The Objective and Subjective Evaluation of Low … · la expansión en el canal 1 y en los canales 1 y 2; sin embargo, la preferencia no fue significativamente diferente entre la

641

J Am Acad Audiol 18:641–652 (2007)

*Department of Audiology and Speech Pathology, University of Tennessee, Knoxville

Patrick N. Plyler, Ph.D., Department of Audiology and Speech Pathology, University of Tennessee, 578 South StadiumHall, Knoxville, TN 37996-0740; Phone: 865-974-7588; Fax: 865-974-1539; E-mail: [email protected]

The Objective and Subjective Evaluation of Low-Frequency Expansion in Wide Dynamic Range Compression Hearing Instruments

Kristy J. Lowery*Patrick N. Plyler*

Abstract

The effects of low-frequency expansion on the objective and subjectiveevaluation of four channel in-the-ear hearing instruments was investigated. Threeexpansion settings were programmed in each device: expansion off, expansionrestricted to channel one, and expansion restricted to channels one and two.Objective evaluations were conducted in quiet (Connected Speech Test) andin noise (Hearing in Noise Test) with speech levels fixed at 40 dB SPL.Subjectively, each participant rated expansion satisfaction in quiet and listeningto low-level speech in a sound-treated room then indicated the expansioncondition preferred overall. Listeners performed significantly better in quiet andin noise for the Off and Channel 1 conditions than the Channels 1 and 2 condition;however, performance was similar between the Off and Channel 1 conditions.Expansion effects on listener satisfaction ratings depended on the listeningenvironment. Overall, listeners preferred expansion in Channel 1 to expansionin Channels 1 and 2; however, preference was not significantly differentbetween the Channel 1 and Off conditions. Results indicate restrictingexpansion to 1000 Hz overcomes speech-recognition deficits observed withexpansion active across a broader spectrum without significantly reducingsubjective benefit or preference.

Key Words: WDRC, wide dynamic range compression, connected speechtest, HINT test, expansion, compression, noise reduction

Abbreviations: WDRC = wide dynamic range compression, CST = ConnectedSpeech Test, HINT = Hearing in Noise Test, NAL-NL1 = National AcousticLaboratories Non-Linear, SNR = signal-to-noise ratio, ANSI = American NationalStandards Institute

Sumario

Se investigaron los efectos de la expansión en las frecuencias graves sobrela evaluación objetiva y subjetiva de instrumentos auditivos intra-auricularesde cuatro canales. Se programaron cuatro ajustes de expansión en cadadispositivo: expansión apagada, expansión restringida al canal uno, y expansiónrestringida a los canales uno y dos. Se condujeron evaluaciones objetivas ensilencio (Prueba de Lenguaje Conectado) y en ruido (Prueba de Audición enRuido) con los niveles de lenguaje fijos a 40 dB SPL. Subjetivamente, cadaparticipante calificó la satisfacción en cuanto a la expansión en silencio yescuchando lenguaje a bajo volumen en una cabina sono-amortiguada, y luegoexpresó la condición de expansión preferida sobre las demás. Los sujetosfuncionaron significativamente mejor en silencio y en ruido para las condicionesde apagado y canal 1 que para la condición de canales 1 y 2; sin embargo,el desempeño fue similar entre las condiciones de apagado y canal 1. Losefectos de la expansión sobre la calificación de la satisfacción de los sujetos

JJoouurrnnaall ooff tthhee AAmmeerriiccaann AAccaaddeemmyy ooff AAuuddiioollooggyy/Volume 18, Number 8, 2007

642

Wide dynamic range compression(WDRC) hearing instrumentsprovide level-dependent gain in

order to increase the audibility of low-level, high frequency speech cues neces-sary for accurate speech recognition(Johnson, 1993; Killion, 1996). However,signal levels at or below the compressionkneepoint receive the maximum amplifica-tion available in WDRC instruments,thereby improving the audibility of low-level environmental noises and/or noisesgenerated by the hearing instrument aswell. As a result, WDRC hearing instru-ment users may report an audible "hiss-ing" sound when in low-level environ-ments (Venema, 2000). Expansion wasdesigned to provide reduced gain for sig-nals below the compression kneepointrather than maximum gain in order toreduce the audibility of these unwanted,low-level signals (Venema, 1998; Ghent etal, 2000; Kuk, 2002). Recent research hasdemonstrated that single-channel expan-sion improves subjective performance forWDRC hearing instrument users (Plyleret al, 2005a; Plyler et al, 2006) despitedegrading the recognition of low-levelspeech (Plyler et al, 2005a, Plyler et al,2005b).

