtoexplainthecontinueddifÞ-cultyexperiencedbyyoungchildrenusinghearingaidswhenperceivingandproducingthesesoundscomparedwithvowelsandotherconsonantsoundclasses.Thegravityofthisproblemiscompoundedbytheregularitywithwhich/s/anditsvoicedcognate,/z/,occurintheEnglishlanguage8%ofallspokenconsonants)andbytheirlinguisticimportance.RudminidentiÞesover20linguisticusesfor/s/and/z/,includingplurality,third-personpresenttense,pastversuspresenttense,toshowpossession,possessivepronouns,contractions,andsoon.Develop-mentally,inconsistentexposuretothesepho-nemesforachildwithSNHLmayhavelong-termconsequencesformorphosyntacticdevel-opment.24,27Findingslikethesehavebeenthe Figure1SpectrogramofthesentenceÒchildrenlikestrawberriesÓbeforeprocessingasspokenbyafemaletalker.Arrowsdenotethefricative,affricate,andstopsoundclassesthatcontainsigniÞcanthigh-frequencyenergyandarethereforethesubjectoffrequency-loweringtechniques. RECOGNITIONOFFREQUENCY-LOWEREDSPEECH/ALEXANDER inspirationforavarietyoffrequency-lowering(FL)techniques(i.e.,methodsofmovinghigh-frequencyspeechinformationintolower-fre-quencyregions)incommerciallyavailablehear-ingaids.IncontrasttoafewyearsagowhenFLseemedlikejustasignalprocessingnovelty,nowmorethanhalfoftheworldÕsmajorhearingaidmanufacturersincludeFLasanoptionalfeature.Table1providesasummaryofpeer-re-viewedresearchonmodern,commerciallyavailableFLtechnologiesalongwithafewstudiesunderreviewinwhichtheauthorhaspersonalinvolvement.Asupplementaryde-scriptionofearlyFLtechniquesandasummaryofrelatedresearchÞndingscanbefoundinSpeciÞcstudiesinTable1arediscussedwhereappropriateandspeciÞctech-nologiesarediscussedinthenextsection.Attheoutset,itisimportanttomakenotethatthedataindicatethatwe,asaÞeldofcliniciansandresearchers,donotyetknowenoughtopredictwhowillandwhowillnotbeneÞtfromFLtechnology.First,becausemodernFLtechniqueshaveonlybeenaroundforashortperiodcomparedwithotherhearingaidprocessingstrategies,therejustsimplyisnotalotofdatabyindependentresearchersascanbeinferredfromTable1.Furthermore,wehavethesamedifÞcultiesaswedowhentryingtounderstandindividualvariabilitywithconven-tionalhearingaidsinadditiontonuancesassociatedwithFLtechnology.SpeciÞcally,wenowhavetheaddeddifÞcultyofunder-standing(1)howthedifferenttechnologiesalterthespeechsignalthatistransducedbytheimpairedauditoryperipheryand(2)howthisinteractswithindividualcharacteristicsthatinßuencetheabilitytolearnit.EXTRINSICFACTORSRELATEDTOTHESIGNALPROCESSINGANDIMPAIRMENTFig.2providesaframeworkforunderstandingthedifferentfactorsthatlikelyinßuenceindi-vidualvariabilityinrecognitionoffrequency-loweredspeech.TheÞrstsetoffactorstobediscussedareextrinsictothelistenerandrelatestotherepresentationofthesignalattheaudi-toryperiphery.ThespeciÞcspeechsoundanditsenvironmentalcontextserveasinputtothedigitalsignalprocessorthatalterstheacousticrepresentationinawaythatisspeciÞctothesignalandtotheFLtechnologyanditssettings.HowthisnewlycodedinputisthenÒseenÓbythecentralauditorysystem,andbeyond,de-pendsontheintegrityofperipheralprocessing(e.g.,thresholdelevationandbroadenedaudi-toryÞltersassociatedwithvaryingdegreesofouterandinnerhaircellimpairment).Itisarguedherethatthesefactorscontributethemosttodifferencesinindividualoutcomes.Becausethereisnothingonecandotocontroltheinputsignalortheperipheralprocessing,itisincumbentuponthehearingaiddispensertochoosetheappropriatetechnologyandsettingsthatmatchthespeechperceptiondeÞcitwiththeimpairment.Aswillbediscussed,withafewexceptions,therearenoÞrmguidelinesonhowthisshouldbedoneduringÞtting,whichputstheonusonthedispensertouseappropriatemeasurestovalidateindividualoutcomes.SignalProcessingDetailsTomakeinformedchoicesaboutwhen,what,andhowtoimplementFLinahearingaidÞtting,itiscriticalthatthedispenserunder-standsthetechnologyandhowthehandlesintheÞttingsoftwaremanipulatethesignal.Unfortunately,thesedetailsaresometimeshardtocomeby.Presentedbelowisasum-maryofinformationgatheredfromdetailedacousticanalysesbytheauthorforallbutthemostrecentFLtechniques.IndustryinterestinFLhadaslowstart,buthasrecentlysurged.Currently,ÞveFLtechniquesareimple-mentedincommerciallyavailablehearingLinearfrequencycompressionbyAVRSo-novationofIsrael(introduced1991)Linearfrequencytransposition(LFT)byWidexofLynge,Denmark(introducedNonlinearfrequencycompression(NFC)byPhonakofStafa,Switzerland(introduced2008).AlsousedbyUnitronofKitchener,Ontario,Canada(starting2012)SpectralenvelopewarpingbyStarkeyHear-ingTechnologiesofMinnesota(EdenPrarie,MN),USA(introduced2011) SEMINARSINHEARING/VOLUME34,NUMBER22013 Table1SummaryofStudiesUsingModernFrequency-LoweringTechnologyReference/TechniqueParticipantDetailsMethods/OutcomeMeasuresResultsKuketal(2009)/LFT8adultswithseveretoprofound2kHzDailytrainingforÞrstmowithLFT;nocontrolfortrainingwithoutConsonantrecog(nonsenseCVCVCs)inquiet(50-&68-dBSPL)&babblenoise,assessedafterinitialÞt,1mopostÞttraining,2moAt50-dBSPL,fricativesimproved5Ð10%initially&anadditional10%after2moAt68-dBSPL,fricativesimproved10%after2mo;aninitialdecreaseinstopswasbacktobaselineafter2moNodifferenceforspeechinnoise,exceptfricatives(15%improvement)Auriemmoetal(2009)/LFT10childrenwithseveretoprofound3kHzPhonemerecog(nonsenseCVCVs)inquiet&fricativeproductionConditions:participantÕsownaid,newdevicewithoutLFT(3wk),&withLFT(6wk)WeeklyauditorytrainingwithoutLFT&withLFTAfter6wk,consonantrecogwithLFTat30-dBHLwas20%betterthanwithoutLFT;nosigniÞcantdifferenceat50-dBHLWorstperformerswithoutLFThadgreatestimprovementwhenactivatedNovowelrecogdifferencesProductionaccuracyof/s/&/z/improvedbywhenactivatedSmithetal(2009)/LFT6childrenwithseveretoprofound1kHzAudio&audiovisualrecogoflivevoicemonosyllables(CVCs)Consonantproductioninwords&sentencesParental&teacherreportsNorandomizationorcontrolformaturationeffectsImprovementsinCVCrecogat3&6moforaudioonly(notaudiovisual)&productionat6moAlexanderetal(2008)/LFT&NFC24adultswithmildtomoderateSNHL&24normal-hearingcontrolsRecogoffricatives&affricates(VCs)innoiserecordedthroughaidswithLFT&NFCAgainstmanufacturersÕcandidacyConditions:noLFT/noNFC(conventional),LFT,NFC,wideband(byaddinghigh-passÞlteredspeechtotherecordings)WidebandimprovedoverconventionalNFCwithinputbandwidth9kHzimprovedoverconventionalLFTsigniÞcantlydegradedrecog,especially/s/&/z/,forbothgroupsSimpsonetal(2005)/NFC17adults(experiencedHAusers)withmoderatetoprofoundSNHLRecogofCVCmonosyllabicwordsConditions:conventionalaid,experimentalNFC(4Ð6wk)Noevidencethataudibilityfortheloweredspeechwasmeasured8hadimprovementinhigh-freq.wordrecogwithNFC,8showednodifference,&1performedworseSimpsonetal(2006)/NFC7adultswithprecipitousSNHLRecoginquiet(open-setmonosyllabicwords,closed-setVCVs)&noise(open-setsentences)Subjectivemeasure:APHAB2conditions:conventionalaid,experimentalNFC(4Ð6wk)Startfreq.rangedfrom1.0to1.6kHzNoevidencethataudibilityfortheloweredspeechwasmeasuredNosigniÞcantdifferenceinmonosyllabicword&consonantrecoginquietOnly1listenershowedimprovementinsentencerecoginnoisewithNFCAPHAB:higherglobalscoresfor4listenerswithconven-tionalaid RECOGNITIONOFFREQUENCY-LOWEREDSPEECH/ALEXANDER Table1(Reference/TechniqueParticipantDetailsMethods/OutcomeMeasuresResultsGlistaetal(2009)/NFC13adults&11childrenwithslopinghigh-freq.SNHLrangingfrommoder-atelyseveretoprofoundAidedspeechdetection,speechrecogforconsonants,plurals,&3phases:acclimatizationtoconventionalprocessing,treatmentwithprototypeNFC,treatmentwithdrawal(conventional5adultshadsigniÞcantimprovement,mostlyforplurals7childrenhadsigniÞcantimprovement,mostlyforBeneÞtforpluralsdependedonagegroup&high-freq.lossadultspreferredNFC1adulthadworsevowelperceptionBohnertetal(2010)/NFC11adultswithseveretoprofoundNFCcomparedwithlistenersÕowndeviceafter2&4moStartfrequencies2kHzfor7listenersSRT:nonsensesentencesinnoiseSubjectivequestionnairesComparisonbetweendevicesnotvalid;also,noblindingtotheReportsoffricativesassoundingunnaturalwithNFC,despitebettersound-qualityratingsWolfeetal(2010)/NFC15childrenwithmoderatetomoder-atelysevereSNHLNFCenabledversusNFCdisabled(6-wkacclimatizationperiodAidedspeechdetection,pluralrecoginquiet,sentencerecoginmultitalkerbabble,high-freq.consonantdiscriminationinquietPluralrecogsigniÞcantlybetterwithNFC(84%versusImprovedconsonantdiscrimination(/s/&/d/)withNFCNosigniÞcantdifferenceforsentencerecoginnoiseWolfeetal(2011)/NFCSeeWolfeetal(2010)6-mofollow-uponWolfeetal2010usingthesamemeasuresNocontrolorwithdrawalcondition,thus,cannotruleoutmatura-tion/learningeffectsImprovedconsonantdiscriminationwithNFCafter6moversus6wkImprovedsentencerecoginnoiseonlyafter6moMcCreeryetal(2012)/NFC20adultswithnormalhearingProcessingfor3audiogramswithvaryingdegreesofhigh-freq.hearinglossRecogoffricatives&affricates(nonsenseCVCs)3conditions(usingthePurduehearingaidsimulator):conventionalprocessing,manufacturerÕsdefaultNFCsettings,NFCsettingswithoptimizedbandwidthSpeechrecogimprovedacrossconditionsasestimatedaudibility&bandwidthincreasedOptimizedÞttingmethodresultedinsigniÞcantlyhigherspeechrecogcomparedwithotherconditionsGlistaetal(2012)/NFC6childrenwithatleastmoderatelyseverehigh-freq.SNHLAidedspeechdetection,pluralrecog,/s/-//discrimination(CVpairs),consonantrecog3phases:conventionalprocessing(baseline),NFC(treatmentwk),conventionalprocessing(withdrawal)Single-subjectdesign;cannotdiscountmaturationeffectsSigniÞcantbeneÞt(upto20%)fromNFCfollowingaccli-matizationfor5listenersonatleast1test,especiallyfricatives;theotherlistenerwasnearceilingwith&withoutNFCformosttestsImprovedpluralrecog(3listeners)atvaryingtimesdur-ingtreatmentphaseImproved/s/-//discrimination(3listeners)overtimeAcclimatizationtrendswerevariable SEMINARSINHEARING/VOLUME34,NUMBER22013 Table1(Reference/TechniqueParticipantDetailsMethods/OutcomeMeasuresResultsEllis&Munro(2013)/NFC15adultswithnormalhearingSentencerecoginnoiseCognitivemeasures:workingmemoryspan(readingspantest),executivecontrolfunction(trailmarkingtest)3conditions:noNFC&2NFCwitha1.6kHzstartfreq.