KISEP RESEARCH PAPER AUDIOLOGY 청능재활 6;2:33-39 양이청취조건하에서잡음특성이 곰 / 공 음소변별에미치는영향 ABSTRACT 한림대학교대학원청각학, 1 한림대학교언어청각학부청각학전공, 청각언어연구소 2 정두환 1 구성민 1 임덕환 2 Effects of Noise Parameters on Discriminating Gom and Gong Under Binaural Condition Doohwan Chung, 1 Sungmin Koo 1 and Dukhwan Lim 2 1 Graduate Program in Audiology, 2 Section of Audiology, Audiology and Speech Pathology Research Institute, Hallym University, Chuncheon, Korea The effects of noise spectral bands on phoneme discrimination between gom and gong were studied under binaural hearing conditions for twenty young adults. The competing phonemes were provided under the broadband noise of.2-1 khz, lowpass filtered noise of 1 khz cutoff frequency, bandpass filtered noise of 1-3 khz band, and highpass filtered noise of 3 khz cutoff frequency. The phase of each filtered noise was shifted under binaural condition by 1 degree. Presentations with background noises were grouped in terms of signal to noise levels of -1 db, db, and +1 db SNR. Stimuli were digitally synthesized and two alternative forced choice (2AFC) method was used to collect the corresponding discrimination scores. The results were as follow:(a) In low pass noise, critical frequency region for discriminating final phonemes of gom and gong was approximately below 1 khz. (B) High pass noise band had little effects on discrimination task and this indicated the noise range could be outside the critical band of the phonemes. (C) The effect of BMLD also could not be observed in data under high pass filtered noise since the frequency band was outside the critical band of the phonemes. (D) Some SNRs had significant effects on discriminating competing phonemes. The outcome indicated the needs for similar studies on competing phonemes for various placements. The data can be used to elaborate new methods of auditory evaluation, optimal fitting of hearing aids, and effective mapping of cochlear implants under the binaural condition. KEY WORDS:Two alternative forced choice (2AFC) BMLD Critical band Binaural hearing. INTRODUCTION 난청인뿐만아니라정상청력을가진사람에게도어음에대한변별에방해가되는요인중하나는주변환경에서발생하는잡음이다. 그런잡음속에서어음변별을향상시키기위한많은연구가진행되고있다. 특히양이청취를이용한잡음상황에서의어음청취력과방향성개선효과등이입증되어지고있다. 1) 이런향상의요인중하나는 BMLD(binaural masking level difference) 에의한것일수있다. BMLD는신호에 논문접수일 :6 년 3 월 5 일심사완료일 :6 년 5 월 11 일교신저자 : 임덕환, -72 강원도춘천시옥천동 1 번지한림대학교언어청각학부청각학전공, 청각언어연구소전화 :(33) 248-2217 전송 :(2) 62-9133 E-mail:dlim@hallym.ac.kr 대한차폐된탐지역치가 signal 과 masker 의상대적인차이에의해서향상되는것을의미한다. BMLD 는비교적미세한크기차이 (interaural difference cues of intensity, ILD) 와시간차이 (interaural difference cues of time, ITD) 를처리하기위한청각시스템의기능에기인하는것으로알려져있다. 7) Goldstein 은청각시스템은 phase deaf 는아니며제한된범위내에서인지적으로중요하게작용한다고제안했다. 15) 이 phase 는소리의변별에중요한요인으로작용할수있다. 최근에는복합음을비롯한어음에대한연구도진행되고있다. 15) 차폐잡음의 narrow band 에서 5 Hz 의순음위상을 1 변위했을때 (N S π ) 역치는위상변위를하지않았을때 (N S ) 보다 BMLD 역치가약 15 db 낮게나타났다. 17) 어음자극에대해서는약 1.5~8 db 더낮은 BMLD 역치가관찰되기도했다. 1)2)6)12) 33