Expansion effects differ if multi-channelexpansion is used instead of single-chan-nel expansion. Plyler et al (2007) exam-ined the effects of multi-channel expan-sion on the objective and subjective per-formance of twenty current WDRC hear-ing instrument users fitted binaurally

with four-channel digital in-the-ear prod-ucts (Starkey Axent II). Three memoriesof each instrument were programmed suchthat expansion was activated in all fourchannels, in channels one and two only(<2000 Hz), and deactivated in all fourchannels across memories. All other fit-ting parameters were held constant acrossthe three memories. Objective perform-ance was evaluated in quiet usingConnected Speech Test (CST) (Cox et al,1987; Cox et al, 1988) and in noise usingthe Hearing in Noise Test (HINT) (Nilssonet al, 1994) for speech levels of 40 dB, 50dB, and 60 dB SPL. Subjective perform-ance was evaluated by having each subjectrate their satisfaction regarding theamount of noise reduction they perceivedin each expansion condition on a dailybasis over a two-week period in real-worldlistening environments. In addition, eachsubject was asked to indicate whichexpansion condition they preferred follow-ing the completion of the two-week trial.Objective results in quiet and in noiseindicated that restricting expansion tochannels one and two significantlyimproved objective performance relative tothe four-channel expansion condition;however, objective results for the restrict-ed condition were significantly poorer thanthose for the no-expansion condition.Furthermore, subjective results indicatedthat restricting expansion to channels oneand two preserved the subjective perform-ance observed with four-channel expan-sion without significantly affecting listen-

dependieron del ambiente de escucha. Globalmente, los sujetos prefirieronla expansión en el canal 1 y en los canales 1 y 2; sin embargo, la preferenciano fue significativamente diferente entre la condición de canal 1 y apagado.Los resultados indican que la restricción en la expansión a 1000 Hz compensalas deficiencias en el reconocimiento del lenguaje observada con una expansiónactiva a lo largo de un espectro más amplio, sin reducir significativamente elbeneficio subjetivo o la preferencia.

Palabras Clave: WDRC, compresión de rango dinámico amplio, prueba delenguaje conectado, prueba de HINT, expansión, compresión, reducción delruido

Abreviaturas: WDRC = compresión de rango dinámico amplio, CST = Pruebade Lenguaje Conectado; HINT = Prueba de Audición en Ruido; NAL-NL1 =Laboratorios Nacionales de Acústica – No Lineal; SNR = tasa señal-ruido; ANSI= Instituto Americano Nacional de Estándares

LLooww--FFrreeqquueennccyy EExxppaannssiioonn/Lowery and Plyler

643

er preference. These findings suggestedrestricting expansion to below 2000 Hzpreserves subjective benefit observedwhen expansion is active across the entirespectrum but does not restore low-levelspeech recognition observed when expan-sion is deactivated.

The purpose of the present study wasto extend the work of Plyler et al (2007) todetermine the effects of low-frequency(<1000 Hz) expansion on the objectiveand subjective performance of hearing-instrument users. Improving access tospeech cues below 2000 Hz may improveaudibility of speech cues necessary foraccurate feature identification, therebyimproving the recognition of low-levelspeech. Furthermore, in-situ measuressuggested that subjective benefit withexpansion is attributed to expansion acti-vation below 1235 Hz (Plyler et al, 2007).Given this, it is reasonable to postulatethat restricting expansion to lower fre-quency regions than those examined byPlyler et al (2007) would improve therecognition of low-level speech withoutsignificantly reducing subjective benefitfor WDRC hearing-instrument users.The following research questions wereaddressed:

(1) Does low-frequency expansion affectthe recognition of low-level speech inquiet and in noise?(2) Does low-frequency expansion affectsatisfaction?(3) Does low-frequency expansion affectoverall listener preference?