(2:1or3:1compressionration)NFCsigniÞcantlydecreasedspeechrecogSigniÞcantcorrelationbetweenrecogofunprocessedspeech&3of4cognitivemeasuresNosigniÞcantrelationshipbetweenrecogofNFCspeech&cognitivemeasuresKopunetal(2012)/NFC12children&24adultswithmildtosevereSNHLRecogofmonosyllabicwordswith&withoutNFCusingPurduehearingaidsimulator,assessedbefore&aftershort-termaudiovisu-alexposurewithNFCNFCsigniÞcantlyimprovedwordrecogNopreferentialeffectforexposureorforagegroupOnly2adultsperformedworsewithNFCBrennanetal(2012)/NFC19children&24adultswithmildtosevereSNHLSound-qualityratingsforspeech(sentenceswith3fricatives)&foravarietyofmusicgenres3conditionsprocessedwithPurduehearingaidsimulator:restrictedbandwidthat5kHz,extendedbandwidthat10kHz,&NFCImprovedsoundqualitywithextendedbandwidth/NFCcomparedwithrestrictedbandwidthforspeech;nodif-ferencesformusicListenersÕpreferenceforextendedbandwidth/NFCposi-tivelycorrelatedwithampliÞcationexperience,butnorelationshipwithaudiogramAlexander(2012)/NFC28adultswithmildtomoderatelysevereSNHLRecogofconsonants&vowelsfromnonsensesyllablesinnoisePurduehearingaidsimulatorusedtocreateacontrol&6NFCconditionswithÞxedoutputbandwidth(2groupsat3.3&5.0kHz)butvaryingstartfreq.&compressionratioImprovementinfricativerecogformostNFCsettingsrelativetocontrol1.6kHzstartfreq.decreasedvowel&nonfricativecon-sonantrecog(increasingstartfreq.wasmoreeffectiveatrestoringrecogthandecreasingcompressionratio)Abbreviations:APHAB,AbbreviatedProÞleofHearingAidBeneÞt;C,consonant;Freq.,frequency;HA,hearingaid;HL,hearinglevel;LFT,linearfrequencytransposition;NFC,nonlinearfrequencycompression;recog.,recognition;SNHL,sensorineuralhearingloss;SPL,soundpressurelevel;SRT,speechrecognitionthreshold;V,vowel. RECOGNITIONOFFREQUENCY-LOWEREDSPEECH/ALEXANDER FrequencycompressionbySiemensofErlangen,Germany(introduced2012)LinearFrequencyCompressionOneoftheearliestcommerciallyavailableFLtechniqueswaslinearfrequencycompressionintroducedbyAVRSonovationinanalogde-vicesin1991andlaterindigitaldevicesin2004.Akeyelementofthistechniqueisitsswitchingbehavior.AspectralbalancedetectorisusedatthefrontendoftheprocessingtodeterminewhenFLshouldoccur.SpeciÞcally,iftheenergyabove2,500Hzisgreaterthanthatbelow2,500Hz,thenFLoccurs,otherwiseampliÞcationisprovidedasusual.Hence,FLisdynamicbecauseofitsallornonebehavior.Atleastforspeechinquiet,thisisagoodmethodforidentifyingthespeechsoundschar-acterizedbyhigh-frequencyaperiodicenergy.FLisimplementedquitesimplyusingaÒslowplayÓeffect;infact,theearlyanalogversionusedamagnetictapethatwasplayedataslowrateofspeed.Digitally,thisisdonebyusingtwodifferentanalog-to-digitalconverters(ADCs)atthefrontend.TheÞrst,ADC1,hasasam-plingrateequaltothesamplingrateofthedigital-to-analogconverter(DAC).Theother,ADC2,hasaprogrammablesamplingratethatisanintegermultiple(2,3,4,or5)oftheDACsamplingrate,calledthedynamicfrequencycompressioncoefÞcient.Theformer,ADC1,isusedwhennoloweringoccurs(i.e.,whenthesignalislow-frequencydominated,aswithvowels)andtheotherisusedforlowering(i.e.,whenthesignalishigh-frequencydomi-nated,aswithfricatives,etc.).BecausethesamplingrateoftheDAC(output)isslowerthantheADC2(input),theentirefrequencyrangeisshiftedproportionallylowerbyafactorequaltothedynamicfrequencycompressioncoefÞcient(Figs.3and4).LinearFrequencyTranspositionIn2006,WidexintroducedLFTtothemarketastheÒAudibilityExtender.ÓWhereasAVRSonovationhearingaidswereprimarilynicheproducts,thiswastheÞrsttimethattheconceptofFLwentmainstream.LFTwassimplyanoptionaladd-onfeatureofanalreadyfullydevelopedlineofproducts.WhentheLFTfeatureisactivated,thealgorithmcontinuallysearchesforthemostintensespectralpeakinalimitedfrequencyrangeknownastheÒsource Figure2Aframeworkforunderstandingfactorsthatlikelyinßuenceindividualdifferencesinspeechrecognition.Extrinsicfactorsaffecttherepresentationofthefrequency-loweredspeechattheauditoryperiphery,includingthespeciÞctechniqueandthesettingschosenforit,andintrinsicfactorscontributetoanindividualÕsabilitytolearnandbeneÞtfromthissignal. Figure3Relationshipbetweeninputandoutputfrequenciesforlinearfrequencycompression.Whentheinputisdominatedbyenergybelow2,500Hz,noloweringoccurs.Whenitisdominatedbyenergyabove2,500Hz,theentirefrequencyrangeiscom-pressedbyafactordeterminedbythedynamicfrequencycompressioncoefÞcient.Abbreviation:DFC,dynamicfrequencycompression. SEMINARSINHEARING/VOLUME34,NUMBER22013 region.ÓThesourceregionisdeterminedbyaprogrammableÒstartfrequencyÓthatincludestheone-third-octavebandfrequenciesfrom630to6,000Hz.Thesourceregionbeginsahalfoctavebelowthestartfrequencyandex-tendsoneoctaveaboveitortothelimitoftheinputbandwidthofthemicrophoneand/orADC.ThefrequencyregiontowhichtheinputistransposediscalledtheÒtargetregionÓandisoneoctavebelowthesourceregion(Table2).Anoctave-wideband(relativetothetargetdestination)isÞlteredaroundthedominantspectralpeakinthesourceregionandisthenresynthesizedoneoctavedown(afactoroftwo),thusmixingwithanylow-frequencyenergythatmightbepresent.Anoptionalexpandedmodeexistsforstartfrequencies2,500Hzwherethesourceregionbeginsahalfoctavethestartfrequencyandextendsforanadditionaloctave(Table3).Aswiththebasicmode,anoctave-wideband(relativetothetargetdestination)isÞlteredaroundthedomi-nantspectralpeakinthesourceregion,butisthenresynthesizeddownbyafactorofthreeinsteadoftwo.Hence,belowthenominalstartfrequencythereisapotentialmixingofenergyfromtheoriginalinputsignal,thetransposedsignalfromthebasicmode,andthetransposedsignalfromtheexpandedmode(Fig.5A). Table2TheApproximateSourceandTargetRegionsoftheBasicTransposedSignalsforDifferentLFTStartFrequencies(inHz)StartFreq.SourceRegionTarget630445Ð1260223Ð630800566Ð1600283Ð8001,000707Ð2000354Ð10001,250884Ð2500442Ð12501,6001131Ð3200566Ð16002,0001414Ð4000707Ð20002,5001768Ð5000884Ð25003,2002263Ð64001131Ð32004,0002828Ð80001414Ð40006,0004242Ð(max)2121Ð(max/2)Abbreviations:Freq,frequency;LFT,linearfrequency;Note:Actualvalueswilldependontheaudibilityoftheloweredsignal.ThevalueforÒmaxÓcorrespondstothemaximumfrequencyrepresentedbytheanalog-to-digital Figure4AspectrogramofthesentenceÒchildrenlikestrawberriesÓfromFig.1afterprocessingwithlinearfrequencycompressionandadynamicfrequencycompressioncoefÞcientofthree.Energyfrom0to8kHziscompresseddownto0to2.67kHzwhenthespectralbalanceoftheinputsegmentishigh-frequencydominated.BoxeshighlightthevisuallyidentiÞableenergyalteredbytheprocessing. Table3TheApproximateSourceandTargetRegionsoftheExpandedTransposedSignalsforDifferentLFTStartFrequencies(inHz)StartFreq.SourceRegionTarget630891Ð1782297Ð5948001131Ð2263377Ð7541,0001414Ð2828471Ð9431,2501768Ð3536589Ð11791,6002263Ð4525754Ð15082,0002828Ð5657943Ð18862,5003536Ð(max)1179Ð(max/3)Abbreviations:Freq,frequency;LFT,linearfrequency;Note:Actualvalueswilldependontheaudibilityoftheloweredsignal.ThevalueforÒmaxÓcorrespondstothemaximumfrequencyrepresentedbytheanalog-to-digital RECOGNITIONOFFREQUENCY-LOWEREDSPEECH/ALEXANDER Unliketheprevioustechnique,whenactivated,FLiscontinuous,althoughitsbehaviorisdynamicbecausewhatisloweredandwheredependsonthespectralcontentoftheinputsignal(Fig.6).NonlinearFrequencyCompressionIn2008,PhonakintroducedtheirÞrsthearingaidswithNFC,knownasÒSoundRecoverÓtothemarket.Insomerespects,frequencycom-pressionasimplementedbyPhonakisoppositeoffrequencycompressionasimplementedbyAVRSonovation.Whereasthelatterisselec-tiveintimeandisimplementedlinearlyinthetemporaldomain(i.e.,byexploitingsamplingrate)acrosstheentirefrequencyrange,theformerisfrequencyselectiveandisimple-mentednonlinearlyinthespectraldomainacrosstime.Thatis,whenactivated,NFCisalwaysoperating,butitonlydoessooveralimitedanalysisbandthatisdeterminedbyaprogrammablestartfrequencythatrangesfrom1,500to6,000Hz.FrequenciesbelowthestartfrequencydoundergoFL(Fig.7).ThisisakeydifferencebetweenhowLFTandNFCarecontrolledbythehandleswithintheWidexandPhonakprogrammingsoftware.WithWidex,alltheFLoccursthestartfrequencywhereaswithPhonak,alltheFLoccursthestartfrequency.WiththeÞrstgenerationproduct(NaõUP),onlyfrequenciesupto6.3kHzweresubjecttofrequencycompression.Withthesecondgenerationofproducts,abandabout4.5to4.8kHzwidebeginningjustbelowthestartfrequencywassubjecttofrequencycom-pression,withanupperlimitof10kHz.Withthelatestgenerationofproductsintroducedin2011,allfrequenciesbeginningwiththestartfrequencyandcontinuingthrough10kHzun-dergofrequencycompression.Therelationshipbetweeninputandoutputfrequenciesisdeter-minedbythecompressionratio,whichcanvaryfrom1.5:1to4.0:1.AsconÞrmedbycalcula-tionsperformedbytheauthor,becauseFLoccursonalogscale,thecompressionratiocorrespondstothepsychophysicalreductioninspectralresolutionintermsofauditoryÞlters(e.g.,2.0:1meansthatinformationthatwouldnormallyspantwoauditoryÞltersinthe Figure5Relationshipbetweeninputandoutputfrequenciesforthreedifferentstartfrequencies(differentpanels)forlinearfrequencytransposition.ThedottedlineineachpanelcorrespondstoahypotheticallossinwhichthemaximumaudibleoutputfrequencywithampliÞcationis1,800Hz.Theuntransposedsignalisrepresentedbycircles,thetransposedsignalbysquares,andtheexpandedtransposedsignalbytriangles.