Dynamic Characteristics of Hearing and Its Applications

Transcription

1

2 Dynamic Characteristics of Hearing and Its Applications S. Namba a and S. Kuwano b a Takarazuka University of Art and Design, Japan b Osaka University, Japan In psychophysical experiments in laboratories, dependent and independent variables are well controlled. It leads to find a clear relation between both variables. When this relation is confirmed, we can propose a psychophysical law based on the results. On the other hands, in the case of practical situations, many factors contribute to determining subjective impressions and it is not easy to find exact law between subjective responses and physical parameters of noise. When we take the limit of a psychophysical law and combination of several laws into consideration, it is possible to apply the results of basic research to practical problems. Some examples are introduced in the case of frequency correction and the treatment of temporal variations in the evaluation of noise. INTRODUCTION Psychophysics is defined as the scientific study of the relation between stimulus and sensation [1].Several psychophysical laws and methods for measurement have been proposed. When a psychophysical law is valid, we can estimate the sensation from physicalmeasurement. This is very useful for applied purposes. For example, in the case of noise problems, the psychological effects of noise can be predicted from the results of physical measurement. However, it must be taken into account that psychophysical law usually has some limitations and that within the limit the law is valid. FREQUENCY CORRECTION Usually a good relationship is found between A- weighted sound pressure level and loudness. Sound level meter is a useful tool for the estimation of loudness of several kinds of environmental noise. When the noise consists of broad band frequency component without any prominent pure tone component, both A-weighted sound pressure level and loudness-level based on ISO 532B show good correlation with its loudness. An example is shown in Figs.1-a and b [2]. However, when the noise has strong pure tone components, there is some discrepancy between A-weighted sound pressure level and loudness. In this case, loudness level is a more appropriate index of loudness. As shown in Fig.2-a [3], A-weighted sound pressure level shows poor correspondence with the loudness of air-conditioner noise with pure tone components, while as shown in Fig.2-b, loudness level shows fairly good correspondence with the loudness. It is well known that sound energy is a good physical measure of loudness. Sone scale, the relation between sound energy and loudness, is a typical example of psychophysical law. In psychophysics, there have been long and severe discussions about the validity of ratio scaling of magnitude of sensation4,5. But the above examples concerning loudness scaling are not problems in ratio scaling of sound energy but those in the effects of frequency region. In laboratory situations, the physical parameters of stimuli are well controlled. But in practical situations, many factors other than energy may affect the results. A-weighting has only one filter, on the other hand, the method for calculating loudness level uses 24 filters based on the concept of critical bands. This concerns the advantage of critical bands than simple A-weighting. It is clear that when noise has pure tone components, A-weighting has poor ability for frequency correction compared with critical band theory. REPRENTATIVE VALUE OF LEVEL- FLUCTUTING NOISE There is another important problem concerning the difference between laboratory conditions and practical situations. In laboratory situations, steadystate sounds are usually used as stimuli since they can simplify the experimental conditions in time domain. This is an advantage of steady-state sounds in order to control physical conditions. However, in actual situations, almost all sounds are level-fluctuating.it is important to find what kind of statistical or mathematical value is appropriate as representative for the variation. To make clear, psychophysical experiments are needed using systematically controlled level patterns. The results of our former experiments suggest that the mean energy level is a good measure of the loudness of level-fluctuating sounds [6]. It means that the mean energy of level fluctuating sounds are the same, then the sounds have the same loudness even if the level patterns are different. This equal energy model is relatively robust, and it can be applied to the loudness of various kinds of non steady-state sound. In practical

3 purposes, the mean energy level corresponds to Equivalent Continuous A-weighted Sound Pressure Level (L Aeq ) which is standardized as ISO In former section, the priority of A-weighted sound pressure level and/or loudness level was discussed without considering level-fluctuation. ISO 532B is limited to apply to steady-state sounds, not to non steady-state sounds. If it is allowed to apply our equal energy model to ISO 532 B, this standard can be used for non steady-state sounds as well. It was found in our former experiments that the revised Loudness Level (abbreviated as LLz) is a good measure of the loudness of non steady-state sounds with prominent frequency components [7]. SUMMARY Basic research of psychophysics has contributed to noise evaluation. But the psychophysical laws based on the laboratory experiments have some limit of application to practical purposes. A combination of several laws from broader viewpoints makes useful application to noise problems. Some examples were introduced. REFERENCES [1] G. S. Gesheider, Psychophysics Method, Theory, and Application, (LEA, 1985). [2] S. Namba et al., J.A.S.J.(E), 14, (1993).. [3] S. Namba, et al., J.A.S.J. (E), 13, (1992). [4] S. S. Stevens, Psychophysics, (Wiley, 1975) [5] J. C. Baird, Sensation and Judgement, Complementary Theory of Psychophysics. (LEA1997) [6] S. Namba and S. Kuwano, J.A.S.J.(E), 5, (1983). [7] S. Kuwano, et al., Noise Cont. Eng. J., 33, (1989). Fig.1-a Fig.1-b Fig.2-a Fig.2-b