34 AUDIOLOGY 청능재활 6;2:33-39 BMLD 의대표적인조건은위상을변위하지않은경우 (N S ) 와신호음의위상을 1 변위한경우 (N S π ) 이다. N S 은잡음과신호모두같은위상으로나타나게하는것이고, N S π 는잡음의위상은변위하지않고신호음의위상을 1 변위한것이다. 어음의탐지와인지는 BMLD 신호음의위상의변위여부에영향을받는다는결과가보고되었으나, 1-3)12)21) 잡음을위상변위한연구는상대적으로적은편이다. 추가하여, 양이청취조건하에서배경잡음의위상변위가특정음소의변별에어떠한영향을주는지, 잡음의 spectral band 가특정음소의변별에어떠한영향을주는지에대한참고자료가청능훈련과정에서필요하다. 본연구의목적은양이청취조건하에서특정음소 ( 곰 / 공 ) 의변별에위상변위된잡음 (N π S ) 은어떤영향을주는지, noise spectral band 의어느주파수대역이음소별별에영향을주는지살펴보고, 다양한조건의신호대잡음비는그결과에어떠한영향을줄것인지살펴보고자하였다. MATERIALS AND METHODS Subjects 정상청력의성인남녀각 1명씩을대상으로하였다. 연령분포는 ~28세로평균 22.95세 ( 남 :24.1세, 여 :21.8 세 ) 이며, 청력검사결과는모든주파수에서 dbhl 이하로정상청력을보였다. 이경검사및고막운동성검사를실시하였으며, 대상자모두양이의 WRS(word recognition score) 가 % 이상으로정상수치를기록했다. Stimuli 본연구에서사용된자극음은청능평가에서사용되고있는 GASP!-K(Glendonal Auditory Screening Procedure! -Korean version) 에서변별이어려운 곰 / 공 을사용했다. 음소와동시에합성되어제시되는배경잡음은.2~ 1 khz 의전주파수대역의 broad band noise, 1 khz 이하로통과시킨 low pass noise, 1~3 khz를취하는 band pass noise, 3 khz 이상을통과시킨 high pass noise로구성되었다. 신호음과잡음은양측으로동시에제시되며동일한어음신호위상에잡음주파수대역만반대측이 1 위상변위되어동시에합성되었다. 실험전초기조건으로각배경잡음대역을위상변위되지않은잡음이사용되었다. 모든배경잡음은세가지의신호대잡음비 (-1 db, db, +1 db SNR) 에의해제시되었다. 자극음은 digital-to-analog converter(gw) 를이용하여표본추출비율 (sampling rate) 5 khz로전 산합성되었고, low-pass filter(frequency Devices 9) 로 1 khz 에서 filtering 을실시한후강도조절기 (HEWL- ETT PACKARD 35D) 로 7 dbhl 수준에서일정하게유지되었다. 잡음이적절히차단된공간에서대상자의양이에삽입형이어폰 (ER 4, micropro, Etymotic Research) 을통해자극음을제시하였다 (Fig. 1). Procedure Response bias 를배제하기위해전산프로그램을이용한 2AFC(Two Alternative Forced Choice) method가사용되었다. 2AFC 에포함된두개의 stimulus 는 rise/fall time msec 포함하여 1, msec 씩 5 msec 간격으로구성되었으며 (Fig. 2), 각 stimulus 에는실험을위해배경잡음과전산합성된 signal- 곰 과 non-signal- 공 이무작위로피검자에게제시되었다. Voice + Background noise (no phase shift, phase shift) 2AFC Stimulus A Stimulus B Low-pass filter 1 (Frequency Devices 9) Attenuator 1 (HEWLETT PACKARD 35D) Synthesizing D/A converter (GW) Computer Low-pass filter 2 (Frequency Devices 9) Attenuator 22 (HEWLETT APCKARD 35D) Fig. 1. System block diagram for recording and analyzing the data. Selection of Stimulus A and Stimulus B in the successive presentation was randomly chosen for signal gom and nonsignal gong. Stimulus A Stimulus B 9 5 9 Fig. 2. Illustration of the profile of a single trial in 2AFC task. One stimulus contained the signal gom and the other stimulus included the non-signal gong.