(A)Theboxesindicatefrequenciesofoverlapbetweentheuntransposed,transposed,andexpandedtransposedsignalswhenthestartfrequencyis1,600Hz.(B)Theboxindicatesanislandofintermediateinputfrequenciesthatareinaudiblewhenthestartfrequencyis3,200Hz.(C)Thearrowsindicateanidealscenariowhenthestartfrequencyis2,500Hzinwhichtheinputfrequencycorrespondingtowhereaudibilityfortheuntrans-posedsignalendsisclosetothestartfrequencyfortranspositionandwhereaudibilityforthissignalendsisclosetothestartfrequencyforexpandedtransposition. SEMINARSINHEARING/VOLUME34,NUMBER22013 unimpairedearbeforeprocessingwillonlyspanoneauditoryÞlterafterprocessing).Topreventdispensersfrombeingoverwhelmed,optionsintheprogrammingsoftwarearelimitedto12to15presetcombinationsofstartfrequencyandcompressionratio.Theexactcombinationsdependonthehearinglossenteredintothesoftware.ShowninFig.8isafamilyoffre-quencyinput-outputcurvesthatweremeasuredempiricallyusingahearingaidwiththesecondgenerationalgorithm.SiemenshasrecentlyimplementedaformofFLsimplyknownasÒfrequencycompres-Detailsaboutitssignalprocessingareunknownatthistime.DispensercontroloverFLisdifferentfromSoundRecoverinthattwohandles,fandf,determinethestartandendfrequenciesofthetargetregion,respective-ly.Thecompressionratioisthendeterminedbythesettingsforthetwohandles,whichhavelowerandupperlimitsof1.5and8.0kHz,SpectralEnvelopeWarpingStarkeyisalsooneofthelatestcompaniestoimplementFL,knownasÒSpectraliQ,Óinitshearingaids.ThealgorithmisdescribedasÒspectralenvelopewarping.ÓThetermfeaturedetectionisusedinthealgorithmde-scriptionbecauseaclassiÞerlooksforspectralfeaturesinthehigh-frequencyspectrumthatarecharacteristicofspeech.Thetermhasbeenusedtodescribethealgorithm Figure6AspectrogramofthecarriersentenceÒchildrenlikestrawberriesÓfromFig.1andthestimulusÒeeSHÓafterprocessingwithlinearfrequencytranspositioninwhichpeakenergyfrom4,242to7,000Hzistransposeddownto2,121to3,500Hz.BoxeshighlightthevisuallyidentiÞableenergyalteredbytheprocessing. Figure7AspectrogramofthesentenceÒchildrenlikestrawberriesÓfromFig.1afterprocessingwithnonlinearfrequencycompressioninwhichtheentirebandfrom1.5to6.0kHziscompressedto1.5to3.5kHz.OneconsequenceofalowstartfrequencythatcanbeseenfromtheÞgureisthatformanttransitionsbecomeßattenedcomparedwiththeinputinFig.1. RECOGNITIONOFFREQUENCY-LOWEREDSPEECH/ALEXANDER behaviorasthesefeaturesareaddedtothelow-frequencysignal(Fig.9)inawaythatpreservestheirnaturalharmonicstructure(Fig.10).An-otherkeyfeatureisthat,unlikeotherFLtechniques,thistechniquedoesnotrolloffthehighfrequenciesbeyondthetwoupperchannels(5.7kHz).Thisisdonetominimizetheriskofthedispenserunintentionallylimit-ingaudiblebandwidthbychoosingtooaggres-siveaFLsetting,theimportanceofwhichwillbediscussedlater.DispensercontrolofSpectraliQtakestheformofaseven-pointscalethatcorrespondstothebandwidthofthesourceregion,withhighersettingsreservedformoreseverehearinglosses.Againcontrolisalsoprovided;thiscontrolindependentlyadjuststhelevelofthetranslatedspectralfeature.SummaryofTechniquesFig.11providesavisualschematicforcompar-ingthecommerciallyavailableFLtechniques.FrequencyisrepresentedbythecolormapÑthelowestfrequenciesarered(bottom)andthehighestfrequenciesareviolet(top).TheÞrstbarshowstheinputbandandindicatesthatthesourceregionisthefrequencyrangeforwhichthehearingaidcannotprovidesufÞcientaudi-bility.Thefollowingfourplotsshowhoweachtechniqueapproachesthegoalofbringingdownthisinformationintoaregionofaidedaudibility.Fig.12showsaclassiÞcationofthesetechniquesalongtwodimensions:algorithmactivation(inputdependentversusalwaysac-tive)andtechniquethatservesasthebasisforFL(compressionversustransposition).Keycharacteristicsoffrequencycompressionare(1)thetargetregion(thefrequencyrangewhereinformationismovedto)iscontainedwithinthesourceregion(thefrequencyrangeoranalysisbandthatissubjecttolowering),(2)thebandwidthofthesourceregionisreduced,and(3)thestartfrequency(whichcanbe0Hz)islikeananchorthatdoesnotmove.Keycharacteristicsoffrequencytransposition/trans-lationare(1)thereislessoverlapbetweentargetandsourceregions,(2)thebandwidthofthesourceregionisnotreduced,and(3)thestartfrequencyismovedtoalowerfrequency(i.e.,thereisamixingofloweredandunloweredsignals). Figure8Afamilyoffrequencyinput-outputcurvesfromthesecondgenerationimplementationofnon-linearfrequencycompressionforoneparticularhearingloss.Foreachsetting,theÞrstnumberisthestartfrequency,followedbythecompressionratio.Thedottedlinerepresentsthemaximumaudibleoutputfrequencyachievablewiththehearingaidforahypotheticalhearingloss.Asterisksindicatetheonlyappropriatesettingsforthishypotheticallossbasedoncriteriaoutlinedinthetext(i.e.,audibilityfortheloweredsignalinawaythatdoesnotrestricttheaudiblebandwidthoftheprocessedsignal). Figure9AspectrogramofthesentenceÒchildrenlikestrawberriesÓfromFig.1afterprocessingwithspectralenvelopewarping.BoxeshighlightthevisuallyidentiÞableenergyalteredbytheprocessing. SEMINARSINHEARING/VOLUME34,NUMBER22013 ModelingtheInteractionbetweentheSignalProcessing,HearingLoss,andHistorically,attemptstoimplementFLhavebeenlimitedtoindividualswithseveretoprofoundSNHL.Fortheseindividua-ls,argumentsfortheuseofFLarerelativelyeasytomakebecausedeÞcitsinspeechrecog-nitionincreaseashigh-frequencyaudibilitydecreases.However,asthecutofffre-quencyofaudibilitydecreases,thechallengesinvolvedwithFLincrease.Reasonsforthisincludethefactthatthereismorelostinforma-thatneedstoberecoveredandacorre-spondinglysmallerregionforrecodingit.Inaddition,therecodedinformationmustbemovedtoregionswherecriticallow-frequencyspeechinformation(i.e.formants)mightal-readyexist.Ontheotherhand,argumentsjustifyingtheuseofFLforindividualswithmildtomoderateSNHLaremoredifÞculttomakebecausetheoveralldeÞcitattributedtoband-widthreductionfortheselossesisless,whichmeansthatpotentialbeneÞtisalsoless.Inaddition,earlyFLtechnologyinvolveddrasticalterationsofthesignalandpossibleaudibleAssuch,thelikelihoodofÒdoingmoreharmthangoodÓwasrelativelyhigh.ModerntechniquesandadvancesinsignalprocessinghavereducedsomeoftheserisksassociatedwithFLand,ingeneral,maybemoreamenabletolistenerswithmilderdegreesofSNHL,especiallychildrenwhodependonthefullbandwidthofspeechfornormalspeechandlanguagedevelopment(seefollowingAsjusthighlighted,therelationshipbe-tweenSNHLandFLseemstobegiveandtake.Astheseverityoflossincreases,thedeÞcitandcorrespondingpotentialbeneÞtincreasealongwiththerisksassociatedwithusingFL.35,40,41Withlesssevereloss,thereislessdeÞcitandpotentialbeneÞt,butalsolessriskinvolvedwithThisrelationshipcanbebetterunder-stoodusingtheschematicinFig.13.OntheabscissaistheamountofFL.BecausetheacousticÞdelityofthesignalisradicallyalteredbyFL,theabscissaalsoislabeledÒdistortion.