4 Assimilation and Asymmetry of Loudness Ratios for Increasing or Decreasing Sequences of Signal Intensities G. Canévet Laboratoire de Mécanique et d'acoustique, CNRS, F Marseille cedex 20. In magnitude estimation of loudness, assimilation produces different loudness ratios for level increases and decreases between successive sounds, as described here both theoretically and experimentally. As a consequence, due to assimilation, a given level change does not produce the same loudness change for an increase and for a decrease in level: A bias is observed, in favour of decreasing levels. Assimilation has been found to bias the judgments, in magnitude estimation measurements, in such a way that the loudness assigned to a sound is attracted towards the value assigned to the preceding sound (Ward, 1973; Cross, 1973). A well-known consequence of assimilation is that its produces an underestimation of the exponent of the loudness function. Another consequence of assimilation, that has received less attention and that is described here, is that the ratios between successive judgments are different if the level increases (than if it decreases) from one sound to the next. The purpose of the present study is to investigate the influence of assimilation on loudness ratios, both theoretically by using the current models of assimilation and loudness functions, and experimentally on the basis of original data. CALCULATED LOUDNESS RATIOS In magnitude estimation, assimilation pulls the number assigned to the loudness of a sound at a given level toward the number given on the preceding trial. Cross (1973) demonstrated that the series of numbers produced, for a sound at any given level, follows a power function of the level of the preceding sounds. Combined with the classical loudness function, this can be modelled by a product of two power functions: S i = k(p i ) (P i-1 ), S i being the loudness assigned to a sound at pressure P i preceded by a sound at pressure P i-1. Using such a model, we can calculate the ratio between estimations of any pair of levels, presented in either increasing or decreasing order. As shown below, these ratios also follow a power function of the preceding level, but the exponent of the function differs for increasing and decreasing orders of presentation. Assume that N subjects have been measuring the loudness of sounds at n levels, from L1 to Ln. For increasing orders, the average ratio of estimations for any given level Li preceded by L1 can be expressed as a product of N individual ratios : R Li,L1 =[(E Li,L1 )/(E Li,X1 )* (E Li,L1 )/(E Li,X2 )... *(E Li,L1 )/(E Li,XN )] 1/N in which E Li,L1 is the estimation of level Li when preceded by level L1, and E Li,Xj the estimation of level L1 when preceded by that particular level X that corresponded to the sequence presented to subject j (j=1,n). The same calculation holds for all ratios from R L1,L1 to R Ln,L1. It can be observed that the numerators correspond to the estimations of all levels from L1 to Ln when preceded by L1. They follow a power function of Li with the exponent mentioned above. The denominators, on the contrary, are the estimations of L1 when preceded by any other level. If N is big enough, one can evaluate the limit value of the denominators as the average value of the estimation for L1, which is a constant given by S 1 = k(p 1 ). Therefore, the data points corresponding to those ratios will follow a power function of the preceding level with the exponent. In real experiments, N is seldom big enough, but the results are close to this prediction, as will be seen in the next section. The equivalent ratio for decreasing orders is : R L1,Li =[(E L1,Li )/(E Li,X1 )* (E L1,Li )/(E Li,X2 ) *(E L1,Li )/(E Li,XN )] 1/N Now the numerators correspond to the estimations of level L1 preceded successively by all other levels. They thus follow a power function of Li with the exponent mentioned above. The denominators follow a power function with an exponent -. Finally, the data points will follow a power function with an exponent Increasing and decreasing ratios are equal for the highest level tested (Li=Ln); for other cases, decreasing ratios are bigger. An experiment was set up to test these results.

5 MEASURED LOUDNESS RATIOS The experiment was run in an isolated sound-proof room, using a Tucker&Davis workstation for sound generation, a VT 320 terminal for subject's responses, and a Sennheiser HD 545 headphone for sound presentation (monaural stimulation on right ear.) The subjects were three women and fourteen men, with ages ranging from 12 to 57 and an average of 28 years. A loudness function was obtained on 4-kHz test sounds at seven levels: 37.5, 45, 52.5, 60, 67.5, 75 and 82.5 db SPL. The sounds had a duration of 800 ms, with 30 ms of rise and fall times. The sequence of levels presentation was such that each level was presented once and only once after every other levels. A different order was used for each subject. The set of seven levels was thus estimated seven times, and a geometric mean of the seven estimations was calculated for each level and subject. Mean absolute threshold of the group, measured by Bekesy tracking, was 16 db SPL, with a standard deviation of 8 db. The measured loudness function shows the expected shape: The geometric means of loudness estimations can be well fit (correlation coefficient greater than 0.99) by a power function with an exponent of about Assimilation is visible for all levels, although with a variable strength: The exponents that vary from for the higher level (82.5 db) to for the 67.5-line. The average exponent is In summary, our set of data can be represented by the following function: S i = 0.1 [(P i ) 0.49 P( i-1 ) ]. Loudness ratios for all pairs are presented in Figs. 1 and 2. For the case of increasing order of level presentation (Fig. 1), the ratios are plotted as a function of the second sound of the pair, or level on trial N. For example, the upper line of data (unfilled squares in Fig. 1) is for all levels preceded by 37.5 db. As predicted by the calculations above, these data can be acceptably well fit (correlation > 0.98) by power functions. The corresponding exponents are fairly close to each other, with a mean of 0.52, not far from the predicted value of (=0.49). All lines in Fig. 1 have been plotted with this common slope. For the decreasing order, the ratios have been inverted: The data are plotted as a function of the preceding stimuli, or level on trial N-1. In this case, the upper line of data (unfilled squares) is for ratios of 37.5 preceded by all other levels. Again, the data can be fit by power functions. The average exponent is about 0.4, which is very close to the predicted value of (- ), with 0.49 and =0.86. Ratios of estimations 10 1 Figure 1 Ratios of estimations 10 1 Level in db SPL on trial N db 45 db 52.5 db 60 db 67.5 db 75 db 82.5 db INCREASING ORDER Common slope: Level in db SPL on trial N 37.5 db 45 db 52.5 db 60 db 67.5 db 75 db 82.5 db Level in db SPL on trial N DECREASING ORDER Common slope: Figure 2 Level in db SPL on trial N-1 In the case of signals with continuously sweeping intensity, assimilation may then bias the judgments of loudness change, if loudness change is directly related to the ratio, or difference, between end values of loudness in the sweep (Neuhoff, 1998; Canévet et al., 1999; Teghtsoonian et al., 2000.) This will be discussed at the meeting. ACKNOWLEDGMENTS The author is grateful to Mimi and Bob Teghtsoonian, and to Dominique Habault, for highly helpful advices in the development of this research. REFERENCES 1. Canévet G. Scharf B., Schlauch R., Teghtsoonian R. Teghtsoonian M. Nature, 1999, 398, Cross D. V. Perc. & Psychophys., 1973, 14, Neuhoff J. Nature, 1998, 395, Teghtsoonian R., Teghtsoonian M., Canévet G. Perception & Psychophysics, 2000, 62, Ward L. M. Perc. & Psychophys., 1973, 13,