DH Chung, et al:effects of Noise on Discriminating Gom and Gong G 35 무작위로제시되는 5회의실험에서마우스클릭으로선택된결과가전산처리되어변별점수가기록되었다 (Fig. 3). 각측정실험은초기조건의주파수대역별로 Broad band noise, Low pass noise(<1 khz), Band pass noise(1~3 khz), High pass noise(>3 khz) 에신호대잡음비 -1 db, db, +1 db의 session에위상변위한조건의 session 을더해총 24개 session 을각 5회씩실시하여최종변별점수로기록하였다 (Table 1). (correct percentage) (Hit+Correct Rejection) = (Hit+Miss+False Alarm+Correct Rejection) Table 1. Calculation for discrimination scores Response Signal gom Stimulus A Decision [Correct / Wrong] Stimulus Non-signal gong Stimulus B Signal gom Signal gom Hit Miss Randomization 5 trials Fig. 3. Two Alternative Forced Choice (2AFC) procedure. Each stimulus contained either signal gom or non-signal gong and this selection was randomly assigned. Presentation was in the order of Stimulus A and Stimulus B and, then, a subject was asked to choose which stimulus had the signal gom. Non-signal gong Non-signal gong False alarm Correct rejection RESULTS 대상자 명에대한 곰 / 공 의변별실험에서 N S 에대한 N π S 의자료와함께각신호대잡음비에대한결과가수집되었다. 이결과를요약하면, one-way ANOVA 분석에서신호대잡음비에서유의미한통계적인차이를보였고, 일부 N S 의결과와 N π S 의결과사이에서유의미한차이가나타났다. 와 Low pass noise(<1 khz) 에서 N π S 에대해통계적차이가있었고, 이와유사한결과가 Broad band noise 와 Low pass noise(<1 khz) 에서 N S 에대해서도나타났다 (Fig. 4, Table 2 and 3, ANOVA:p<.5). 신호대잡음비와변화와관련된자료에서모든주파수조건사이에서통계적차이를보이는결과가나타났다. Low pass filtered noise(<1 khz) 는 N π S 에서신호대잡음비 db와 +1 db간의유의미한차이를보였다. Low pass filtered noise(<1 khz) 를제외하고신호대잡음비 -1 db, db, +1 db에서다른주파수조건사이에서유의미한차이가없었다 (Fig. 5, Table 4 and 5, ANOVA:p<.5). N S 의결과와 N π S 의결과를 paired t-test 한결과에서, 신호대잡음비가 db 조건에서 Low pass noise(<1 khz) 와 High pass noise(>3 khz) 에서통계적유의미한차이를보였다. 이결과는 critical band 영역이 N S 와 N π S 에미치는영향때문인것으로추정되었다 (Fig. 6, Table 6, paired t-test results:p<.5). DISCUSSIONS AND CONCLUSIONS 본연구에서는다양한주파수대역과세가지다른신호대잡음비에대한위상의변위가 곰 / 공 의변별에미치는영향에대해실험하였다. 변별과관련하여 Low pass filtered noise(<1 khz) 를제외한모든조건에서통계적으로유의미한차이가나타나지않았다. 이는 곰 / 공 의음 (.2-1 khz) Low pass filtered noise (1 khz) Band pass filtered noise (1-3 khz) High pass filtered noise (>3 khz) NπS NS NπS NS NπS NS NπS NS Fig. 4. The comparison of discrimination scores under the various signal to noise levels and phase shifts. Signal to noise levels were noted as SNR of -1 db ( ), db ( ), and 1 db ( ). The subscript π in the figure stands for the binaural condition of 1 degree phase shift and for binaural in-phase condition. S and N in the figure denote signal and background noise, respectively. : Statistically significant difference (p<.5).
36 AUDIOLOGY 청능재활 6;2:33-39 소변별의 critical band 가결과에중요하게작용한것으로결론되었다. 사용된음소 곰 / 공 은변별하는 critical frequency band가 1 khz 이하이기때문에대부분의 Low pass filtered noise(<1 khz) 에서의결과가낮게나타난것으로 추정된다. 이와연관된결과로 Moore 는변별을결정하는 dominance region 이존재한다고보고하였고, 17) 이 dominance region 에서위상과시간분석모두가능하게나타난다고하였다. 4) 다른선행연구에따르면, filtered noise Table 2. ANOVA analysis for groups of noises with binaural phase shift under various SNR conditions (p<.5) Binaural phase shift (Dichotic noise) Source Sum of squares df F p-value Between groups*.212 2 1.731.* Within groups.564 57 Total.776 59 Between groups*.787 2.182.* Low pass filtered noise (<1 khz) Within groups 1.111 57 Total 1.889 59 Between groups*.3 2 1.847.167 Band pass filtered noise (1-3 khz) Within groups.41 57 Total.44 59 Between groups*.1 2.477.623 High pass filtered noise (>3 khz) Within groups.38 57 Total.39 59 *:In terms of various SNRs, :In terms of the individual subjects Table 3. ANOVA analysis for groups of noises with no binaural phase shift under various SNR conditions (p<.5) No binaural phase shift (Diotic noise) Source Sum of squares df F p-value Between groups*.127 2 7.931.1* Within groups.457 57 Total.585 59 Between groups*.442 2.489.* Low pass filtered noise (<1 khz) Within groups.614 57 Total 1.56 59 Between groups*.2 2 1.3.2 Band pass filtered noise (1-3 khz) Within groups.42 57 Total.44 59 Between groups*.3 2 2.314.18 High pass filtered noise (>3 khz) Within groups.32 57 Total.35 59 *:In terms of various signal to noise ratios, :In terms of the individual subjects SNR -1 db SNR db SNR +1 db NπS NS NπS NS NπS NS Fig. 5. The comparison of discrimination scores under the various frequency bands of the background noise and binaural phase shifts. The background noise were noted as broadband noise ( ), low pass filtered noise ( ), band pass filtered noise ( ), and high pass filtered noise ( ). The subscript π in the figure stands for the binaural condition of 1 degree phase shift and for binaural in-phase condition. S and N in the figure denote signal and background noise, respectively. :Statistical difference.