Ó Figure10AuthorÕsrenditionofhowspectralenve-lopewarpingaddsinformationfromthehigh-fre-quencyspectrumtothelow-frequencyspectrumwhilemaintainingtheharmonicstructureofthesourcesignalandtheoriginalhigh-frequencyspectralcontent. Figure11Visualschematicforcomparingthefourfrequency-loweringtechniques.Seetextfordetails. RECOGNITIONOFFREQUENCY-LOWEREDSPEECH/ALEXANDER FLmightbeconsideredÒconstructivedistor-tionÓwhenitaidsinspeechrecognitionandÒdestructivedistortionÓwhenitdoesnot,thetermisappropriateeitherway.Ontheordinateisinformation,whichcanbequanti-Þedusinginformation-theoretictermslikeorunitsfromtheSpeechIntelligibilityIndex,andsoon.Thedashedlinerepresentsthepotentialinformationgainedbymovingother-wiseinaudiblehigh-frequencycontenttolow-er-frequencyregionsofaudibility.Uptoacertainpoint,theamountofhigh-frequencyinformationincreaseswithincreasesinFL,beyondwhich,informationdecreasesascochle-arlimitationstakeover(e.g.,severeamountsoffrequencycompression).Thelinedoesnotgobelowzerobecauseitrepresentsonlythatinformationthatisinaccessiblewithconven-tionalampliÞcation,sopresumablyFLcannotmakethisinformationworsethannothavingitatall.ThedottedlinerepresentsthepotentialdecreaseininformationforthatpartofthespectrumthatcanbeampliÞednormally.ItnevergoesabovezerobecauseitishardtoimaginethatFLcanactuallymakethisinfor-mationbetter.ForsmalldegreesofFL,thisinformationmaybeundisturbed;forexample,whenNFChasahighstartfrequency.However,asFLincreases,theinformationcontainedinthelow-frequencyspectrumbe-comesdegraded.Themagnitudeofthelineforthehigh-frequencycontentisintentionallylessthanthatforthelow-frequencycontenttorespectthedifferencesintheamounteachcontributestooverallspeechrecognition.Theabsolutemagnitudeandtheunderlyingformofeachfunctionwillthenshiftdependingontheseverityofloss,thespeciÞcFLtechnolo-gyandsettings,andthespeechsoundsinvolved.ThethicksolidlineinFig.13representsthesummationofinformationgainedandinformationlostattributabletoFL.ThegoalofthehearingaiddispenseristochoosetheFLsettingthatmaximizesthisfunctionfortheindividual.Althoughonecannotknowwhattheunderlyingfunctionis,orwhereeachset-tingisatonthefunction,probemicrophonemeasurementsorsubjectivelisteningtestscanhelpguardagainstthetwoextremes.TheÞrstextremeoccurswhentheFLinformationismovedtoaregionthatisstillinaudibleforthelistener.Ifnoadditionalinformationismadeaudible,thennobeneÞtshouldbeexpected.Althoughthisseemsobvious,confusionaboutthetechnologyorafailuretoverifyaidedaudibilitycanresultinthisscenario.Forexam-ple,asalreadymentioned,withWidexalltheFLoccursbelowthestartfrequencywhereaswithPhonakalltheFLoccursabovethestartfrequency.Iftheunderstandingfortheformerwereconfusedforthelatter,itislikelythatlittletononeoftheFLinformationwouldbeaudi-ble.VeriÞcationofaudibilityforfrequency-loweredspeechshouldalsobeaconsiderationwhenevaluatingresearchoutcomesasinTable1.Forexample,theearliestreportedinvestigationsofNFCwithadultsdidnot Figure12ClassiÞcationofthefrequency-loweringtechniques.InsertscorrespondtotheplotsshowninFig.11. Figure13Schematicofhowinformationinthespeechsignalisaffectedbythedistortionintroducedbyfrequencylowering.Thedashedlinerepresentsthepotentialincreaseofinformationgainedbymovingotherwiseinaudiblehigh-frequencycontenttolower-frequencyregionsofaudibility.ThedottedlinerepresentsthepotentialdecreaseininformationforthatpartofthespectrumthatcanbeampliÞednormally.Thesolidblacklinerepresentsthesumma-tionofinformationgainedandinformationlostattributabletofrequencylowering.ItishypothesizedthattheunderlyingformsofthefunctionswilldependonthefactorsdepictedinFig.2. SEMINARSINHEARING/VOLUME34,NUMBER22013 reportiforhowaudibilityforthefrequency-loweredspeechwasobtained,therebylimitinginferencesorcomparisonsonemightdraw.TheotherextremethatmustbeguardedagainstisunintentionallylimitingtheaudiblebandwidthbychoosinganoverlyaggressiveFLsetting.Afterall,thetreatmentshouldfollowtheprincipleofÒdonoharm.ÓFig.13illustratesthepossiblenegativeeffectsofFL.Shownbythesolidblackline,asFLismademoreaggressiveandabroaderinputbandwidthisaffected,apatientÕsabilitytoextractinforma-tionfromtheloweredsignaldecreases.NotonlydoestheÒinformationvalueÓoftherecodedhigh-frequencycontentdecrease,buttheinfor-mationfromlow-frequencycontentisprogres-sivelydegradedaswell.Theexactdetailswill,ofcourse,dependonthespeciÞctechnology.AdultsbeingÞtwiththisFLwillverballyobjectiftheyfeelthattheyarelosingtoomuchinformation,butyoungchildrencannot.ThisriskofimpedinginformationextractionhasleadmanufacturerstointentionallylimittheadjustmentparametersassociatedwithFLand/orprovidethedispenserwithrecommendedcandidacyguidelines.RegardlessofwhoisbeingÞtwithFLhearingaids,aguidelinetheauthorrecommendsistouseprobemicro-phonemeasurementstoobtainthemaximumaudiblefrequencyafterÞne-tuningthehearingaidwithFL,andthendothesamewithFLactivatedusingthesettingsunderconsiderationtoensurethataudiblebandwidthisnotlimitedbyFL.ToassistthedispenserinchoosingFLsettingsforNFCandLFT,theauthorhasdevelopedonlinetoolsthatvisuallyplothowfrequenciesarealteredbythedifferentsettings(availableatwww.tinyURL.com).ForNFC,thebasicprinciplescanbevisualizedinFig.8.Foragivenaudiogram,onlyalimitednumberofcombinationsofstartfrequencyandcom-pressionratioareavailabletothedispenser.Inthisexample,wewillassumethatthemaxi-mumaudiblefrequencyintheoutputthatcanbeobtainedwithNFCdeactivatedis3.6kHz,asindicatedbythedashedline.FromtheÞgure,itshouldbeclearthatonlyaselectnumberoftheavailableoptions,asindicatedbytheasterisks,avoidthetwoextremesjustdescribed.Ontheoneextreme,noneoftheFLsignalwillbeaudibleforstart3.6kHz.Ontheotherextreme,settingswithstartfrequencies1.7kHzbe-gintorestricttheaudiblebandwidthoftheoutput.Oftheremainingsettings,thedis-pensermightreasonablychoosetheonethatobjectivelymaximizesthebandwidthoftheinputsignalthatismadeavailableintheoutputafterFLortheonethatissubjectivelymostpleasingtothelistener.ForLFT,inadditiontothesetwoextremes,onemightwanttoconsiderhowtheinformationintheinputisrepackagedintheoutput.Forexample,Fig.5Ashowshowalowstartfrequencymightleadtolessthanoptimaloutcomesbecausetoomanyfrequenciesintheinputareoverrepresentedintheoutputduetotheoverlapbetweentheuntransposed,transposed,andexpandedtransposedsignals.Fig.5Bontheotherhand,showstheoppositeinwhichahighstartfrequencycanleadtoanislandofintermediateinputfrequenciesthatarestillinaudibleafterFL.Fig.5Cshowsanidealscenarioinwhichtheinputfrequencycorrespondingtowhereaudibilityforanuntransposedsignalendsisclosetothestartfrequencyfortranspositionandwhereaudibilityforthissignalendsisclosetothestartfrequencyforexpandedtransposition.SomeDataTwostudiesbytheauthorhighlighthowthechoiceoftechnologyanditssettingscaninßu-enceoutcomesfordifferenthearinglosses.AlexanderetalinvestigatedtheefÞcacyofLFTandNFCusing24adultswithmildtomoderateSNHLand24normal-hearingcon-Participantslistenedmonaurallythroughheadphonestoaseriesofninefricativesandaffricatesspokenbythreewomeninavowel-consonantcontextthathadbeenmixedwithspeech-shapednoiseat10-dBsignal-to-noiseratioandrecordedthroughhearingaidswithLFTorNFCthatwereprogrammedforamildtomoderatelosssuchthatFLoccurredonlyforinputfrequencies4kHz.ItshouldbenotedthattheusesofLFTandNFCfortheselosseswereoutsidetherecommendedcandidacyguidelinesestablishedbythe RECOGNITIONOFFREQUENCY-LOWEREDSPEECH/ALEXANDER manufacturers.Controlstimuliincludedre-cordingsmadewhileFLwasdeactivatedineachhearingaid(restrictedbandwidthcondi-tion)andthesesamerecordingsmixedwithhigh-passÞlteredversionsoftheinputstimuli,sothataverageaudibilityextendedoutto9kHz(widebandwidthcondition).Consis-tentwithpreviousÞndingsthatdemonstratedtheimportanceofhigh-frequencyinformationforfricativeidentiÞcation,7,19performanceforthehearing-impairedlistenersinthewidebandwidthconditionswassigniÞcantlybetterthantherestrictedbandwidthconditions.In-dividualperformanceforconditionswhereNFCprovidedaudibilityforinputfrequenciesupto8to9kHzwassimilartothewidebandwidthconditions,indicatingthattheben-eÞtobservedwithincreasingbandwidthalsocanbeobtainedusingNFC.Incontrast,per-formancewithLFTforbothhearing-impairedandnormal-hearinglistenerswassigniÞcantlyworsecomparedwiththerestrictedbandwidthconditionbyabout10and20%,respectively.SigniÞcantdifferencesbetweenmostcondi-tionscouldbelargelyattributedtoanincreaseordecreaseinconfusionsforthephonemes/s/and/z/.DifferencesinoutcomesbetweenthetwoFLtechniquesmightbeattributedtothedegreetowhichLFTalteredthelow-frequen-cyspectrumcomparedwithNFCinthispar-ticularpopulationoflisteners.Asdiscussedlater,itispossiblethatthisdifferencewouldhavebecomelessfollowingextendedexperi-encewiththetechnology.UsingsimulatedhearingaidprocessinginMATLAB(i.e.,ÒthePurduehearingaidsim-ulatorÓ)withNFCmodeledafterSimpsonetalandßexiblemultichannelwidedynamicrangecompression,Alexanderinvestigatedtheef-fectofvaryingthestartfrequencyandcom-pressionratiofortwoÞxedoutputTwenty-eightlistenerswithmildtomoderatelysevereSNHLidentiÞedconsonantsandvowelsfromnonsensesyllablesinnoise.Allspeechoutputwaslow-passÞlteredat3.3or5.0kHzacrosstwogroupsoflistenerstocontrolforhigh-frequencythresholdswhensimulatingtwoclinicalsce-narioswherebythedispenserhasavarietyofNFCoptionsforrepackagingdifferentamountsofhigh-frequencyinformationinalimitedbandofaudibility.ForbothgroupstherewassigniÞcantimprovementinfricative/affricateidentiÞcationformostNFCsettingsrelativetothelow-passcontrolconditions.However,whenstartfrequencywaslow(1.6kHz),therewasadecreaseinvowelandnon-fricativeconsonantidentiÞcation.Recognitionofthesesoundsimprovedwhenthestartfrequencywasincreased(2.2kHz)eventhoughthecompressionratioalsohadtobeincreasedtoprovideaudibilitytothesamebandofinputfrequencies,aprocessthatreducedspectralresolutionwithintheFLsignal.Alternatively,whenlesscompressionwasusedwiththelowstartfrequency,vowelandnonfricativeconsonantidentiÞcationwasclosertothatforthelow-passcontrol,al-thoughthiscameattheexpenseofbringinglesshigh-frequencyfricativeinformationdownintotherangeofaudibility(cf.Fig.13).Overall,theresultsofthisstudyindicatethatmanyfactorslikelydeterminehowmuchinformationindividuallistenerscanextractfromfrequency-loweredspeech,includingthefrequencyregionsalteredbyFLandtheseverityoflossintheregionstowhereinformationismoved.INTRINSICFACTORSRELATEDTOTHEABILITYTOLEARNANDBENEFITFROMFREQUENCY-LOWEREDSPEECHThesecondsetoffactorsinFig.2tobediscussedarehighlyinterconnectedandrelatetointrinsiccharacteristicsoftheindividuallistenersandtheirabilitiestomakeuseofthenewandalteredspeechcuesassociatedwithFL.ThesefactorsincludeageofÞtting,ageofhearinglossonset,listeningexperience,andcognitivefactors.Theprecedingsectionindi-catesthateachFLstrategyisaformofspeechrecodingthatuniquelyalterstheinformationcontainedinindividualspeechsoundsattheauditoryperipheryinacomplexmanner.ForlistenerstobeneÞtfromtherecodedspeechinformation,theymustÞrstlearnhowtointer-pretthenewsignal.ForsomeFLsettings,littletonolearningmaybenecessaryandimmediatebeneÞtcanobserved,eveninthelaborato-29,45,46Forexample,ifNFChasahighstart