MMEETTHHOODDSS

PPaarrttiicciippaannttss

Fifteen of the 20 adults that participat-ed in the Plyler et al (2007) experimentagreed to participate in the presentexperiment. All participants had sen-sorineural hearing impairment with nomore than a 15 dB HL difference in pure-tone thresholds at any octave frequencyfrom 250 through 8000 Hz between ears(ANSI S3.6-1996 [American NationalStandards Institute, 1996]), had normaltympanograms bilaterally, and were cur-rent binaural hearing instrument users(Figure 1). All qualification and experi-mental testing was conducted in a sound-treated examination room (IndustrialAcoustic) with ambient noise levels suit-able for testing with ears uncovered(ANSI S3.1-1991 [American NationalStandards Institute, 1991]).

HHeeaarriinngg IInnssttrruummeennttss aanndd FFiittttiinngg

Each hearing instrument used was thesame digital in-the-ear device with four-channel WDRC processing and multiplememory capability (Starkey Axent II)used in the Plyler et al (2007) experi-ment. Each hearing instrument was pre-viously programmed using the NAL-NL1fitting strategy (Byrne et al, 2001). Thecompression parameters were determinedby the Starkey software and varied acrossparticipants based on their audiometric

FFiigguurree 11.. Mean right and left air conduction thresholds and standard deviations of the participants.

data and resulting in-situ targets. Theexpansion parameters were also deter-mined by the Starkey software but wereconstant across all hearing instruments(Figure 2). Three memories of each hear-ing instrument were programmed in a ran-dom order such that expansion was acti-vated in channel one only (<750 Hz) and inchannels one and two only (<1750 Hz) anddeactivated in all four channels (Off)across memories; however, participantswere unaware of which expansion condi-tion they were using at all times. All otherfittings parameters were identical acrossthe three memories, and the noise suppres-sion feature and volume control were deac-tivated for the experiment.

Probe-microphone measures were con-ducted on each ear using pink noise at 50,65, and 80 dB SPL to verify match toNAL-NL1 targets (Byrne et al, 2001)(±6 dB from 500-4000 Hz) and with theloudspeaker deactivated to verify appro-priate functioning of the expansion fea-ture (Mueller, 2001). All probe-micro-phone measurements were downloaded toa personal computer and were averagedacross 26 ears for 13 of the 15 partici-pants (note: data from two participantswere excluded due to an equipment mal-function). Probe-microphone measuresobtained with the loudspeaker deactivat-ed revealed a mean in-situ level of 28 dBSPL for the unaided condition, 38 dB SPLfor the Channels 1 and 2 condition, 41 dB

SPL for the Channel 1 condition, and 44dB SPL for the Off condition averagedover the frequency range of 200 to 2000Hz (Figure 3).

OObbjjeeccttiivvee EEvvaalluuaattiioonn

Objective evaluations were conductedin quiet using the CST and in noise usingthe HINT (see Plyler et al, 2007, forreview). The CST is normally conductedusing a fixed signal-to-noise ratio (SNR);however, no background noise was uti-lized in order to evaluate the effects oflow-frequency expansion in a quiet set-ting. The HINT protocol utilized reflect-ed a slight modification of the originalHINT protocol in that noise levels werevaried and speech levels were fixed. Thismodification ensured consistent speechlevels between the CST and the HINTstimuli. All speech stimuli and back-ground noise were produced by a com-pact-disc player and routed through atwo-channel diagnostic audiometer (GSI-61) to a loudspeaker located at 0º azimuthin the sound-treated examination room(Industrial Acoustic). Participants wereseated 1m from the loudspeaker, and theCST and the HINT were administered at40 dB SPL (peak) for the three expansionconditions (Off, Channel 1, Channels 1and 2). Two CST and two HINT trialswere conducted for each expansion condi-tion, and an average of the two trials


644

FFiigguurree 22.. Expansion parameters determined by the Starkey software and crossover frequenciesbetween the first and second, second and third, and third and fourth channels of each hearing instru-ment. Mean in-situ levels averaged across ears for 13 of the 15 participants for the expansion off con-dition are shown for reference purposes.

served as the CST or HINT score for thatparticipant in the given condition. Priorto data collection, an experimental sched-ule was generated listing a completelyrandomized assignment for expansioncondition, CST passage, and HINT sen-tence list for each participant.