DH Chung, et al:effects of Noise on Discriminating Gom and Gong G 37 Table 4. ANOVA analysis for groups of noises with binaural phase shift under various frequency bands (p<.5) Binaural phase shift (Dichotic noise) Source Sum of squares df F p-value Between groups* 1.265 3 24.317.* SNR -1 db Within groups 1.318 76 Total 2.583 79 Between groups*.51 3 35.3.* SNR db Within groups.356 76 Total.857 79 Between groups*.14 3 4.384.7* SNR +1 db Within groups. 76 Total.94 79 *:In terms of various frequency bands, :In terms of the individual subjects Table 5. ANOVA analysis for groups of noises with no binaural phase shift under various frequency bands (p<.5) No binaural phase shift (Diotic noise) Source Sum of squares df F p-value Between groups*.673 3.9.* SNR -1 db Within groups.849 76 Total 1.523 79 Between groups*.57 3 7.6.* SNR db Within groups.188 76 Total.245 79 Between groups*. 3 6.584.1* SNR +1 db Within groups.76 76 Total.96 79 *:In terms of various frequency bands, :In terms of the individual subjects SNR -1 db Low pass filtered noise (<1 khz) Band pass filtered noise (1-3 khz) High pass filtered noise (>3 khz) SNR db Low pass filtered noise (<1 khz) Band pass filtered noise (1-3 khz) High pass filtered noise (>3 khz) Fig. 6. The results of paired t-test analysis between N πs ( ) and NS ( ). *:stands for statistically significant difference. The subscript π in the figure stands for the binaural condition of 1 degree phase shift and for binaural in-phase condition. S and N in the figure denote signal and background noise, respectively. SNR +1 db Low pass filtered noise (<1 khz) Band pass filtered noise (1-3 khz) High pass filtered noise (>3 khz) Discrimination socre
38 AUDIOLOGY 청능재활 6;2:33-39 Table 6. The paired t-test analysis between dichotic and diotic conditions shown in Fig. 6 SNR -1 db SNR db t-value df p-value N πs (BBN)-NS (BBN) -.546 19.592 N πs (<1 khz)-ns (<1 khz) -2.88 19.5 N πs (1-3 khz)-ns (1-3 khz) 1.165 19.258 N πs (>3 khz)-ns (>3 khz).45 19.658 N πs (BBN)-NS (BBN) 1.696 19.16 N πs (<1 khz)-ns (<1 khz) -4.876 19.* N πs (1-3 khz)-ns (1-3 khz) -1.6 19.125 N πs (>3 khz)-ns (>3 khz) -2.268 19.35* N πs (BBN)-NS (BBN) -.181 19.858 N πs (<1 khz)-ns (<1 khz).3 19.841 SNR +1 db N πs (1-3 khz)-ns (1-3 khz) -1.165 19.258 N πs (>3 khz)-ns (>3 khz) -1.371 19.186 *:stands for statistically significant difference (p<.5). The subscript π in the figure stands for the binaural condition of 1 degree phase shift and for binaural in-phase condition. S and N in the figure denote signal and background noise, respectively 의영역도어음과잡음이합성된신호음의명료도에형향을미치는것으로보고되었다. 18) 이러한요인들이 Low pass filtered noise(<1 khz) 에서의변별에많은영향을미쳤을것으로판단된다. 본자료에서는신호대잡음비에의한영향은뚜렷하게나타나는반면에 BMLD effect 는의미있는차이가나타나지않았다. 이는 BMLD effect 가변별하려는음소간의 critical band 를벗어났기때문에, 그리고신호대잡음비의영향이 BMLD 효과보다상대적으로더우세했기때문이라고추정이되며, 이부분에대해서추가적인후속실험연구가필요하다고판단된다. Goldstein은청각시스템은 phase deaf 는아니며제한된범위내에서인지적으로중요하게작용한다고제안하였다. 15) 이 phase 는소리의변별에중요한요인으로작용할수있고, 최근에는복합음을비롯한어음에대한위상의효과에대한연구도진행되고있다. 15) 따라서위상은변별이어려운음소의변별수행에서와같이복합음변별을결정하는중요한요인으로작용할수있을것으로여겨진다. BMLD 는배경잡음이있을때, 잡음에대한신호음의변별을향상시키기위해양이의속성을사용하는청취자의능력을반영한다. 11) BMLD 는순음에서만측정되는것은아니라, 유사한효과가복합음, 클릭음, 어음에서도나타나는것으로보고되고있다. 17) 따라서 BMLD 와관련된잡음과신호음의위상변위는음소변별을향상시킬수가있다. 이런위상변위의효과를보다명확하게밝혀내기위해서는 BMLD 를여러형태의난청에적용하는연구도필요할것으로판단된다. 