SEMINARSINHEARING/VOLUME34,NUMBER22013 frequency(e.g.,3,500Hz),FLwillprimarilyaffectinformationcontainedintheaperiodichigh-frequencyspectrumbymakingitnar-rowerinbandwidth.BecausetheprocessedsignalcanhaveatleastthesamesoundqualityasconventionalampliÞcation,47,48listenerswithSNHLmaynotevennoticethealterationformostspeechsounds.ThepotentialforbeneÞtisrelatedtotheextentthatthenewinformationmadeavailablewithFLreducesuncertaintyaboutthespokenmessage.ThedegreetowhichlistenershaveuncertaintyintheÞrstplacewilldependonthelinguisticcontextofthemessageandthelistenersÕabili-tiestousethiscontext.Theroleoflinguisticcontextwasdiscussedearlier.ListenersÕabili-tiestousecontexttobootstrapunderstandingofamessagefragmentedbyÒinformationdrop-outsÓ(inaudibilityofcriticalacousticcues)dependsontheirknowledgeofgrammar,se-mantics,andpragmatics.Thisiswherewemightexpecttoseedifferencesbetweenchil-drenandadults.Italsointroducesafurtherdistinctionbetweenwhetherthelossispre-orpostlinguistic,becauselistenerswhohaveau-ditoryexperiencewiththenaturalproductionsoftherecodedsoundsmightadaptmorequicklytothem.ForotherFLsettings,moreimplicitand/orexplicitlearningmaybenecessarybeforefullbeneÞtcanberealized.Thesesettingslikelyinvolveamanipulationoftheprimaryformantsofspeech,whichgenerallyresideinthepartofthespectrumbelow3,500Hzorso.Forthesesettings,alteredsoundqualitymightbeob-servedalongwithnewperceptualconfu-Whetherlistenerscanadapttothealteredsoundqualityisanimportantcon-siderationbecausethetechnologywilllikelyberejectediftheycannot.ThereissomeindicationthatsoundqualityorpreferenceratingscanimproveforbothadultsandchildrenovertimewithmodernFLstrategies.Assumingthatsoundqualityisnotanissue,otherfactorsthatmightdeterminebeneÞtarelistenersÕexperi-encewiththetechnology,whichcanbegainedimplicitlythroughacclimatizationand/orex-plicitlythroughtraining,andthelistenersÕunderlyingcognitiveabilitytoapplytheneces-saryeffortinvolvedwiththeperceptuallearningChildrenversusAdultsChildrenmightexperiencegreaterbeneÞtfromFLcomparedtoadultssimplybecausetheyappeartohaveagreaterÒdeÞcitÓwhenidentify-ingspeechunderidenticalconditions.Inotherwords,inconditionswhereadultsareperform-ingneartheceilingoftheirabilitiesusingconventionalampliÞcation,childrenmightstillbeabletobeneÞtfromadditionalinformationgainedviafrequency-loweredspeech.Asmen-tioned,whenlinguisticknowledgeconfersanadvantage,adultÐchilddifferencesinspeechrecognitionmightbeexpected.Forexample,whenidentifyingwordsorsentencesinnoise,childrenrequiremorefavorablesignal-to-noiseratiosforsimilarperformance.53Ð57However,evenwhenlinguisticknowledgeisoflittleapparentvalue,aswhenidentifyingnonsensesyllables,childrenstillunderperformadultsandrequirehigherlevelsofaudibility.18,58Ð61Rea-sonsforthisarebeyondthescopeofthisarticle,butlikelyincludedifferencesinphonologicaldevelopmentandphonotacticknowledge.Thatis,adultshaverelativelystablecategoriesforthevariousacousticformssoundsinalanguagecantakeandhaveanimplicitknowl-edgeabouttheallowablesoundsequences.Thishelpstorestricttherangeofpossibleresponseoptionsevenwhennonwordsaretested.MorerelevanttothisdiscussionareÞndingsthatsuggestthatchildrenbeneÞtmorethanadultswhentheavailabilityofspeechinformationisaugmentedbyincreasesinsignalbandwidth.7,64WhetherthisÞndingholdswhenthesameinformationismadeavailableviaFLhasyettobedemonstrated.Onlyasmallnumberofstudieshavetestedadultsandchildrenusingidenticalmethodolo-gies,whichisanimportantevaluationcriterionbecausedifferentprocedures,especiallydifferentmethodsforselectingindividualFLsettings,cansigniÞcantlyaffectoutcomesasnotedpre-viously.Glistaetalreportedoutcomesforbothchildrenandadultswithslopinghigh-frequencySNHLrangingfrommoderatelyseveretopro-found.AwithdrawaldesignwasusedinwhichperformancefollowingexposuretoNFCforaminimumof3to4weekswascomparedwithterminalperformancefollowingexposuretoconventionalprocessingforanother4weeksorso.AgegroupwasasigniÞcantpredictorof RECOGNITIONOFFREQUENCY-LOWEREDSPEECH/ALEXANDER performanceonatestofpluralrecognition,with4of13adultsand7of11childrenshowingsigniÞcantimprovementwithFL.However,itisunknownwhetherthisresultisduetodevelop-mentaldifferencesortothefactthatchildrenweregiven5-to10-dBmoregainthanadultspertheDesiredSensationLevelv5.0prescrip-tiveguidelines.UsingthePurduehearingaidsimulator,KopunetaltestedadultsandchildrenwithmildtosevereSNHLonmonosyllabicwordrecognitionbeforeandafter22minutesofau-dio-visualexposuretotwochildrenÕsstoriesthathadbeenprocessedwithcustomizedampliÞca-tionandNFCprocessing.TheyfoundthatbothchildrenandadultsperformedsigniÞcantlybetterbeforeandafterexposurewithNFCcomparedtowithout.ExposuredidnotprovideadditionalbeneÞtforNFCbecauseperformancealsoim-provedfortheprocessingconditionwithoutNFC.Importantly,theydidnotÞndasigniÞcantdifferenceinbeneÞtforchildrenandadults.Inastudythatexclusivelyexaminedsound-qualityjudgmentsforspeechandmusicfollow-ingprocessingusingthePurduehearingaidsimulatorwithNFCandwithoutNFCattwobandwidths,BrennanetaldidnotÞndprefer-encedifferencesbetweenadultsandchildrenwithmildtosevereSNHL.Forspeechstim-uli,bothagegroupspreferredwidebandwidthprocessing(10,000Hz)orNFCmoreoftenthanrestrictedbandwidthprocessing(5,000Hz).ThispreferencewaspositivelycorrelatedwithampliÞcationexperience,buttherewasnorelationshipwiththeaudiogram.Formusicstimuli,nosigniÞcantdifferenceswerefoundbetweenthethreeconditions.ItshouldbenotedthatthestartfrequencyofNFCwas3,000Hzformostlisteners,whichlimitedtheprocessingtofrequenciesthatdonotcon-tributemuchtopitch.Theseresultsaresome-whatincontrasttothoseofGlistaetal,foundthatchildrenhadagreaterpreferenceforNFCthanadults,andtothoseofAuriemmoetwhonotedthatchildrenmoreoftenpre-ferredLFTthanadultsafterinitialÞtting,butnotafter2weeksofusageduetoincreasesinpreferencebytheadults.Insummary,despiteseveralvalidreasonstoexpectadultÐchilddifferencesinbeneÞtand/orpreferenceforFL,supportingevidenceisnotavailablewithinstudiesthatusedamatchedexperimentaldesignacrossagegroups.AgeofHearingLossOnsetDiscussionofchildÐadultdifferencesimpliesmorethanauditoryandlinguisticdevelopment.Mostoftheadultsinthestudiesdescribedintheprecedingsectionhadacquiredlosses(namelypresbycusisandnoiseexposure),andmostofthechildrenhadcongenitallosses(e.g.,geneticcauses,meningitis,hyperbilirubinemia,anoxiaatbirth,etc.).Differencesinetiologyhaveimplicationsforthedegreeofinvolvementofthevariouscochlearstructures(e.g.,outerversusinnerhaircellloss)andtheaudiometricconÞgurationÑfactorsdiscussedintheÞrstpartofthisarticle.Forexample,PittmanandStelmachowiczanalyzedalmost500audio-gramsfromaclinicalpopulationandfoundthatchildrencomparedwithadultshadagreatervarietyofaudiometricconÞgurationsotherthantheclassicslopingloss,hadgreatervariabilityinthresholdsacrossfrequency,andhadagreaterprevalenceofasymmetricalhear-inglosses.Asdiscussed,audiometricdiffer-encesinßuencetheamountofunaideddeÞcitforindividualspeechsoundsandthechoiceofFLtechnologyanditssettings.Anotherfactortoconsiderinthisdiscus-sionistheroleofprelinguisticversuspostlin-guistichearingloss.IthasbeendocumentedthatyoungchildrenwhoaredeprivedofrichauditoryinputbecauseofuntreatedSNHLorunaidablehigh-frequencySNHLoftenhavedelayedphonologicaldevelopmentcomparedwiththeirnormal-hearingpeers.otherwords,theacousticpropertiesthathelptocategorizeonesoundasbeingdifferentfromanotherarelessdeÞned(acoustic-phoneticboundariesaremorevariable),resultinginperceptualconfusionsandexpressivedeÞcits.Itisunknownwhateffectthiscouldhaveonperceptionoffrequency-loweredspeechrelativetoconventionalhearingaidprocessing,buttheamountofdeÞcitandtheabilitytolearnthenewspeechcodearelikelymediatingfactors.Forexample,listenerswithbetterphonologicalrepresentationsofthephonemes/s/and/z/duetopriorauditoryexperiencemighthavefewerdeÞcitswithconventionalhearingaids,hence SEMINARSINHEARING/VOLUME34,NUMBER22013 