SSuubbjjeeccttiivvee EEvvaalluuaattiioonn

In previous expansion studies, subjec-tive evaluations were conducted in theparticipants' own homes instead of in thelaboratory to examine expansion in thereal-world settings for which it wasdesigned (Plyler et al, 2005a; Plyler et al,2006; Plyler et al, 2007). In the presentstudy, subjective evaluations were con-ducted in laboratory to determine if labo-ratory findings agree with those obtainedin more real-world listening environ-ments. All participants were asked torate their satisfaction for each memoryunder two listening conditions while seat-ed in a sound-treated examination room:(1) in quiet, and (2) while listening toCST sentences (Cox et al, 1987) present-ed at 40 dB SPL (peak). The presentationlevel of 40 dB SPL ensured the activationof expansion for those memories in whichexpansion could be activated. For eachexpansion condition, participants wereasked to rate their satisfaction regardingthe amount of background-noise reduc-tion they perceived in quiet (i.e., listening

to the ambient noise in the examinationroom) and when listening to low-levelspeech using a five-point scale (1=unsat-isfied, 2=somewhat satisfied, 3=satisfied,4=very satisfied, 5=extremely satisfied)(Appendix). Two evaluation sessionswere conducted for each listening condi-tion for each participant. Prior to datacollection, an experimental schedule wasgenerated for each participant listing acompletely randomized assignment forlistening condition, expansion condition,and CST passage. At the completion ofthe experiment, each participant wasasked to indicate which setting they pre-ferred overall to determine if low-fre-quency expansion affected overall prefer-ence.

RREESSUULLTTSS


Prior to statistical analysis, individualpercent-correct scores were converted torationalized arcsine transform units(raus) to stabilize error variance(Studebaker, 1985). Connected SpeechTest scores and HINT SNRs were aver-aged within participants for each expan-sion condition (Figures 4 and 5 respec-tively). Two one-way repeated measuresanalyses of variance were performed toevaluate the effects of low-frequencyexpansion on speech recognition in quiet


645

FFiigguurree 33.. Mean in-situ levels averaged across ears for 13 of the 15 participants for each expansion con-dition and the unaided condition with the loudspeaker deactivated.

and in noise. The dependent variable wasCST score for one analysis and HINTSNR for the other analysis. For eachanalysis, the within-subject factor wasexpansion with three levels (Off, Channel1, Channels 1 and 2). A significantexpansion effect was revealed for the CST[F(2,28) = 5.246, p<0.05, partial η2 = 0.273,Ω = 0.790] and for the HINT [F(1,14) =8.644, p<0.05, partial η2 = 0.382, Ω = 0.780]results. Paired samples t-tests were con-ducted to further investigate each expan-sion main effect controlling for family-wise error rate across the tests at the0.05 level, using the Holm’s sequentialBonferroni procedure. Results indicatedthat CST scores and HINT SNRs weresignificantly poorer for the Channels 1and 2 condition than for the Channel 1 orOff conditions; however, CST scores andHINT SNRs were not significantly differ-ent between the Channel 1 and Off condi-tions (Table 1).


SSaattiissffaaccttiioonn RRaattiinnggss

Individual satisfaction ratings werecompleted by all participants and wereaveraged across two trials for each expan-sion condition and each listening environ-ment (Figure 6). Two one-way repeatedmeasures analyses of variance were per-formed to evaluate the effects of low-fre-quency expansion on listener satisfactionratings in quiet and when listening tolow-level speech. The dependent variablewas mean satisfaction rating, and thewithin-subject factor was expansion withthree levels (Off, Channel 1, Channels 1and 2) for each analysis. A significantexpansion effect was revealed for the quiet[F(2,28) = 5.589, p<0.05, partial η2 = 0.285,Ω = 0.816] and for the low-level speech[F(2,28) = 7.408, p<0.05, partial η2 = 0.346,Ω = 0.914] ratings results. Paired sam-


646

Table 1. Comparisons from the CST and HINT Data Controlling for Familywise Error Rate acrossthe Tests at the 0.05 Level, Using the Holm’s Sequential Bonferroni Procedure within Condition

Test Pair Mean Standard t (14) Adjusted p Difference Deviation Value

CST Off, Channel 1 3.73 16.17 0.89 0.386

Off, Channels 1 & 2 14.33 21.59 2.57 0.004

Channel 1, Channels 1 & 2 10.60 14.87 2.76 0.045

HINT Off, Channel 1 1.00 4.09 0.95 0.359

Off, Channels 1 & 2 -4.80 6.32 -2.94 0.022

Channel 1, Channels 1 & 2 -5.80 6.60 -3.40 0.001

FFiigguurree 44.. Mean CST in quiet scores and standard deviations for participants at 40 dB SPL for eachexpansion condition.