6 Cochlear Hearing Loss Reduces the Dynamic Range for Loudness: Implications for Hearing Aids M. Florentine a,b and S. Buus a,c a Institute of Hearing, Speech and Language b Dept. of Speech-Language Pathology and Audiology (133 FR) c Communications and Digital Signal Processing Center, Dept. of Electrical and Computer Engineering (442 DA) Northeastern University, Boston, MA U.S.A. New psychoacoustical measurements in people with hearing losses of primarily cochlear origin indicate that the rate of loudness growth near threshold is normal. These data imply that loudness at threshold is greater than normal when threshold is elevated by a cochlear hearing loss. Therefore, persons with cochlear hearing losses not only have reduced dynamic range of audibility; they also have reduced dynamic range of loudness. These findings disagree with the classical definition of recruitment. Amplification strategies for the design of hearing aids are discussed. A NEW UNDERSTANDING OF LOUDNESS IN HEARING LOSS People with cochlear hearing losses cannot hear a lowlevel sound in the frequency range of their hearing losses. However, once the sound is about 30 db above threshold, it approaches normal loudness (although some loss of loudness often exists). This empirical phenomenon is called "recruitment." The concept of recruitment comes from loudnessmatching data in which the loudness of a tone in a frequency range of normal threshold is matched to the loudness of a tone in the frequency range of an elevated threshold. These data show that whenever a tone in the frequency range of an elevated threshold is raised 1 db, a tone in the frequency range of normal hearing may need to be raised 2-3 db to maintain equal loudness. Whereas such findings show that loudness level grows faster than normal, it does not necessarily follow that loudness itself grows faster [1]. Nevertheless, for over 60 years authors have erroneously concluded that cochlear hearing loss leads to an abnormally rapid growth of loudness immediately above an elevated threshold [2, for review, see 3]. This conclusion arose from the untested assumption that loudness at threshold is the same whether threshold is normal or is elevated by cochlear hearing loss. Recent evidence against this pervasive notion comes from an innovative psychoacoustical method using loudness matches between a tone and four- or ten-tone complexes, which provides a reliable way of measuring the rate of loudness growth near threshold [4]. Using this method, data from listeners with hearing losses of primarily cochlear origin show a normal rate of loudness growth at threshold [5]. Figure 1 shows loudness functions for a subset of these listeners. It is evident that the loudness functions near elevated thresholds of the listeners with hearing losses have slopes similar to that obtained near a normal threshold. Clearly, loudness does not grow more Loudness [sones] Tone Level [db SPL] FIGURE 1. Loudness functions derived from loudnessmatching data. Results are shown for average normal hearing (solid line) and three listeners with relatively flat hearing losses (dashed lines). The x s mark threshold and loudness at threshold for each individual or group.

7 rapidly near threshold in these listeners than in normal listeners. These data agree with loudness functions obtained at levels above 4 db SL for a large group of people with cochlear hearing losses [6]. Of course, it is possible that some types of cochlear losses may behave differently, but such cases are probably rare. Because loudness grows at a normal rate near threshold in listeners with cochlear hearing loss and approaches normal at high SPLs, loudness at threshold must be greater in listeners with cochlear hearing losses than in normal listeners. This is illustrated in Fig. 1. As shown by the x s, the loudness at threshold is larger than normal when the threshold is elevated by hearing loss. This finding indicates that listeners with cochlear hearing losses lose their ability to hear soft sounds, which reduces their dynamic range for loudness. overall impression of speech in the presence of various types of noise [8]. IMPLICATIONS FOR HEARING AIDS Two implications of these new findings on loudness are clear. First, complete restoration of normal loudness in people with cochlear hearing losses is an impossible task. If low-level sounds are made audible, they are likely to be louder than normal. Thus, it may not be beneficial to amplify very low-level sounds to audible levels. Second, the optimal compression ratio for people with cochlear hearing losses is likely to be less than that required to map the normal dynamic range into their reduced dynamic range in terms of SPL. This is consistent with clinical experience. Low compression ratios generally are found to be appropriate when fitting hearing aids with widedynamic-range compression [7]. A NEW UNDERSTANDING OF CONSUMER COMPLAINTS This new understanding of loudness in people with cochlear hearing losses gives insight into consumer complaints about hearing aids. For example, hearingaid users often complain that low-level sounds such as refrigerator noise are too loud. These subjective complaints are clearly understood in the light of these new data. Because loudness at threshold is greater in people with cochlear hearing losses [5] once a lowlevel sound is amplified so they can hear it, it will be louder than for normal listeners. Therefore, the subjective consumer complaint that "refrigerator noise is too loud" is validated in light of these data. Another consumer complaint relates to the compression ratios that are recommended by some hearing-aid manufacturers and fitters. Compression is often intended to map the normal dynamic range of sounds into the reduced dynamic range that lies between the elevated thresholds and the discomfort level of a person with cochlear hearing loss [7]. This goal is based on the erroneous assumption that the reduced dynamic range of SPL still encompasses the entire range of loudness experienced by a person with normal hearing. The compression ratios derived from such considerations are too large, because they map the entire range of loudness experienced by normal listeners into the smaller range of loudness that can be perceived by listeners with cochlear hearing losses. In fact, results from 20 people with sensorineural hearing loss reveal that increasing the compression ratio caused negative ratings on scales of clarity, pleasantness, background noise, loudness, and the ACKNOWLEDGMENTS This research was supported by NIH/NIDCD Grant No. R01 DC REFERENCES 1. S. T. Neely and J. B. Allen, Relationship between the rate of growth of loudness and the intensity DL, in Modeling Sensorineural Hearing Loss, edited by W. Jesteadt, Erlbaum, Mahwah, NJ, 1997, pp J. C. Steinberg and M. B. Gardner, J. Acoust. Soc. Am. 9, (1937). 3. M. A. Brunt, Tests of cochlear function, in Handbook of Clinical Audiology, 4th ed., edited by J. Katz, Williams and Wilkins, Baltimore, MD, 1994, pp S. Buus, H. Msch and M. Florentine, J. Acoust. Soc. Am. 104, (1998). 5. S. Buus and M. Florentine, A reexamination of loudness recruitment: Evidence for normal loudness growth near threshold in cochlear hearing loss, in Proc. 24th Mid- Winter Mtng. Assoc. Res. Otolaryngol., 2001, pp R. P. Hellman, Ear Hearing 20, (1999). 7. H. Dillon, Hearing Aids, Thieme, New York, A. C. Neuman, M. H. Bakke, C. Mackersie, S. Hellman and H. Levitt, J. Acoust. Soc. Am. 103, (1998).