lessroomforimprovementasperformancenearstheceiling.Theselistenersmightalsobeabletolearntousethenewandalteredspeechcuesmorequickly,whereasthosewithpoorerphonologicalrepresentationsmightre-quireanextendedperiodtoacclimatizeand/orrequireexplicittraining.AcclimatizationandTrainingPerceptualacclimatizationreferstothebywhichindividualsadapttoalteredsensoryinputtomaintainoptimumperformanceintheirenvironment.Forspeechperception,theeffectsofacclimatizationareoftengaugedbyimprovementinrecognitionscoresoverbuttheyalsocouldincludegradualimprovementsinratedsoundquality,decreasedlisteningeffort,andsoon.Thelatterareoftenoverlooked,butareimportanttoconsiderbe-causeinitialreactionstofrequency-loweredspeech,whichcontributetotheoverallexperi-ence,mightexhibitevengreaterchangeovertimecomparedwithconventionalampliÞcationbecauseoftheincreasedamountofsignalalteration.Theseconsiderationsarementionedbrießyelsewhereinthisarticle.Twostudiesthathaveexplicitlyexam-inedacclimatizationeffectsforfrequency-low-eredspeechwithouttraininghavebeenconductedonchildren.BothstudiesreportedsigniÞcantimprovementsinspeechrecognitionfollowingextendedexperiencelisteningthroughhearingaidswithNFC.Wolfeetalpresentedresultsfroma6-monthfollow-upon15youngchildren(ages5to13years)withmoderatetomoderatelysevereSNHLwhowereinitiallytestedaftertwocounterbalanced6-weekintervalsinwhichtheylistenedthroughthesamehearingaidswithNFCactivatedordeactivatedduringtheirdailyroutines.Discriminationofsingular/pluralcontrastsinquiet,whichtestsperceptionof/s/and/z/,anddiscriminationofhigh-frequencyconsonantsembeddedinnonsensevowel-consonant-vowelutterancesinquietweresigniÞcantlyimprovedbyNFCafter6weeks.Sentencerecognitioninmultitalkerbabbledidnotimproveduringthistime,however.Whentestedagainat6months,pluraldiscriminationmaintainedatceilingper-formancelevelsandtherewerecontinuedim-provementsinthediscriminationofhigh-frequencyconsonantsrelativetotestingat6weeks.Interestingly,therewassigniÞcantimprovementthesignal-to-noiseratiothresh-oldforsentencerecognitionthatwasnotpres-entafter6weeks.Theseresultsindicatethatsomechildrenmayundergoanacclimatizationperiodlastingatleast6monthsforNFC.However,maturationand/orlearningeffectscannotberuledoutascontributingfactorsintheWolfeetalstudybecausetherewasnocontrolconditionorwithdrawalconditionwherebylistenersweretestedfollowinganacclimatizationperiodwithNFCdeactivated.Tominimizethelikelihoodofmaturationeffects,Glistaetaltested6olderchildren(ages11to18years)withatleastmoderatelyseverehigh-frequencySNHLinasingle-subjectde-signonphonemedetection,pluralrecognition,discriminationbetween/s/and//,andconso-nantrecognitionatregularintervalsovera4-monthperiod.Onelistenerdidnotshowimprovementovertimeduetoceilingeffects.FortheotherÞvelisteners,theyfoundsigniÞ-cantimprovement(upto20%)withNFConatleastonetest,especiallywithfricatives,althoughthepatternofimprovementovertimewasquitevariablewithinandbetweenlisteners.There-fore,theauthorscautionaboutextrapolatingfromindividualresultsandwarnthatmatura-tioneffectsarestillpossible,althoughunlikely.Explicitauditorytrainingisanotherway,andinsomecasesmaybetheonlyway,thatlistenerscangettheexperiencetheyneedtolearnthenewandalteredspeechcuesintro-ducedbyFL.Theconceptoftrainingonfrequency-loweredspeechgoesbacktotheearliesttechniquesdescribedintheliterature,whereitwasmorecommonplacethanmoderntechniques(seeSimpsonforareview).Onereasonforthisshiftinapproachesmaybethatthemoderntechniquesdescribedintherecentliteratureareusuallyimplementedinwearabledevices,whichprovideslistenerswithanop-portunitytoacclimatizetotheprocessing.Theonlyexceptionsthatincorporatedformaltrain-ingproceduresutilizedLFT.Auriemmoetaltested10children(ages6to13years)withseveretoprofoundSNHLonconsonantandvowelrecognitioninquietbeforeandaftera3-weekcontrolperiodduringwhich RECOGNITIONOFFREQUENCY-LOWEREDSPEECH/ALEXANDER theyworeahearingaidwithLFTdeactivatedduringtheirdailyroutines.Thentheyrepeat-edthesametestingaftertwoadditional3-weektreatmentperiodsduringwhichLFTinthehearingaidwasactivated.Throughoutthecontrolandtreatmentphasesoftheexperiment,listenersattendedweeklyauditorytrainingses-sions.ThemainÞndingswerethatduringthe6-weekperiodwhenLFTwasactivated,conso-nantrecognitionatalowpresentationlevel(30-dBhearinglevel)signiÞcantlyimprovedbymorethan20%andtheproductionaccuracyof/s/and/z/improvedbymorethan10%.TherewasnosigniÞcantdifferenceinvowelrecognitionwhenLFTwasactivatedversusdeactivatedbecauseperformancewasnearceil-inglevelsduringinitialtestingforbothcon-ditions.BecauselistenersreceivedtrainingalongwiththeirdailyexposuretoLFT,itcannotbedeterminediftheobservedeffectswereduetoacclimatizationand/ortotraining.KuketaltestedeightadultswithseveretoprofoundSNHLonconsonantrecognitioninquietandinbabblenoise,ÞrstwithLFTdeactivatedandactivatedaftertheinitialÞt,thenafter1monthofdailyexposureandtrain-ing(20to30minutesperaday),andthenafteranothermonthofdailyexposurebutwithouttraining.AfeatureanalysisoftheconfusionsrevealedthatthemainsourceofimprovementwithLFTwasfricatives.Atapresentationlevelof50-dBsoundpressurelevel(SPL),fricativerecognitionimproved5to10%initiallyandthenanadditional10%after2months.Atapresentationlevelof68-dBSPL,asigniÞcantdifferencewasonlyseenforfricatives,whichimprovedby10%after2months.Atthislevel,therewasalsoadecreaseinrecognitionofstopconsonantsaftertheinitialÞt,whichsigniÞcantlyimprovedafter2monthstoperformancelevelsslightlybetterthanbaseline.Forspeechinnoise,theonlysigniÞcantdifferencewasabouta15%improve-mentforfricativesafter2monthsoftraining.LiketheAuriemmoetalstudy,onecannotdetermineiftheobservedeffectswereduetoacclimatizationand/ortotraining,becauselis-tenersreceivedtrainingalongwiththeirdailyexposuretoLFT.Furthermore,becausetherewasnocontrolgroupthattrainedonthesamehearingaidwithLFTdeactivated,onecannotdetermineiftheobservedmaineffectsareduetotheadditionalinformationintroducedbyLFTortosimplepracticeeffects.Toaddressthisconcern,abrieffollow-upreportbyKukandKeenanindicatesthattherewerenotrain-ingeffectspersewhenusingthehearingaidswithLFTdeactivated.Insummary,thefewreportsthathaveexaminedtheeffectsofextendedlisteningexperiencewithfrequency-loweredspeechsup-portthenotionthatatleastsomelistenersrequireaperiodofacclimatizationand/ortrain-ingwiththenewspeechcodebeforefullbeneÞtcanberealized.OnecanspeculatethatthemoredrasticthealterationofspeechcuesfollowingFL,namelyinvolvementoflow-frequencyspeechcues,thegreatertheneedforextendedorexplicitlisteningexperienceforthelistenertolearnhowtomosteffectivelyprocessthenewlyintroducedandalteredinformation.CognitiveFactorsTherehasbeenrecentinterestinunderstandinghowcognitiveprocessesinßuenceoutcomesforhearingaidalgorithmsotherthanFL(seeLunneretalforareview).Cognitiveprocess-inghasbeenimplicatedinatleasttwoways.Onewayisasamoderatingvariable(e.g.,verbalworkingmemory)thatinßuenceswhobeneÞtsandunderwhatcircumstances,aswithwidedynamicrangecompression.Anotherwayisasadependentvariable;thatis,theconstructbywhichbeneÞtismeasured(e.g.,intermsoflisteningeffort,recall,reactiontime),aswithdigitalnoisereduction.Asnotedinthepreviousparagraphs,FLtechniquesandset-tingsthataltermoreofthenaturalspeechcodemayrequirealearningperiodconsistingofexposurewithorwithouttrainingtoachievemaximumpotential.Ontheonehand,iflearn-ingputsahighdemandoncognitiveresources,thenFLmightnotbebeneÞcialormightevenbedetrimentaltothoselistenerswhohavefewerresourcesavailable,suchastheelderly.theotherhand,ifFLcanprovideadditionalinformationthathelpsreduceuncertaintyaboutthespokenmessagewithouttheneedforlearn-ing,thenitmightbeabletoreducethecogni-tiveloadassociatedwithlisteninginreal-worldscenarios,therebybeneÞtingthesesame SEMINARSINHEARING/VOLUME34,NUMBER22013 