ples t-tests were conducted to furtherinvestigate each expansion main effectcontrolling for familywise error rateacross the tests at the 0.05 level, usingthe Holm’s sequential Bonferroni proce-dure. Results in quiet indicated that sat-isfaction ratings were significantly lowerfor the Off condition than for theChannels 1 and 2 condition; however, sat-isfaction ratings for the Channel 1 condi-tion were not significantly different fromratings for the Off or the Channels 1 and2 conditions (Table 2). Conversely, low-level speech results indicated that satis-faction ratings were significantly higherfor the Off and Channel 1 conditions thanfor the Channels 1 and 2 condition; how-ever, satisfaction ratings for the Off con-dition were not significantly differentfrom ratings for the Channel 1 condition(Table 2). These results indicated thatthe effects of expansion on listener satis-faction ratings depend on the listeningenvironment.

OOvveerraallll PPrreeffeerreennccee

Overall preference results wereobtained from the 15 participants and aredisplayed in Figure 7. A one-sample chi-square test was conducted to determinewhether listeners had a preferred expan-sion condition (Off, Channel 1, Channels1 and 2) (note: the exact option was usedto determine significance due to the smallsample size). The result of the test wassignificant, χ2(1, N = 15) = 8.40, p<0.05.The proportion of listeners that preferredthe Channel 1 condition (P = 0.66) wasgreater than the hypothesized proportionof 0.33 while the proportion of listenersthat preferred the Off (P = 0.26) andChannels 1 and 2 conditions (P = 0.06)was less than the hypothesized propor-tion of 0.33. Follow-up testing indicatedthat the proportion of listeners preferringthe Channel 1 condition differed signifi-cantly from the proportion of listeners pre-ferring the Channels 1 and 2 condition,


647

Table 2. Comparisons from the Satisfaction Ratings Data Controlling for Familywise Error Rate acrossthe Tests at the 0.05 Level, Using the Holm’s Sequential Bonferroni Procedure within Condition

Environment Pair Mean Standard t (14) Adjusted p Difference Deviation Value

Quiet Off, Channel 1 -0.93 1.58 -2.29 0.076

Off, Channels 1 & 2 -1.23 1.72 -2.78 0.045

Channel 1, Channels 1 & 2 -0.30 1.10 -1.06 0.308

Low-Level Off, Channel 1 0.20 0.10 0.78 0.450

Speech Off, Channels 1 & 2 1.20 1.37 3.38 0.012

Channel 1, Channels 1 & 2 1.00 1.46 2.65 0.038

FFiigguurree 55.. Mean HINT SNRs and standard deviations for participants at 40 dB SPLfor each expansion condition.

χ2(1, N = 11) = 0.007, p<0.05; however, theproportion of listeners preferring theChannel 1 condition did not differ signifi-cantly from the proportion of listeners pre-ferring the Off condition, χ2(1, N = 14) =0.109, p>0.05. These results suggestedthat listeners preferred expansion inChannel 1 only to expansion in Channels 1and 2; however, overall preference was notsignificantly different between theChannel 1 and Off conditions.

DDIISSCCUUSSSSIIOONN


One purpose of the present study wasto determine if low-frequency expansionaffected the recognition of low-levelspeech in quiet and in noise. Resultswere in agreement with previous findingsobtained using multi-channel expansion


648

FFiigguurree 66.. Mean satisfaction ratings and standard deviations for participants when listening in quietand when listening to low-level speech (CST sentences) in the sound treated booth for each expansioncondition using a five-point scale (1=unsatisfied, 2=somewhat unsatisfied, 3=satisfied, 4=very satisfied,5=extremely satisfied).

FFiigguurree 77.. Overall preference results for participants when listening in quiet and when listening to low-level speech (CST sentences) for each expansion condition.

in that multi-channel expansion restrict-ed to 2000 Hz significantly degradedobjective performance in quiet and innoise relative to the Off condition, pre-sumably due to reduced audibility ofspeech cues necessary for accurate fea-ture identification during expansion acti-vation (Plyler et al, 2005a; Plyler et al,2007). Results further indicate, however,that restricting expansion to lower fre-quency regions than those examined byPlyler et al (2007) did not significantlydegrade the recognition of low-levelspeech in quiet and in noise.