8 Environmental Corrections to Acoustic Descriptions of Natural Sounds A. Preis Institute of Acoustics, Adam Mickiewicz University, Poznan, Poland It was assumed in all previous studies that it is sound itself that is of primary importance for people living in their typical environment. This view, however, should be corrected. Environmental sound is not only heard as a result of its impact on the auditory organs but is also an important source of information of the world outside. It is information about objects which, being the sources of sound, may influence the hearer s actions. The auditory system is capable of detecting from a variety of acoustic events detailed information about the distance of the sound source, its velocity, direction of its movement, and even its size and weight. This approach was applied in our study on the annoyance of the railway noise. It has been found that there is a close relationship between the velocity of the train and the sharpness of the noise generated by it. Consequently, the velocity of the moving sound source can be expressed in sharpness values, and used as a correction of the noise index. The paper discusses other cases where this approach could be applied as well. PSYCHOACOUSTIC RESEARCH AND AUDITION IN NATURAL SOUND ENVIRONMENT The psychoacoustic study concentrates on the low level audition. It tries to establish a direct relationship between the basic parameters of sound and the response of subject s auditory system. To eliminate interference of non-acoustic factors the psychoacoustic experiments are designed for highly idealized situations. In these situations elementary acoustic signals (tones) are presented and subjects responses studied. Generally, this way of conduct is effective and agrees with the methodological standards. However, in certain experiments this idealizing strategy goes too far and the obtained results are misleading or at best irrelevant. This happens, for example, when the research on pitch perception is confined to sinusoidal tones. In such experiments the stimuli used differ significantly from typical acoustic events occurring in natural sound environment. A subject exposed to such artificial signals responds artificially. Knowledge about reactions to arbitrarily designed signals can be useful only when it helps to account for the subject s response to the natural sound. We should bear in mind that psychoacoustics is not a self-contained discipline. Its aim is to contribute in building the full-fledged theory of audition. Therefore psychoacoustic research should be extended to signals which are clearly specified but which at the same time preserve the important traits of the sounds met in the natural environment. The aim of the present paper is to propose certain exemplary extensions of the psychoacoustic research program. PERCEPTION OF SOUND AND AN IDENTIFICATION OF THE SOUND SOURCE When subjects are asked to judge definite characteristics of sound such as loudness, pitch, timbre or duration they are expected to base their judgements solely on the heard signal and to neglect possible associations with auditory experiences evoked in the natural sound environment. Expectations of this kind can be best met when presented signals are exotic for the hearer. It is a mystery how the research on uncommon signals can contribute to our knowledge about the regularities of the human audition. When signals resemble natural sounds the demand to cut off the standard interpretations associated with them is hard to fulfil. This discrepancy between the expectations of the researchers and the subjects responses arises because of the false assumption. The psychoacoustic research is based on the assumption that the auditory perception is completed when characteristics of sound are detected. This picture of the auditory perception process is inadequate. It is a common knowledge that the processing of the visual information does not stop when the characteristics of the light wave are identified. The visual system works until it detects the information about objects in an environment from the light wave. Similarly, the ultimate aim of the auditory system is not the complete

9 information about sound events but the detection of the traits of objects that are the sources of the heard sounds. Our auditory system is shaped in the natural selection process and is prepared and well trained to detect such characteristics of the sound source as its location, its distance from the hearer, its velocity and even its weight and size. All these features of objects are decoded from the spectral and time characteristics of sound. Psychoacoustics can help to account for how the auditory system matches the features of objects with the specific characteristics of the sound signal. Below, I will consider how the auditory judgments of distance, velocity, weight and size of the moving sound source influence annoyance assessments. PERCEPTION OF DISTANCE We can expect that the perception of noise produced by sources located at different distances influences annoyance assessment of these auditory events. Noise produced by source located farther from a listener is assessed as less annoying than noise produced by source of the same type but located closer. There are three acoustic effects responsible for this auditory distance perception: decrease of sound energy with distance, selective attenuation by passage through the air and the ground effect. The papers on perceived auditory distance [1], [2] show that changes in sound level and/or in high-frequency spectral content can produce changes in the perception of an apparent distance from a source. To determine how the annoyance judgement depends on the perceived distance of the sound source we have to consider how changes in the frequency spectra and in sound level evoke the different auditory distance perception. Furthermore, it can be shown that the perception of sound source located at different distances may be evoked by equally loud noises that strongly differ in frequency spectra [4]. PERCEPTION OF VELOCITY The results of my recent research indicate, that the higher is the perceived velocity of the moving sound source the more annoying is noise evoked by this source. When the same sound source moves with different velocities the value of its L AE changes as well. The research confirms that from three main acoustic cues of the velocity of the moving sound source (convection effect, ITD, Doppler effect) [3], the most important is the Doppler effect. The relation between the velocity and L AE is not so simple and additional parameters have to be considered. For example, L AE of the train of 264m and passing at 126km/h is 97.3 dba, while L AE of the train of 257m and passing at 140km/h is 92.7 dba. Generally, perception of velocity of the moving source depends on the frequency shift which is larger for the faster sound sources. It can be added that approaching sound sources are perceived as more annoying than those which only pass by or recede. It means that increase in frequency that causes increase of annoyance is often associated with the detected increase in the velocity of a moving sound source. The figure 1 illustrates how sound sharpness correlates with the velocities of the moving sound source. sharpness [acum 2,0 1,5 1,0 0,5 0,0 FIGURE 1. Relation between the sound sharpness and velocity of the moving sound. PERCEPTION OF WEIGHT AND SIZE OF THE SOUND SOURCE When we listen to noise we usually try to estimate weight and size of its source. When cues in sound suggest that the source is large and heavy the noise produced by it is perceived as louder and in consequence as more annoying than noise associated with a source of a smaller weight and size. The typical example is noise of the passenger car and of the truck. Even if the L AE measured for the truck were the same as for the passenger car the sound produced by the former would be perceived as more annoying than the sound of the passenger car. It means that the increase of the number of the heavy vehicles on the road may result in an increase of the perceived annoyance even if the total value of noise index, such as L dn, will not change. REFERENCES y =0,0086x +0,2854 R 2 =0, velocity [km/h] [1] T. Z. Strybel, D. R. Perrott., J. Acoust. Soc. Am. 76, (1984). [2] A. D. Little, D. H. Mershon and P. H. Cox., Perception 21, (1992). [3] R. A. Lutfi, W. Wang., J. Acoust. Soc. Am. 106, (1999). [4] A. Preis, R. Golebiewski. Psychoacoustic correlates of time and spectral characteristics of railway noise. ICA 2001.