본실험에서사용된변별음소는종성자음이다른경우였다. 이와유사한형태의실험이초성자음, 중성모음에 서필요하고, 더나아가단어와문장이포함된실험도필요할것이다. 최근한쪽귀청취조건하에서의위상뿐만아니라양이청취조건하의위상이청각시스템상에서중요한요인으로고려되고있다. 1) 따라서양이청취조건하에서위상변위에대한더많은연구가이루어지면, 이결과들이양이청취조건하에서의새로운진단법, 보청기적합, 효율적인인공와우 mapping 등에임상적으로적용될수가있을것이다. 중심단어 :2AFC 양이차폐레벨차이 Critical band 양이청취. REFERENCES 1. Bronkhorst AW, Plomp R. Binaural speech intelligibility in noise for hearing-impaired listeners,.j Acoust Soc Am. 1989;86(4):1374-1383. 2. Bronkhorst AW, Plomp R. Effect of multiple speechlike maskers on binaural speech recognition in normal and impaired hearing. J Acoust Soc Am. 1992;92(6):3132-3139. 3. Culling JF, Edmonds BA, Hodder KI. Speech perception from monaural binaural information. J Acoust Soc Am. 6;119(1):559-565. 4. Gockel H, Carlyon RP. Dominance region for pitch: Effects of duration and dichotic presentation. J Acoust Soc Am. 5;117(3):1326-1335. 5. Greenwood DD. Auditory masking and the critical band. 1961;33 (4):484-52. 6. Grose JH, Poth EA. Masking level differences for tones and speech in elderly listeners with relatively normal audiograms. Journal of Speech & Hearing Research. 1994;37(2):422-428. 7. Hall JW, Buss E, Grose JH, Dev MB. Developmental effects in the masking-level Difference. Journal of speech, Language, and Hearing Research. 4;47:13-. 8. Hall JW, Harvey AD. N S and NoSpi thresholds as a function of masker level for narrow-band and wideband masking noise. J Acoust Soc Am. 1984;76(6):1699-173. 9. Henning GB, Gaskell H. Binaural masking level differences with a variety of waveforms. Hear Res. 1981;4(2):175-184. 1. Hong BN, Kim JS. Binaural benefits for children who use hearing aids and cochlear implants. Audiology. 5;1(1):19-27. 11. Jiang D, McAlpine D, Palmer AR. Detectability index measures of
DH Chung, et al:effects of Noise on Discriminating Gom and Gong G 39 binaural masking level difference across populations of inferior colliculus neurons. J Neurosci. 1997;17(23):9331-9339. 12. Johansson MS, Arlinger SD. Binaural masking level difference for speech signals in noise. Int J Audiol. 2;41(5):279-284. 13. Katz J. Handbook of clinical audiology. 5 th edition, Lippincott Williams & Wilkins, Baltimore;2. 14. Kent RD, Read C. The Acoustic Analysis of Speech, 2nd ed, University of Wisconsin-Madison, Madison, WI;2. 15. Koo S, Lim D. Effects of relative phases on discrimination of harmonic complex tones. Audiology. 5;1(1):35-39. 16. Meister H, Kluser H, v Wedel H, Walger M. A measurement system for assessing binaural masking level difference (BMLD) in children. HNO. 5;53(8):695-7. 17. Moore BC. an introduction to the psychology of hearing. 5 th ed, Academic Press;3. 18. Niederjohn RJ, Mliner DG. The effects of high-pass and of low-pass filtering upon the intelligibility of speech in white noise. J Aud Res. 1982;22(3):189-199. 19. Persson P, Harder H, Arlinger S, Magnuson B. Speech recognition in background noise: monaural versus binaural listening conditions in normal-hearing patients. Otol Neurotol. 1;22:625-63.. Sever JC Jr, Small AM Jr. Binaural critical masking bands. J Acoust Soc Am. 1979;66(5):1343-135. 21. Wilson RH, Hopkins JL, Mance CM, Novak RE. Detection and recognition masking-level differences for the individual CID W-1 spondaic words. J Speech Hear Res. 1982;25(2):235-242.