Audibility of low-level speech cues mayexplain the speech-recognition patternsobserved in the present examination oflow-frequency expansion. Probe-micro-phone measures revealed that in-situ lev-els decreased by an average of 7.3 dBfrom 1000 to 2000 Hz for the Channels 1and 2 condition relative to the Off condi-tion (Figure 3). Conversely, in-situ levelsdecreased by an average of 1.4 dB from1000 to 2000 Hz for the Channel 1 condi-tion relative to the Off condition (Figure3). As a result, the audibility of speechcues necessary for accurate feature iden-tification may have been significantlyreduced for the Channels 1 and 2 condi-tion relative to the Off condition; howev-er, audibility of necessary speech cuesmay have been similar between theChannel 1 and Off conditions. Therefore,findings of the current investigation sug-gest that limiting expansion activation to1000 Hz and below does not result in sig-nificant speech-recognition deficitsobserved when expansion is active across abroader spectrum, possibly due toimproved audibility of necessary speechcues between 1000 and 2000 Hz (Figure 3).


QQuuiieett

A second purpose of the present studywas to determine if low-frequency expan-sion affected satisfaction with low-levelnoise reduction. Satisfaction ratingswith low-frequency expansion were eval-uated in quiet and in the presence of low-level speech. Results indicated that sat-isfaction with low-frequency expansionwas dependent on the listening environ-

ment. In quiet, listener satisfaction rat-ings were significantly greater for theChannels 1 and 2 condition than the Offcondition. This result was in agreementwith previous findings obtained usingmulti-channel expansion in real-worldsettings in that multi-channel expansionactivated in Channels 1 and 2 significant-ly improved satisfaction ratings in quietrelative to the Off condition, presumablydue to reduced audibility of low-levelenvironmental noises as well as noisesgenerated by the hearing instrument(Plyler et al, 2007).

Results in quiet further indicated thatlistener satisfaction ratings were similarbetween the Channel 1 condition andboth the Channels 1 and 2 and Off condi-tions. This outcome seemed paradoxicalat first glance. On one hand, limitingexpansion to Channel 1 maintains subjec-tive benefit observed when expansion isactive in Channels 1 and 2 or across theentire spectrum (Plyler et al, 2007). Onthe other hand, limiting expansion toChannel 1 does not provide significantlygreater subjective benefit than the Offcondition. This result was not expectedand is difficult to interpret. Althoughsatisfaction ratings appeared greater forthe Channel 1 condition than for the Offcondition, ratings differences approachedbut did not reach statistical significance(p = 0.07). It should be noted, however,that statistical significance is reachedbetween the conditions if adjustments formultiple comparisons are not made,thereby suggesting that satisfaction rat-ings may be significantly greater whenexpansion is limited to 1000 Hz andbelow (Channel 1) than when expansionis deactivated when listening in a quietenvironment. Nonetheless, listener satis-faction ratings between these two condi-tions (Channel 1 and Off) did not differsignificantly in quiet and warrants fur-ther study in the future.

LLooww--LLeevveell SSppeeeecchh

Results obtained when listening to low-level speech indicated that listener satis-faction ratings were significantly reducedwhen expansion was activated inChannels 1 and 2; however, satisfactionratings were significantly improved when


649

expansion was limited to Channel 1 ordeactivated. These results were not inagreement with previous findingsobtained when listening in low-level envi-ronments using either single-channel ormulti-channel expansion which indicatedthat any form of expansion significantlyimproved listener satisfaction ratings ineveryday environments despite the factexpansion significantly reduced speech-recognition in quiet and in noise (Plyleret al, 2005a; Plyler et al, 2007).