10 Loudness scaling of traffic noise: Perceptual and cognitive factors J. Hellbrück a, T. Kato b, A. Zeitler a, A. Schick c, S. Kuwano d, S. Namba e a Catholic University of Eichstätt, Germany b Otemon Gakuin University, Osaka, Japan c University of Oldenburg, Germany d University of Osaka, Japan e Takarazuka University of Art and Design, Japan There is convincing evidence that overall loudness judgment, i.e. loudness of long-term noise judged after listening to the whole noise, is higher than the average of continuous loudness scalings. These findings have been interpreted by Namba and Kuwano in terms of Gestalt Psychology, i.e. the judgment is supposed to be primarily based on prominent parts contained in the noise. The present investigation focuses on the role of memory in overall loudness scaling. In an experimental study we explored not only the relationship between overall loudness judgment and continuous instantaneous loudness scaling, but also the relationship between overall loudness judgment of long-term noise and the loudness of short meaningful auditory events included in the longterm noise. On the one hand we could confirm the above-mentioned effect that overall loudness is higher than the average of instantaneous loudness judgments, and on the other hand we could reveal a good correspondence between overall loudness and the mean loudness judgments of the auditory events included. INTRODUCTION There is no doubt that cognitive factors such as memory play an important role in global loudness judgments[1]. In the following study a comparison will be made between instantaneous judgments on loudness and overall loudness (OL) on the one hand, and on the other hand between OL and loudness judgments on short-term sections each comprising an individual auditory event. While instantaneous judgments are assumed to reflect loudness sensations based on sound energy transformation processes and echoic memory, loudness judgments on relatively long-term meaningful sounds, however, are assumed to need long-term memory scanning, and are, therefore, mainly based on nonsensory categorical memory codes. We expect good correspondence between OL and the mean of loudness judgments on auditory events since both are hypothetically based on long-term memory. Experiment 1 Experimental conditions Nine students and co-workers of Osaka University, aged between 22 and 48 years, participated in the experiment. All of them reported normal hearing. The stimulus was a traffic noise scene 20 minutes in duration, which was recorded by use of a dummy head near a village on a main road at a gated level crossing (L Aeq = 76.3 db) [2]. Sounds were presented via headphones in a sound proof room. Each subject was first exposed to the whole noise, instructed just to listen to the noise and to imagine the scene. After the sound had been fully presented, the subjects were requested to judge the OL of the whole sound by Category Subdivision Scale (CS Scale), which comprises the following five verbally distinguished categories and fine graduations: 1-10 ( very quiet ), ( quiet ), ( medium ), ( loud ), and ( very loud ). In a second trial some days later, subjects were requested to evaluate the same noise by continuously matching the loudness to line length. The task required the subjects to adjust a horizontal line displayed on a computer screen through moving the mouse. Beneath the line the CS scale was depicted in order to anchor subjects loudness judgment according to the loudness scaling procedure in the first session. Loudness judgments (measured in number of pixels) as well as the corresponding LA eq,100ms were registered every 100 ms. Immediately after the noise was finished, subjects were prompted to judge the OL by line length. Then they had to recall all events they could remember. Having finished this memory task, subjects were asked to judge the loudness with the CS scale again. Results and Discussion In the first session, the mean OL scaling value (CS scale) amounted to X=33.1 (SD=5.9).

11 For reason of comparison the individual line lengths in pixels gained in the second session were transformed into units of the CS Scale. In Figure 1 mean loudness scalings and standard deviations of the judgments in the course of time are depicted. Both are related to the right-hand y-axis (CS Scale). Sound pressure level L Aeq,100ms is related to the left-hand y-axis (Decibel Scale). db(a) Sound pressure level CS Mean SD Time (ms x 100) CS-units 60 FIGURE 1. Instantaneous loudness judgments and sound pressure level in the course of time. The OL judgment measured by line length was X=27.4 (SD=5.7) making a statistically significant difference to the first OL judgment by 5.76 scale units [t(8)=3.16; p<.01]. The third OL judgment, however, X=31(SD=3.1), was not significantly different from the first one. Experiment 2 Experimental Conditions Experiment 2 was carried out in a sound proof room of Oldenburg University. Ten normally hearing subjects participated in the study. We used the same noise as in Experiment 1. Except for the second trial, the procedure of Experiment 2 was the same as in Experiment 1. In the second trial, the noise was segmented into auditory events such as crossing cars and trains, passages of calm situations, etc. We identified 13 sections of the like, the duration of which lasted between 15 seconds and 4 minutes. Auditory events were separated by pauses of 5 seconds, and subjects had to judge the loudness of the preceding event during the pause. Results and Discussion For each of the 10 subjects in Experiment II the mean was calculated for the loudness judgments regarding the 13 auditory events (fig. 2, squares), which are compared to the individuals overall judgments. Likewise, instantaneous judgments of each of the 9 subjects in Experiment I were averaged, and plotted against the respective overall judgments as well (circles). Moreover, group means were calculated over both experimental groups (black symbols). It is shown that OL and mean loudness judgments of auditory events (squares) correspond very well. By contrast, instantaneous loudness judgments and respective OL (circles) differ widely in terms of substantially higher overall judgments, thus supporting the findings in [1]. Overall loudness ludgments (CS-units) Instant. judgm. - individuals Instant. judgm. - group Auditory events - individuals Auditory events - group Means of short-term loudness judgments (CS-units) FIGURE 2. Relationship between OL judgments and mean instantaneous loudness judgments or mean auditory events respectively. SUMMARY AND CONCLUSIONS Subjects use different judgment strategies depending upon the task and the duration of the sounds presented. Accordingly, instantaneous loudness judgments are assumed to be direct reflections of pure loudness sensations shortly stored. By contrast, OL judgments are due to the averaging of loud and soft events which are probably recalled in categorical terms from long term memory. ACKNOWLEDGEMENTS Part of the study was granted by JSPS (Japan Society for the Promotion of Science) and DAAD (German Academic Exchange Service). We are thankful to Andrea Hahn who carried out Experiment 2 for diploma thesis. REFERENCES [1] Namba, S. and Kuwano, S. The Journal of the Acoustical Society of Japan (E) (1), (1980). [2] Hellbrück, J. The Journal of the Acoustical Society of Japan (E) (21), (2000).