listenersintermsoflisteningeffort,recall,comprehension,multitasking,andsoon.Tworecentstudieshaveexploredtherela-tionshipbetweencognitiveprocessingandrec-ognitionofspeechwithNFC.Usingolderadults(ages62to92years)withmildtosevereslopingSNHL,Arehartetalmeasuredwork-ingmemoryandrecognitionofsentencesinnoisethathadbeenprocessedwithahearingaidsimulatorthatimplementedaformofNFCsimilartothatdescribedbySimpsonetal.Thesignalprocessingdifferedinthatonlythemostintensepeaksinthehigh-frequencyspec-trumwerecompressed,insteadoftheentirehigh-frequencyspectrum.NFCsettingswereintentionallyaggressivewithstartfrequenciesaslowas1,000Hzbecausetheprimarypurposewastoobtaincorrelatesforspeechrecognitionofdistortedspeech,notFLperse.Astheamountofdistortionincreased,speechrecog-nitionwasmoreadverselyaffectedinlistenerswiththepoorestworkingmemoryabilitiescomparedwiththosewithbetterworkingmemoryabilities.Incontrast,EllisandusedsimilartestmaterialsandthesamecognitivetestasArehartetalbutdidnotÞndasigniÞcantrelationshipbetweentheworkingmemoryofyoungnormal-hearingadults(ages18to50years)andrecognitionofspeechwithanNFCstartfrequencyof1,600Hz.Interestingly,theyfoundasigniÞcantrelationshipbetweenworkingmemoryandrecognitionoftheunprocessedspeech,whichindicatesthatthevariationinworkingmemoryabilitiesforthisnonclinicalpopulationshouldhavebeensufÞcientlylargetocaptureaneffectforNFCprocessinghadtheretrulybeenone.Furtherresearchisclearlyneededtobetterunderstandtherolethatcognitionplaysinprocessingfrequency-loweredspeechandhowthisinßuencescandidacyforthistechnology.Afterreviewingallthestudiesmentionedinthisarticle,itshouldbeclearthatdifferencesinoutcomeswithFLvaryalongwiththedepen-dencyofthetestmaterialsandindividuallisten-ersoninformationinthehigh-frequencyspectrum.Accordingly,whenFLshowsabeneÞtforspeechrecognitionitisprimarilyforfricativeconsonantswithrestrictedlinguisticcontext.NotalloftheindividualsinthestudiesreviewedshowedimprovementinspeechrecognitionwithFL.Formostofthese,speechrecognitionwithFLwasnotstatisticallydifferentfromspeechrecognitionwithoutFL,whichmightoffersomeassurancefromaÒdonoharmÓperspective.However,forsomelisteners,FLnegativelyaffectedspeechrecognition.InlightofthemodelpresentedinFig.13,onemightquestionwheth-erthesettingsinthesecasesactuallyoptimizedthetotalinformationinthesignal.Inaddition,fullbeneÞtmightnotberealizeduntilafterseveralmonthsofexperiencewiththetechnolo-gy,especiallyforspeechinnoise.AlthoughthefocusofmeasuringbeneÞthasbeenonspeechrecognition,oneshouldbeopentothepossibilitythatbeneÞtfromFLmightbelessevidentintermsofdecreasedlisteningeffort,improvedspeechproductionandvocalquality,improvedlocalizationandspatialunmasking,andsoon,evenwhentherearenoobservableimprovementsinspeechrecognitionasmeasuredinthelaboratory.Finally,thereaderisremindedoftheneedto(1)understandthespeciÞcFLtechnologybeforeimplementingit(i.e.,whatfrequenciesarelow-ered,wherearetheyloweredto,whenaretheylowered,andhowaretheylowered),(2)verifyelectroacousticallythatFLmovesadditionalspeechinformationtoaregionofaudibility,butdoesnotundulylimittheaudiblebandwidthoftheentiresignal,and(3)validateoutcomestoensureeffectivenessfortheindividual.VarshaHariramhelpedpreparethereferencesandtheinformationinTable1,althoughanyerrorscontainedthereinaresolelytherespon-sibilityoftheauthor.1.HoganCA,TurnerCW.High-frequencyaudibil-ity:beneÞtsforhearing-impairedlisteners.JAcoustSocAm1998;104:432Ð4412.ChingTY,DillonH,ByrneD.Speechrecogni-tionofhearing-impairedlisteners:predictionsfromaudibilityandthelimitedroleofhigh-frequencyampliÞcation.JAcoustSocAm1998;103:1128Ð1140 RECOGNITIONOFFREQUENCY-LOWEREDSPEECH/ALEXANDER 3.TurnerCW,CummingsKJ.Speechaudibilityforlistenerswithhigh-frequencyhearingloss.AmJAudiol1999;8:47Ð564.BaerT,MooreBCJ,KlukK.EffectsoflowpassÞlteringontheintelligibilityofspeechinnoiseforpeoplewithandwithoutdeadregionsathighfrequencies.JAcoustSocAm2002;112(3Pt1):5.VickersDA,MooreBCJ,BaerT.Effectsoflow-passÞlteringontheintelligibilityofspeechinquietforpeoplewithandwithoutdeadregionsathighfrequencies.JAcoustSocAm2001;110:1164Ð11756.BoothroydA,MedwetskyL.Spectraldistributionof/s/andthefrequencyresponseofhearingaids.EarHear1992;13:150Ð1577.StelmachowiczPG,PittmanAL,HooverBM,LewisDE.Effectofstimulusbandwidthontheperceptionof/s/innormal-andhearing-impairedchildrenandadults.JAcoustSocAm2001;110:8.FoxRA,NissenSL.Sex-relatedacousticchangesinvoicelessEnglishfricatives.JSpeechLangHearRes2005;48:753Ð7659.MannVA,ReppBH.Inßuenceofvocaliccontextonperceptionofthe/s/-/s/distinction:spectralfactors.PerceptPsychophys1980;28:213Ð22810.WhalenDH.EffectsofvocalicformanttransitionsandvowelqualityontheEnglish[s]-[s]boundary.JAcoustSocAm1981;69:275Ð28211.HeinzJM,StevensKN.Onthepropertiesofvoicelessfricativeconsonants.JAcoustSocAm12.NittrouerS,Studdert-KennedyM.Theroleofcoarticulatoryeffectsintheperceptionoffricativesbychildrenandadults.JSpeechHearRes13.ZengFG,TurnerCW.Recognitionofvoicelessfricativesbynormalandhearing-impairedsubjects.JSpeechHearRes1990;33:440Ð44914.NittrouerS.Age-relateddifferencesinperceptualeffectsofformanttransitionswithinsyllablesandacrosssyllableboundaries.JPhonetics1992;20:15.NittrouerS,MillerME.Predictingdevelopmentalshiftsinperceptualweightingschemes.JAcoustSocAm1997;101:2253Ð226616.NittrouerS,MillerME.Developmentalweightingshiftsfornoisecomponentsoffricative-vowelsyllables.JAcoustSocAm1997;102:572Ð58017.HedrickMS.Effectofacousticcuesonlabelingfricativesandaffricates.JSpeechLangHearRes18.PittmanAL,StelmachowiczPG.Perceptionofvoicelessfricativesbynormal-hearingandhear-ing-impairedchildrenandadults.JSpeechLangHearRes2000;43:1389Ð140119.StelmachowiczPG,LewisDE,ChoiS,HooverB.Effectofstimulusbandwidthonauditoryskillsinnormal-hearingandhearing-impairedchildren.EarHear2007;28:483Ð49420.PittmanAL.Short-termword-learningrateinchildrenwithnormalhearingandchildrenwithhearinglossinlimitedandextendedhigh-frequen-cybandwidths.JSpeechLangHearRes2008;51:21.HornsbyBWY,JohnsonEE,PicouE.EffectsofdegreeandconÞgurationofhearinglossonthecontributionofhigh-andlow-frequencyspeechinformationtobilateralspeechunderstanding.EarHear2011;32:543Ð55522.StelmachowiczPG,PittmanAL,HooverBM,LewisDE.Aidedperceptionof/s/and/z/byhearing-impairedchildren.EarHear2002;23:23.MoellerMP,HooverBM,PutmanCA,etal.Vocalizationsofinfantswithhearinglosscomparedwithinfantswithnormalhearing:PartIÑphoneticdevelopment.EarHear2007;28:605Ð62724.MoellerMP,McClearyE,PutmanC,Tyler-KringsA,HooverB,StelmachowiczP.Longitu-dinaldevelopmentofphonologyandmorphologyinchildrenwithlate-identiÞedmild-moderatesensorineuralhearingloss.EarHear2010;31:25.WolfeJ,JohnA,SchaferEC,etal.Long-termeffectsofnon-linearfrequencycompressionforchildrenwithmoderatehearingloss.IntJAudiol26.RudminF.Thewhyandhowofhearing/s/.VoltaReview1983;85:263Ð26927.ElfenbeinJL,Hardin-JonesMA,DavisJM.Oralcommunicationskillsofchildrenwhoarehardofhearing.JSpeechHearRes1994;37:28.SimpsonA.Frequency-loweringdevicesforman-aginghigh-frequencyhearingloss:areview.TrendsAmplif2009;13:87Ð10629.KukF,KeenanD,KorhonenP,LauCC.EfÞcacyoflinearfrequencytranspositiononconsonantidentiÞcationinquietandinnoise.JAmAcadAudiol2009;20:465Ð47930.SermanM,HannemanR,KornagelU.Whitepaper:Miconfrequencycompression.SiemensAG,2012.Availableat:http://hearing/siemens.com/Resources/literatures/Global/publications/2012%20-%20white-paper%20mocon%20frequcny%20compression.pdf?_blob=publicationÞle.AccessedonMarch19,201331.BraidaLD,DurlachNI,LippmannRP,HicksBL,RabinowitzWM,ReedCM.HearingaidsÑareviewofpastresearchonlinearampliÞcation,amplitudecompression,andfrequencylowering.ASHAMonogr1979;19:1Ð11432.ParentTC,ChmielR,JergerJ.Comparisonofperformancewithfrequencytranspositionhearingaidsandconventionalhearingaids.JAmAcadAudiol1998;9:67Ð77 SEMINARSINHEARING/VOLUME34,NUMBER22013 33.McDermottHJ,DorkosVP,DeanMR,ChingTYC.ImprovementsinspeechperceptionwithuseoftheAVRTranSonicfrequency-transposinghearingaid.JSpeechLangHearRes1999;42:34.SimpsonA,HersbachAA,McDermottHJ.Im-provementsinspeechperceptionwithanexperi-mentalnonlinearfrequencycompressionhearingdevice.IntJAudiol2005;44:281Ð29235.SimpsonA,HersbachAA,McDermottHJ.Fre-quency-compressionoutcomesinlistenerswithsteeplyslopingaudiograms.IntJAudiol2006;45:36.TurnerCW,RobbMP.Audibilityandrecognitionofstopconsonantsinnormalandhearing-impairedsubjects.JAcoustSocAm1987;81:1566Ð157337.DubnoJR,DirksDD,EllisonDE.Stop-consonantrecognitionfornormal-hearinglistenersandlis-tenerswithhigh-frequencyhearingloss.I:Thecontributionofselectedfrequencyregions.JAcoustSocAm1989;85:347Ð35438.HorwitzAR,DubnoJR,AhlstromJB.Recognitionoflow-pass-Þlteredconsonantsinnoisewithnor-malandimpairedhigh-frequencyhearing.JAcoustSocAm2002;111(1Pt1):409Ð41639.KukF,KorhonenP,PeetersH,KeenanD,JessenA,AndersenH.Linearfrequencytransposition:extendingtheaudibilityofhighfrequencyinfor-mation.HearingReview2006;13:42Ð4840.GlistaD,ScollieS,BagattoM,SeewaldR,ParsaV,JohnsonA.Evaluationofnonlinearfrequencycom-pression:clinicaloutcomes.IntJAudiol2009;48:632Ð64441.AuriemmoJ,KukF,LauC,etal.Effectoflinearfrequencytranspositiononspeechrecognitionandproductionofschool-agechildren.JAmAcadAudiol2009;20:289Ð30542.AlexanderJM,LewisDE,KopunJG,McCreeryRW,StelmachowiczPG.Effectsoffrequencyloweringinwearabledevicesonfricativeandaffri-cateperoeption,2008InternationalHearingAidConference,August13Ð16,2008;LakeTahoe,CA43.KukF,JessenA,BaekgaardL.Ensuringhigh-Þdelityinhearingaidsoundprocessing.HearingReview2009;16:34Ð4344.ANSI.ANSIS3.5Ð1997(R2007),MethodsforCalculationoftheSpeechIntelligibilityIndex.NewYork,NY:AmericanNationalStandards45.McCreeryRW,BrennanMA,HooverB,KopunJ,StelmachowiczPG.Maximizingaudibilityandspeechrecognitionwithnonlinearfrequencycom-pressionbyestimatingaudiblebandwidth.EarHear2012;20:1Ð446.AlexanderJM.Nonlinearfrequencycompression:balancingstartfrequencyandcompressionratio.Paperpresentedat:39thAnnualmeetingoftheAmericanAuditorySociety;March8Ð10,2012;Scottsdale,AZ47.BohnertA,NyffelerM,KeilmannA.Advantagesofanon-linearfrequencycompressionalgorithminnoise.EurArchOtorhinolaryngol2010;267:1045Ð48.BrennanM,McCreeryR,KopunJ,HooverB,StelmachowiczPG.Signalprocessingpreferenceformusicandspeechinadultsandchildrenwithhearingloss,39thAnnualmeetingoftheAmericanAuditorySociety,March8Ð10,2012;Scottsdale,49.HillenbrandJ,GettyLA,ClarkMJ,WheelerK.AcousticcharacteristicsofAmericanEnglishvow-els.JAcoustSocAm1995;97(5Pt1):3099Ð311150.KukF,PeetersH,KeenanD,LauC.Useoffrequencytranspositiononinthin-tube,open-earÞttings.HearingJournal2007;60:59Ð6351.GlistaD,ScollieS,SulkersJ.Perceptualacclimati-zationpostnonlinearfrequencycompressionhear-ingaidÞttinginolderchildren.JSpeechLangHearRes2012;55:1765Ð178752.NyffelerM.StudyÞndsthatnon-linearfrequencycompressionboostsspeechintelligibility.HearingJournal2008;61:22Ð2653.NittrouerS,BoothroydA.Contexteffectsinphonemeandwordrecognitionbyyoungchildrenandolderadults.JAcoustSocAm1990;87:2705Ð54.GravelJS,FauselN,LiskowC,ChobotJ.Child-renÕsspeechrecognitioninnoiseusingomni-direc-tionalanddual-microphonehearingaidtechnology.EarHear1999;20:1Ð1155.FallonM,TrehubSE,SchneiderBA.ChildrenÕsuseofsemanticcuesindegradedlisteningenviron-ments.JAcoustSocAm2002;111(5Pt1):2242Ð56.HallJWIII,GroseJH,BussE,DevMB.Spondeerecognitioninatwo-talkermaskerandaspeech-shapednoisemaskerinadultsandchildren.EarHear2002;23:159Ð16557.BoninoAY,LeiboldLJ,BussE.Releasefromperceptualmaskingforchildrenandadults:beneÞtofacarrierphrase.EarHear2013;34:3Ð1458.StelmachowiczPG,HooverBM,LewisDE,Kor-tekaasRWL,PittmanAL.Therelationbetweenstimuluscontext,speechaudibility,andperceptionfornormal-hearingandhearing-impairedchildren.JSpeechLangHearRes2000;43:902Ð91459.ScollieSD.ChildrenÕsspeechrecognitionscores:theSpeechIntelligibilityIndexandproÞciencyfactorsforageandhearinglevel.EarHear2008;60.NishiK,LewisDE,HooverBM,ChoiS,Stelma-chowiczPG.ChildrenÕsrecognitionofAmericanEnglishconsonantsinnoise.JAcoustSocAm RECOGNITIONOFFREQUENCY-LOWEREDSPEECH/ALEXANDER 61.McCreeryRW,StelmachowiczPG.Audibility-basedpredictionsofspeechrecognitionforchildrenandadultswithnormalhearing.JAcoustSocAm62.Hnath-ChisolmTE,LaipplyE,BoothroydA.Age-relatedchangesonachildrenÕstestofsenso-ry-levelspeechperceptioncapacity.JSpeechLangHearRes1998;41:94Ð10663.EdwardsJ,BeckmanME,MunsonB.Theinter-actionbetweenvocabularysizeandphonotacticprobabilityeffectsonchildrenÕsproductionaccura-cyandßuencyinnonwordrepetition.JSpeechLangHearRes2004;47:421Ð43664.KortekaasRWL,StelmachowiczPG.BandwidtheffectsonchildrenÕsperceptionoftheinßectionalmorpheme/s/:acousticalmeasurements,auditorydetection,andclarityrating.JSpeechLangHearRes2000;43:645Ð66065.ScollieS,SeewaldR,CornelisseL,etal.TheDesiredSensationLevelmultistageinput/outputalgorithm.TrendsAmplif2005;9:159Ð19766.PittmanAL,StelmachowiczPG.Hearinglossinchildrenandadults:audiometricconÞguration,asymmetry,andprogression.EarHear2003;24:67.JergerS,LaiL,MarchmanVA.Picturenamingbychildrenwithhearingloss:I.Effectofsemanticallyrelatedauditorydistractors.JAmAcadAudiol68.EisenbergLS.Currentstateofknowledge:speechrecognitionandproductioninchildrenwithhear-ingimpairment.EarHear2007;28:766Ð77269.CoxRM,AlexanderGC.MaturationofhearingaidbeneÞt:objectiveandsubjectivemeasurements.EarHear1992;13:131Ð14170.GatehouseS.Thetimecourseandmagnitudeofperceptualacclimatizationtofrequencyresponses:evidencefrommonauralÞttingofhearingaids.JAcoustSocAm1992;92:1258Ð126871.GatehouseS.Roleofperceptualacclimatizationintheselectionoffrequencyresponsesforhearingaids.JAmAcadAudiol1993;4:296Ð30672.BentlerRA,NiebuhrDP,GettaJP,AndersonCV.Longitudinalstudyofhearingaideffectiveness.I:Objectivemeasures.JSpeechHearRes1993;36:73.TurnerCW,HumesLE,BentlerRA,CoxRM.AreviewofpastresearchonchangesinhearingaidbeneÞtovertime.EarHear1996;17(3,Suppl):74.CoxRM,AlexanderGC,TaylorIM,GrayGA.BeneÞtacclimatizationinelderlyhearingaidusers.JAmAcadAudiol1996;7:428Ð44175.HorwitzAR,TurnerCW.ThetimecourseofhearingaidbeneÞt.EarHear1997;18:1Ð1176.SaundersGH,CienkowskiKM.Acclimatizationtohearingaids.EarHear1997;18:129Ð13977.BentlerRA,NiebuhrDP,GettaJP,AndersonCV.Longitudinalstudyofhearingaideffectiveness.II:Subjectivemeasures.JSpeechHearRes1993;36:78.WolfeJ,JohnA,SchaferEC,NyffelerM,BoretzkiM,CarawayT.Evaluationofnonlinearfrequencycompressionforschool-agechildrenwithmoderatetomoderatelyseverehearingloss.JAmAcadAudiol2010;21:618Ð62879.KukF,KeenanD.Frequencytransposition:train-ingisonlyhalfthestory.HearingReview2010;80.LunnerT,RudnerM,RonnbergJ.Cognitionandhearingaids.ScandJPsychol2009;50:395Ð40381.GatehouseS,NaylorG,ElberlingC.LinearandnonlinearhearingaidÞttingsÑ1.Patternsofben-eÞt.IntJAudiol2006;45:130Ð15282.LunnerT,Sundewall-ThorenE.Interactionsbe-tweencognition,compression,andlisteningcon-ditions:effectsonspeech-in-noiseperformanceinatwo-channelhearingaid.JAmAcadAudiol2007;83.CoxRM,XuJ.Shortandlongcompressionreleasetimes:speechunderstanding,real-worldpreferen-ces,andassociationwithcognitiveability.JAmAcadAudiol2010;21:121Ð13884.FooC,RudnerM,RonnbergJ,LunnerT.Recog-nitionofspeechinnoisewithnewhearinginstru-mentcompressionreleasesettingsrequiresexplicitcognitivestorageandprocessingcapacity.JAmAcadAudiol2007;18:618Ð63185.MastersonKM,AlexanderJM.Factorsinßuencingreleasefrommaskingwithfastvs.slowcompres-sion.38thAnnualmeetingoftheAmericanAudi-torySociety,March10Ð12,2011;Scottsdale,AZ86.SarampalisA,KalluriS,EdwardsB,HafterE.Objectivemeasuresoflisteningeffort:effectsofbackgroundnoiseandnoisereduction.JSpeechLangHearRes2009;52:1230Ð124087.PittmanA.ChildrenÕsperformanceincomplexlisteningconditions:effectsofhearinglossanddigitalnoisereduction.JSpeechLangHearRes88.PittmanA.Age-relatedbeneÞtsofdigitalnoisereductionforshort-termwordlearninginchildrenwithhearingloss.JSpeechLangHearRes2011;89.HumesLE.Factorsunderlyingthespeech-recog-nitionperformanceofelderlyhearing-aidwearers.JAcoustSocAm2002;112(3Pt1):1112Ð113290.HumesLE.ThecontributionsofaudibilityandcognitivefactorstothebeneÞtprovidedbyampli-Þedspeechtoolderadults.JAmAcadAudiol91.HumesLE,FloydSS.Measuresofworkingmemo-ry,sequencelearning,andspeechrecognitionintheelderly.JSpeechLangHearRes2005;48:224Ð235 SEMINARSINHEARING/VOLUME34,NUMBER22013 92.SchneiderBA.Howageaffectsauditory-cognitiveinteractionsinspeechcomprehension.AudiolRes93.GosselinPA,GagneJ-P.Olderadultsexpendmorelisteningeffortthanyoungadultsrecognizingaudiovisualspeechinnoise.IntJAudiol2011;94.ArehartKH,SouzaP,BacaR,KatesJM.WorkingMemory,Age,andHearingLoss:SusceptibilitytoHearingAidDistortion.EarHear2013;(Jan):395.EllisRJ,MunroKJ.Doescognitivefunctionpre-dictfrequencycompressedspeechrecognitioninlistenerswithnormalhearingandnormalcogni-tion?IntJAudiol2013;52:14Ð2296.SmithJ,DannM,BrownM.Anevaluationoffrequencytranspositionforhearing-impairedschool-agechildren.DeafnessEducInt2009;97.KopunJ,McCreeryR,HooverB,SpaldingJ,BrennanM,StelmachowiczP.EffectsofExpo-sureonSpeechRecognitionwithNonlinearFrequencyCompression,39thAnnualmeetingoftheAmericanAuditorySociety,March8Ð10,2012;Scottsdale,AZ RECOGNITIONOFFREQUENCY-LOWEREDSPEECH/ALEXANDER IndividualVariabilityinRecognitionofFrequency-LoweredSpeechJoshuaM.Alexander,Ph.D.Frequencyloweringinhearingaidsisnotanewconcept,butmodernadvancesintechnologyhaveallowedittobeperformedmoreefÞcientlyandonselectportionsofthespectrum.Nonlinearfrequencycompressionreducesthefrequencyspacinginabandofhigh-frequencyenergysothatmoreinformationiscarriedintheaudiblebandwidth.Frequencytranspositionandtranslationtechniquesloweronlythepartofthehigh-frequencyspectrumthatlikelycontainsimportantspeechinformation.Theseadvancesmayhelpovercomethelimitedbandwidthinconventionalhearingaids,whichrestrictaccesstohigh-frequencyinformationevenforthosewithmildtomoderatehearingloss.Thisisespeciallyimportantforyoungchildrenlearningspeechandlanguage.Aframeworkisadvancedinwhichfactorsthatinßuenceindividualdifferencesinspeechrecognitioncanbedividedintoextrinsicfactorsthataffecttherepresentationofthefrequency-loweredspeechattheauditoryperiphery,includingthespeciÞctechniqueandthesettingschosenforit,andintrinsicfactorsthatcontributetoanindividualÕsabilitytolearnandbeneÞtfromthissignal.Finally,theimportanceofelectroacousticallyverifyingtheoutputtoavoidtoolittleortoomuchloweringandtheimportanceofvalidatingeffectivenessofoutcomesinindividualusersofthetechnologyareemphasized.Hearingaids,frequencylowering,frequencycompression,frequencytranspositionLearningOutcomes:Asaresultofthisactivity,theparticipantwilldescribehowdifferentfrequency-loweringtechnologiesalterthespeechsignal,howcharacteristicsofthehearingaidwearermightinßuencetheabilitytobeneÞtfromthissignal,andwhattowatchforwhenverifyingoutput. DepartmentofSpeech,Language,andHearingSciences,PurdueUniversity,WestLafayette,Indiana.Addressforcorrespondence:JoshuaM.Alexander,Ph.D.,DepartmentofSpeech,Language,andHearingSciences,PurdueUniversity,WestLafayette,IN47907(e-mail:alexan14@purdue.edu).IndividualVariabilityinAidedOutcomes;GuestEditor,JasonA.Galster,Ph.D.SeminHear2013;34:86Ð109.Copyright2013byThiemeMedicalPublishers,Inc.,333SeventhAvenue,NewYork,NY10001,USA.Tel:+1(212)584-4662.DOI:http://dx.doi.org/10.1055/s-0033-1341346.ISSN0734-0451.