Differences in listener satisfaction rat-ings in low-level environments observedbetween the present and previousresearch could be attributed to themethodological differences used betweeninvestigations. In previous research, par-ticipants evaluated each expansion condi-tion in their own homes instead of in thelaboratory in order to examine the effectsof expansion in the real-world settings forwhich expansion was designed (Plyler etal, 2005a; Plyler et al, 2007; Plyler et al,2006). Participants were instructed toconduct their everyday environment eval-uations while paying close attention tosoft sounds such as having a quiet con-versation with one other person, listeningto leaves rustling, watching television ata low volume, etc. (Plyler et al, 2005a;Plyler et al, 2007; Plyler et al, 2006).Consequently, it is possible some partici-pants did not listen to speech when eval-uating expansion in their everyday envi-ronment while others rarely listened tospeech at such low levels. In either case,speech-recognition deficits from expan-sion activation would have gone unno-ticed and would not impact the subjectiveresults. Conversely, it is also possiblethat the noise-reduction benefits pro-duced by expansion outweighed thespeech-recognition deficits producedwhen listening in a low-level environ-ment given the fact most listeners pre-ferred the use of expansion (Plyler et al,2005a; Plyler et al, 2007; Plyler et al,2006).

In the present study, subjective evalua-tions were conducted in laboratory todetermine if laboratory findings agreedwith those obtained in more real-worldenvironments. Each participant wasasked to rate his or her satisfaction witheach expansion condition when listening

in quiet and when listening to low-levelspeech (40 dB SPL) while seated in asound-treated examination room. Thisdesign ensured that (1) ambient levelsfell below the expansion thresholds andengaged the feature during quiet evalua-tions (Figure 3) and (2) participants actu-ally listened to low-level speech whenevaluating each expansion condition.Interestingly, satisfaction ratings resultsobtained when listening to low-levelspeech were in agreement with objectiveresults in that performance for bothmeasures was significantly reduced whenexpansion was activated in Channels 1and 2 while performance improved signif-icantly when expansion was limited toChannel 1 or was deactivated. Therefore,these results suggest that satisfactionwith expansion depends on the listeningenvironment of the end user. Results fur-ther indicate, however, that limitingexpansion activation to 1000 Hz andbelow (Channel 1) does not significantlyreduce expansion satisfaction when lis-tening in quiet or when listening to low-level speech.

OOvveerraallll PPrreeffeerreennccee

The final purpose of the present studywas to determine if low-frequency expan-sion affected overall listener preference.A significant number of participants pre-ferred limiting expansion activation toChannel 1 (66.6%) to activating expan-sion in Channels 1 and 2 (6.6%); however,the number of participants who preferredno expansion (26.6%) was not significant-ly different from the number of partici-pants who preferred limiting expansionactivation to Channel 1 (66.6%). Thisresult was not in complete agreementwith previous research (Plyler et al,2005a; Plyler et al, 2007), which indicat-ed that listeners preferred any form ofexpansion to no expansion. However,overall preference differences observedbetween the present and previousresearch could be attributed to themethodological differences betweeninvestigations. In the present research,satisfaction ratings were conducted in thelaboratory and were affected by the lis-tening environment, thereby potentiallyaffecting the overall preference results.


650

Furthermore, preference results werealso in agreement with low-level speechsatisfaction ratings and objective results.For example, 93% of the participants pre-ferred deactivating expansion or limitingexpansion activation to Channel 1, condi-tions which maximized speech recogni-tion in quiet, speech recognition in noise,and satisfaction ratings when listening tolow-level speech. Therefore, results ofthe present study suggest limiting expan-sion activation to 1000 Hz and below(Channel 1) preserves both objective andsubjective performance in quiet and inlow-level environments, and is preferredoverall by a significant number of partic-ipants.

CCOONNCCLLUUSSIIOONNSS

Results of the present study indicatethat hearing-instrument users prefer

the use of low-frequency expansion toexpansion that affects a larger frequencyspectrum. Results further suggest thatrestricting expansion to below 1000 Hzovercomes speech-recognition deficitsobserved when expansion is active acrossa broader spectrum (Plyler et al, 2005a;Plyler et al, 2007) without significantlyreducing subjective benefit for WDRChearing-instrument users. Dispensersshould be aware that subjective perform-ance with expansion depends on whetherthe end user is listening in a quiet envi-ronment or is listening to low-levelspeech. Lastly, any interpretation ofresults should consider the specificexpansion parameters used in this study.Because the effect of expansion on low-level speech and noise depends on theexpansion parameters, these results maynot generalize to other types of expansionsystems (e.g., more than four channels orother expansion parameters such asthreshold, ratio and time constant).Future work should investigate the rolethese parameters play on the effective-ness of expansion.

AAcckknnoowwlleeddggmmeennttss.. The authors would like toacknowledge the support of Starkey Laboratories forproviding the hearing instruments used in this research.