12 Ecological Psychoacoustics: Loudness Change, Distance Relations, and Evolution John G. Neuhoff Department of Psychology, The College of Wooster, Wooster, OH 44691, USA Many psychoacoustic experiments are conducted with stimuli and listening conditions that rarely occur in a natural listening environment. Although there are justifiable reasons for this practice, it is particularly troubling if one accepts the proposition that the auditory system has evolved specifically to deal with sounds that occur naturally. Recent work on loudness change, and auditory motion perception has revealed a bias in the perception of approaching acoustic sources. Specifically, rising intensity can be perceived to change more than equivalent falling intensity, and looming sources are perceived as closer than equidistant sources that recede. These findings are consistent with a mechanism that provides an evolutionary advantage by giving advanced warning of source approach. It is proposed that an ecological and evolutionary perspective in other areas of psychoacoustics would complement traditional psychophysical techniques by providing a richer understanding the relationship between the auditory system and the environment in which it evolved. ECOLOGICAL PSYCHOACOUSTICS Many psychoacousticians have focused their efforts on understanding the specific manner in which acoustical energy is transduced into an electrochemical signal in the nervous system, or on various psychophysical thresholds with stimuli that rarely occur in natural listening environments. Often the goal of this research is to better understand how the peripheral auditory system functions in order to develop better techniques to assist the hearing impaired. The obvious advantage to this approach is that one can gain a better understanding of how the hardware of the peripheral auditory system functions and perhaps develop more effective solutions for auditory pathology. The disadvantage is that the stimuli (often pure tones and noise bursts) and listening conditions employed (headphones, sound attenuating booths, and anechoic chambers) are often unrealistic, leaving the larger question of environmentally important listening behaviors and the linkage between perception and action relatively unexplored. A more recent approach to studying audition has emerged. Influenced by Gibson s ecological approach to studying perception and bolstered by a greater acceptance of evolutionary psychology, researchers have begun to explore listening under more realistic conditions. This perspective has recently been summarized by Gaver [1, 2]. The goal of this line of research has typically been to understand the higher-level processes that occur when a listener hears a sound or acoustic pattern. The focus of these investigations is often on what might be called listening behavior, or more specifically, the link between auditory perception and action. These investigations have shown that listeners can make reasonable estimates of many physical characteristics of sound sources. For example, listeners can use acoustic cues to discriminate object length [3], shape [4, 5], and even use higher order temporal properties to perceive and categorize dynamic events such as breaking, bouncing, and vessel filling [6, 7]. Listeners can also determine whether there is room to pass between a sound source and a barrier [8], and whether there is an occluding object between a sound source and the listener [9, 10]. Loudness, Distance, and Adaptation When asked to predict arrival time of an approaching sound source based on acoustic cues, listeners often err on the side of safety, expecting contact before the source arrives [11-14]. Guski [15] has suggested that this type of error should not be interpreted as such. He proposes that when a sound source approaches, the primary role of the auditory system is that of warning, either to direct the visual system toward the object if time allows, or to initiate appropriate behaviors to avoid the object. In a recent set of experiments listeners reliably overestimated the change in intensity of rising loudness tones relative to equivalent falling intensity tones [16]. In a natural environment this overestimation could provide a selective advantage, because rising intensity can signal movement of the

13 source towards a listener. Canévet and colleagues challenged the evolutionary implications of these findings, suggesting that loudness change in headphones is vastly different from the kind of acoustic change produced by an approaching source [17] To address these concerns, Neuhoff conducted three subsequent experiments [18]. Listeners heard approaching and receding sounds in an open field and indicated the perceived starting and stopping points of the source. Approaching sounds were perceived as starting and stopping closer than receding sounds despite the equal starting and stopping points. The evidence suggests that an asymmetry in the neural coding of approaching and receding auditory motion is an evolutionary adaptation that provides advanced warning of approaching acoustic sources. There is converging evidence that supports the role of the auditory system as a warning mechanism in processing approaching sources. Schiff and Oldak [12] examined accuracy in judging time-to-arrival. In one condition the source was on a collision course with the listener. In another the source traveled a path that would bypass the listener at a safe distance. From an evolutionary perspective one would predict greater underestimation of time-to-contact when the source is on a collision course. Listeners underestimated arrival time across all conditions. However, the underestimation was greater when the source was on a collision course with the observer. Furthermore, the further the angle of approach was from a collision course, the more accurate time-to-arrival estimates were. The authors suggested that such cautious estimates of arrival time would provide survival benefits, particularly in situations in which judgment errors are likely. A New Perspective in Psychoacoustics The relationship between the perception of acoustic source approach and the advantages of perceptual bias in a natural listening environment suggest that an ecological and evolutionary perspective might benefit other areas of psychoacoustics. REFERENCES 1. W. W. Gaver, "How do we hear in the world? Explorations in ecological acoustics," Ecol. Psychol. 5, pp , (1993). 2. W. W. Gaver, "What in the world do we hear? An ecological approach to auditory event perception," Ecol. Psychol. 5, pp. 1-29, (1993). 3. C. Carello, K. L. Anderson, and A. J. Kunkler- Peck, "Perception of object length by sound.," Psychol. Sci., 9, pp , (1998). 4. S. Lakatos, S. McAdams, and R. Causse, "The representation of auditory source characteristics: Simple geometric form.," Percept. & Psychophys., 59, pp , (1997). 5. A. J. Kunkler-Peck and M. T. Turvey, "Hearing shape.," J. Exp. Psychol.: Hum. Percept. & Perf. 26, pp , (2000). 6. W. H. Warren and R. R. Verbrugge, "Auditory perception of breaking and bouncing events: A case study in ecological acoustics," J. Exp. Psychol.: Hum. Percept. & Perf., 10, pp , (1984). 7. P. A. Cabe and J. B. Pittenger, "Human sensitivity to acoustic information from vessel filling," J. Exp. Psychol.: Hum. Percept. & Perf., 26, pp , (2000). 8. M. K. Russell and M. T. Turvey, "Auditory perception of unimpeded passage," Ecol. Psychol., 11, pp , (1999). 9. M. K. Russell, "Acoustic perception of sound source occlusion," in Studies in Perception and Action IV: Ninth International Conference on Perception and Action, M. A. Schmuckler and J. M. Kennedy, Eds. Mahwah, NJ: Erlbaum, (1997), pp H. Ader, "Ein neues Hoerphaenomen.," Monatsschrift fuer Ohrenheilkunde, pp. 7, (1935). 11. L. D. Rosenblum, A. P. Wuestefeld, and H. M. Saldana, "Auditory looming perception: Influences on anticipatory judgments.," Perception, 22, pp , (1993). 12. W. Schiff and R. Oldak, "Accuracy of judging time to arrival: Effects of modality, trajectory, and gender," J. Exp. Psychol.: Hum. Percept. & Perf., 16, pp , (1990). 13. L. D. Rosenblum, M. S. Gordon, and L. Jarquin, "Echolocating distance by moving and stationary listeners.," Ecol. Psychol., 12, pp , (2000). 14. D. H. Ashmead, D. L. Davis, and A. Northington, "Contribution of listeners' approaching motion to auditory distance perception," J. Exp. Psychol.: Hum. Percept. Perf., 21, pp , (1995). 15. R. Guski, "Acoustic tau: An easy analogue to visual tau?," Ecol. Psychol. 4, pp , (1992). 16. J. G. Neuhoff, "Perceptual bias for rising tones," Nature, 395, pp , (1998). 17. G. Canévet, B. Scharf, R. S. Schlauch, M. Teghtsoonian, and R. Teghtsoonian, "Perception of changes in loudness," Nature, 398, pp. 673, (1999). 18. J. G. Neuhoff, "An adaptive bias in the perception of looming auditory motion," Ecol. Psychol., 13, pp , (2001).