RREEFFEERREENNCCEESS

American National Standards Institute. (1991) AmericanNational Standard Criteria for Permissible Ambient NoiseDuring Audiometric Testing (ANSI S3.1-1991). New York:ANSI.

American National Standards Institute. (1996) AmericanNational Standard Specification for Audiometers (ANSIS3.6-1996). New York: ANSI.

Byrne D, Dillon H, Ching T, Katsch R, Keidser G. (2001)NAL-NL1 procedure for fitting nonlinear hearing aids:characteristics and comparisons with other procedures.J Am Acad Audiol 12:37–51.

Cox R, Alexander G, Gilmore C. (1987) Development ofthe Connected Speech Test (CST). Ear Hear 8:119–126.

Cox R, Alexander G, Gilmore C, Pusakulich K. (1988) Useof the Connected Speech Test (CST) with hearing-impairedlisteners. Ear Hear 9:198–207.

Ghent RM, Nilsson MJ, Bray VH. (2000) Uses of expan-sion to promote listening comfort with hearing aids. Posterpresented at the American Academy of Audiology’s 12thAnnual Convention, Chicago.

Johnson WA. (1993) Beyond AGC-O and AGC-I: thoughtson a new default standard amplifier. Hear J 46:63–69.

Killion MC. (1996) Compression distinctions. Hear Rev3:29–32.

Kuk FK. (2002) Considerations in modern multichannelnonlinear hearing aids. In: Valente M, ed. Hearing Aids:Standards, Options, and Limitations. New York: ThiemeMedical Publishers, 186–187.

Mueller HG. (2001) Probe-mic assessment of digital hear-ing aids? Yes, you can! Hear J 54:14–15.

Nilsson M, Soli S, Sullivan J. (1994) Development of theHINT for the measurement of speech reception thresh-old in quiet and in noise. J Acoust Soc Am 95:1085–1099.

Plyler PN, Hill AB, Trine TD. (2005a) The effects of expan-sion on the objective and subjective performance of hearinginstrument users. J Am Acad Audiol 16:101–113.

Plyler PN, Hill AB, Trine TD. (2005b) The effects of expan-sion time constants on the objective performance of hearinginstrument users. J Am Acad Audiol 16:614–621.

Plyler PN, Lowery KJ, Hamby HM, Trine TD. (2007) Theobjective and subjective evaluation of multichannel expan-sion in wide dynamic range compression hearinginstruments. J Speech Hear Res 50(1):15–24.

Plyler PN, Trine TD, Hill AB. (2006) The subjective eval-uation of the expansion time constant in single-channelwide dynamic range compression hearing instruments.Int J Audiol 45:331–336.

Studebaker GA. (1985) A “rationalized” arcsine trans-form. J Speech Hear Res 28:455–462.

Venema TH. (1998) Programmable and digital hearingaids. In: Venema TH. Compression for Clinicians. SanDiego: Singular Publishing Group, 87–93.

Venema TH. (2000) The many faces of compression. In:Sandlin RE, ed. Textbook of Hearing Aid Amplification.San Diego: Singular Publishing Group, 238–239.


651

AAppppeennddiixx

Initials: ____________________ Date: ________________________

Comments: ___________________________________________________________________________________________

______________________________________________________________________________________________________

______________________________________________________________________________________________________

______________________________________________________________________________________________________

______________________________________________________________________________________________________

______________________________________________________________________________________________________


652

Quiet

EvaluationSession 1:

EvaluationSession 2:

Rate your satisfaction withthe amount of backgroundnoise reduction (1=unsatisfied, 2=somewhat satisfied,3=satisfied, 4=very satisfied,5=extremely satisfied)

Memory 1:_____

Memory 2:_____

Memory 3:_____

Favorite:______

Memory 1:_____

Memory 2:_____

Memory 3:_____

Favorite:______

Speech Rate your satisfaction withthe amount of backgroundnoise reduction (1=unsatisfied, 2=somewhat satisfied,3=satisfied, 4=very satisfied,5=extremely satisfied)

Memory 1:_____

Memory 2:_____

Memory 3:_____

Favorite:______

Memory 1:_____

Memory 2:_____

Memory 3:_____

Favorite:______

Documents

The Objective and Subjective Evaluation of Low … · la expansión en el canal 1 y en los canales 1 y 2; sin embargo, la preferencia no fue significativamente diferente entre la