14 Application of Psychological Findings to Musical Composition: Its Possibility and Limitation Y. TANAKA Department of Psychology, Tokyo Metropolitan University, 1-1 Minamiosawa, Hachioji, Tokyo, , Japan Although the importance of collaboration between composers and psychologists is claimed in exploration in music, application of psychological findings to actual musical activity is quite limited. In particular the application to contemporary musical composition is limited to providing sound materials in the genre of electro-acoustic music. In this presentation we consider the reasons of this limitation and discuss the possibility and limitation on the application of psychological findings to composition in contemporary music and the collaboration between composers and psychologists. It has been claimed that collaboration between composers and psychologists is important both in understanding musical behavior for psychologists and in exploration of new possibility of musical expression for composers. For example, various acoustic phenomena in music (e.g., consonance, timber perception, etc.) gave psychologists the clues to understand acoustic process. However, application of findings in psychoacoustics (more generally, psychology of auditory perception and cognition) to actual musical activity seems to be quite limited. In this presentation, we consider in what way psychological findings are applied to actual activity of composition, and what the possibility and the limitation are. Here we focus on composition in contemporary music, and in practical aspect of the activity. WHERE CAN WE APPLY PSYCHOL- OGY TO COMPOSITION? The most well known application of psychological findings to composition is that in electro-acoustic music. Risset emphasized that findings in psychoacoustics were crucially important in sound synthesis [1]. According to him, sound synthesis by a computer revealed the complexity of psychoacoustic" correspondence between physical parameters and sensible effects. In this sense, psychoacoustics provided various tools to control sound materials. For example, the studies with multidimensional scaling method for the relation between timbral impression and physical parameters are very useful to consider the synthesized sound. A study on subjective duration and envelope pattern [2] has implications to music making by computer software. Findings of psychoacoustics also provide new sound materials that were never imagined in the traditional music. For example, some sorts of acoustic illusions revealed completely new experience that is not realizable by traditional means, as infinitely descending (Fall by Risset) or ascending (For An / Rising by James Tenney) sound, or spatial movement of sound source (by John Chowning) [1]. WHY THE APPLICATION IS LIMITED? Although psychological data essentially contributed to provide and control sound materials in the electroacoustic music, it is difficult to find application to other genre of contemporary music (music by traditional instruments) and higher level of musical structure. Risset pointed out that specificity of traditional instruments was so familiar to us that our ears were inscribed the properties of musical instruments [1]. Another reason why electro-acoustic music is the limited genre of application of psychological findings might be the fact that the newly discovered sound materials shown in the previous section require the strict control of physical parameters, which is difficult for human performers to realize. Researches on perception of sound material use much simpler situation than actual listening situation for the strict control of stimuli. But, in the real musical pieces, materials are variously combined with other materials and listened by different subjects in different contexts. As often pointed out, perception and cognition of musical pieces are not the same as those of single sound material. Then the control of sound materials is not sufficient to control perception and cognition of higher level of the musical structure. But it is difficult to control perception of higher levels of structure [3]. Recent researches in music psychology have shown that perception of musical material and structure (e.g., pitch, interval, or tonality) is strongly affected by listeners' musical knowledge or experience. How one listens to a musical piece is also affected by listeners attitudes, which often depend on cultural and historical background or on composers esthetics. In short, perception of a musical piece is sensitive to various factors out of control of psychological experiments (and interests). Studies on cognition of higher level of musical structure might help composers. However, most of current studies on the higher level processing of musi-

15 cal information are based on music theory of traditional western tonal music, which is less interesting to composers of contemporary music. WHAT SEPARATES US? More fundamental obstacle to the collaboration is the gap in interest between composers and psychologists. Roughly speaking, psychologists' interest is in explanation of phenomena in terms of mental processes. They want to know how our mind works to achieve such perception. On the other hand, composers' interest is whether the phenomena are esthetically interesting or not. For example, auditory illusions like phonetic restoration [4] or illusory continuity [5] are interesting to understand an auditory system because of a discrepancy between our knowledge on physical state of the stimuli and their percepts, but they may not be interesting esthetically in their percepts themselves. WHEN CAN WE WORK TOGETHER? if the composers idea is not confirmed by a psychological experiment, it does not mean that the musical piece is not valuable esthetically. REFERENCES 1. J-C. Risset, Musique et perception, In Quoi? Quand? Comment? La recherche musicale, edited by T.Machover, Christian Bourgois Editeur, Paris, 1985, pp S. Namba et al., Journal of Music Perception and Cognition, 4, 71-80(1998). 3. F. Delalande, Perception des son et perception des oevre, In Quoi? Quand? Comment? La recherche musicale, edited by T. Machover, Christian Bourgois Editeur, Paris, 1985, pp R.M.Warren, Science, 167, (1970). 5. G.L. Dannenbring, Canadian Journal of psychology, 30, (1976). 6. D. Pressnitzer et al., Perception and Psychophysics, 62, 66-80(2000). As we have seen above, it is difficult for composers and psychologists to share a fruit. But recently Pressnitzer et al. published interesting research, which might show one possible way to collaboration between psychologists and composers [6]. Pressnitzer and colleagues hypothesized that tension-relaxation relation of non-tonal chords corresponded to their perceptual roughness. In their experiment, subjects listened realistic stimuli taken from an instrumental piece by Joshua Fineberg (a composer and one of the members of Pressnitzer s group), and evaluated its perceptual roughness and relative tension-relaxation. The result showed that relative roughness and tension-relaxation were correlated to each other, and they corresponded to relative tension of chords defined by Fineberg s compositional theory (algorithm). This research is very interesting in the following points. First, they provided an explanation to the prediction of Fineberg s theory in terms of purely psychological concept, perceptual roughness. This was also possible because Fineberg s theory was well formalized to predict the outputs percept. Second, their findings could be applicable to refine the compositional theory. The composer can use a psychological theory on perceptual roughness to refine the theory to control their material. As we see in this example, for fruitful collaboration, it is required for psychologists to connect the composers interest to psychological concepts. Also it is necessary for composers to clarify the target of their control and formalize their theory (or method). Nevertheless, it does not mean that composers must do so, because composers interest often goes beyond the scope of psychology. It should be also noted that even

16 Measurements, Control and Precision in Music Performance J. Tro Department of Telecommunications, Acoustics, Norwegian University of Science and Technology, N-7491 Trondheim, Norway. The Art of Music Listening very often prescribes additional knowledge and information about music history, musical styles and timbre differences among musical instruments - all with the aim of improving your ability to listen analytically. Even if music recording methods and techniques are well developed for storing and distributing purposes, the traditional time and frequency domain signal descriptions seem to be an incomplete set of data when we are looking for relations between the music stimulus production and the listening experience. The present demand for virtualization of audio environment and sources has made the question about sound and music controllers and parametric instrument control an important one. On this background the Art of Music Performance has grown into a scientific research topic that includes a lot of different issues, such as acoustical measurement technology, sound and event recording methods, man-machine interaction, data protocols and evaluation techniques. However, it is of utmost importance to find suitable and reliable data for simulations, modeling and virtualizations. This paper presents piano performance data including anechoic recordings and simultaneously recorded audio and MIDI performances. Important keywords for the discussion are measuring procedures, intentional performance control, control parameters and control ability and deviation. INTRODUCTION Today Music Performance has become a scientific research topic that includes a lot of different issues, such as acoustical measurement technology, sound and event recording methods, man-machine interaction, gesture intention and control, data protocols and evaluation techniques. The increasing interest for understanding the link between music perception processes and performance based artistic intentions has led to a lot of projects dealing with gesture registrations and event recordings. The introduction of MIDIized acoustical instruments made it an easy task to plan data acquisition procedures with an acceptable quality and reliability [1] even if piano mechanics and piano performing techniques have been studied for a long time [2]. DATA ACQUISITION With the present experimental setup (Figure 1) it is possible to incorporate both acoustical as well as MIDI event data suitable for comparison and different analyzing procedures. We can have - One performer playing several pieces of music. Data are suitable for style and genre analyses as well as individual fingerprint definition. One performer playing one piece repeatedly. Data are suitable for statistical definitions of individual deviation. Many performers playing several pieces of music. Data are suitable for style and genre analyses including style variation as well as individual fingerprint definition. Many performers playing one piece repeatedly. Data are suitable for statistical definitions of variation among performers and individual deviation. Performers playing under varying acoustical conditions. Data are suitable for analyzing performance details affected by acoustical feedback factors. And so on Disklavier MIDI in Remote keyboard Mic MIDI out FIGURE 1. Recording Setup. DAT This presentation will touch aspects such as measurements, instrumental mechanics, intentional artistic control and performer deviation. PC

17 MEASUREMENTS Two crucial questions are: Why do we measure a musical performance? How do we measure a musical performance? One simple answer is that simulations based on physical models and, in general, the virtualization of human behavior, i.e. the curiosity driven analysis-by-synthesis research, need basic and realistic input data. The second question is much harder to answer. Measurement procedures may vary from simple event recordings to complex audiovisual documentation. However, the ultimate data acquisition procedure should not only answer the question what happened? It should even tell us why did it happen? In such a context it seems to be correct and important to include details about both performer s intention and physiological and physical factors in the sound source excitation process. In 1929 Ortman [2] presented detailed finger, hand and arm movements for a piano performer registered by the help of a pantograph. controller. As the performing conditions were quite different, we may expect different analysis data. MIDI Key Velocity Performance no. 1, 2 and 3. FIGURE 3: Average MIDI Key Velocity (± STD) for 3 performances (Sinding Serenade). The overall analysis of the Sinding Serenade indicates a difference in performance duration with 96 s, 98 s and 90 s for performance no. 1, 2 and 3, respectively. Figure 3 shows a significant difference in average MIDI Key Velocity, probably due to the different acoustical feedback intensity during the performance. The oral presentation will include lots of details from the Sinding Serenade analyses, emphasizing differences in performance environment, musical dynamics and performer variation. FIGURE 2. Motion patterns of 4 drummers [4]. Similar registrations of the accentuated attack gestures of drummers are presented in Dahl [4] (figure 2). The present experimental equipment setup (figure 1) gives us simultaneously recorded MIDI-data and anechoic high quality digital sound. Gesture data analyses compared with MIDI and acoustical data give us insight in the necessary motion and force and the resulting signal variation and deviation, i.e. control ability. DISCUSSION The present experiment included three recording sessions. The 1 st and the 2 nd intended to be identical anechoic Disklavier performances (identical music performed as close to identical as possible). The 3 rd session was a repetition of the previous one performed in a different acoustical environment using a Roland A-90 keyboard as a Disklavier remote ACKNOWLEDGMENTS This paper is part of the documentation of the ongoing Music Technology research project at the Acoustics Group, NTNU. Discussions and comments from Profs. Ulf Kristiansen and Peter Svensson have been highly appreciated. REFERENCES 1. J. Tro, Data Reliability and Reproducibility in Music Performance Measurements, WESTPRAC VII Conference, Kumamoto 3-5 October O. Ortmann, The Physiological Mechanics of Piano Technique, ISBN , London/New York, 1929, (DaCapoPress reprint 1981). 3. S. Dahl, The Accented Stroke, - Kinematics and Timing, The Gesture Workshop, Centro Tempo Reale, Firenze, 30 April J. Tro, Control Aspects, - Keyboard Precision and Efficiency, The Gesture Workshop, Centro Tempo Reale, Firenze, 30 April 2001.