Directional Sensitivity of Sound Pressure Levels in the Human Ear Canal

Directional sensitivity of sound-pressure levels in the human ear canal

John C. Middlebrooks

Departments of Neuroscience and Surgery (ENT), University of Florida, Gainesville, Florida 32610

James C. Makous

Department of Neuroscience, University of Florida, Gainesville, Florida 32(510

David M. Green

Psychoacoustics Laboratory, University of Florida, Gainesville, Florida 32610

(Received 18 October 1988; accepted for publication 14 March 1989)

Changes in sound pressures measured in the ear canal are reported for broadband sound sources positioned at various locations about the subject. These location-dependent pressures are one source of acoustical cues for sound localization by human listeners. Sound source locations were tested with horizontal and vertical resolution of 10 ø. Sound levels were

measured with miniature microphones placed inside the two ear canals. Although the measured amplitude spectra varied with the position of the microphone in the ear canal, it is shown that the directional sensitivity at any particular frequency of the broadband stimulus is independent of microphone position anywhere within the ear canal. At any given frequency, the distribution of sound pressures as a function of sound source location formed a characteristic spatial pattern comprising one or two discrete areas from which sound sources produced maximum levels in the ear canal. The locations of these discrete areas varied in horizontal and vertical location according to sound frequency. For example, around 8 kHz, two areas of maximum sensitivity typically were found that were located laterally and were separated from each other vertically, whereas, around 12 kHz, two such areas were found located on the horizontal plane and separated horizontally. The spatial patterns of sound levels were remarkably similar among different subjects, although some frequency scaling was required to accommodate for differences in the subjects' physical sizes. Interaural differences in sound-pressure level (ILDs) at frequencies below about 8 kHz tended to increase monotonically with increasing distance of the sound source from the frontal midline and tended to be relatively constant as a function of vertical source location. At higher frequencies, however, ILDs varied both with the horizontal and with the vertical location of the sound source. At some frequencies, asymmetries between the left and right ears in a given subject resulted in substantial ILDs even for midline sound sources. These results indicate the types of horizontal and vertical spatial information that are available from sound level cues over various ranges of frequency and, within a small subject population, indicate the nature of intersubject variability.

PACS numbers: 43.66.Ba, 43.66.Pn, 43.66.Qp, 43.63.Hx [WAY]

INTRODUCTION

In contrast to the visual system, in which the locations of stimuli are mapped directly onto the retinas, the auditory system has no topographic representation of sound locations at the sensory periphery. Instead, the locations of sounds must be computed centrally from acoustical cues present at the two ears. Temporal and intensive cues for sound location are created by the interaction of the sound wave with the torso, head, and external ears. The purpose of this study was to characterize the intensive cues for sound locationsthe characteristics of the amplitude spectrum and of the interaural difference spectrum that vary with sound source location. Several previous studies have characterized the amplitude transfer function of the auditory periphery (e.g., Wiener and Ross, 1946; Shaw, 1966; Blauert, 1969/70; Searle et al., 1975; Mehrgardt and Mellert, 1977). However, these studies all used sound sources restricted to one of three

cardinal planes, either the horizontal plane, the vertical midline plane, or the plane defined by the vertical axis and the interaural axis. The spatial cues for locations outside of these planes cannot be predicted from published data. In this pa- per, we report on the sound:pressure levels measured in the ear canal for 356 source locations separated by 10 ø in both the horizontal and vertical dimensions.

The description of the acoustic stimulus in the ear canal of a listener is a complicated business, especially at higher fi'equencies at which the wavelength of the sound ap- proaches the dimensions of the canal. The cross-sectional area of the typical ear canal is small enough that the primary mode of propagation is a plane progressive wave at least for frequencies below a certain cutoff frequency (see Kuhn, 1987). Morse and Ingard (1968, p. 511) state that, for a rigid tube, the cutoff frequency is about 0.58c/d, where c is the velocity of propagation in free space, and d is the diame-

89 J. Acoust. Soc. Am. 86 (1), July 1989 0001-4966/89/070089-20500.80 @ 1989 Acoustical Society of America 89

Downloaded 06 Apr 2012 to 83.50.155.173. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/terms.jsp

ter of the tube. For an average person, the diameter of the canal is less than 1 cm and the typical cutoff frequency will, therefore, exceed 20 000 Hz. Higher modes of propagation occur only at frequencies above this cutoff frequency or near physical irregularities in the tube, such as the entrance or near the tympanic membrane, where the tube tapers to a point. In the vicinity of these irregularities, cross modes may be present (Rabbitt and Holmes, 1988). These evanescent waves propagate over a limited distance, and their influence on the plane progressive wave is minimal. Thus our aim was to measure the pressure level of this plane progressive wave. Admittedly, the presence of the microphone could disturb the normal distribution of pressure. We report on measurements using two microphones that provide an upper bound on the error introduced by these disturbances.

Since any tube will have longitudinal standing waves, one might assume that the measurements should be made as close as possible to the taper in the tube where the tympanic membrane is located. Because of cross modes present near this taper, and because the tympanic membrane is an ex- tremely sensitive structure, we were forced to make our measurements at some distance from the end of the tube. This

fact makes it difficult to infer the amplitude spectrum at the tympanic membrane from a measurement made elsewhere in the canal. However, several previous studies (Wiener and Ross, 1946; Shaw and Teranishi, 1968; Mehrgardt and Mel- lert, 1977), as well as our own observations, indicate that the dependence of sound pressures on source location is independent of the position of a recording microphone in the ear canal. At any given frequency, the position of the microphone will influence the absolute level of the recorded pressure, but the relative changes in that level with changes in sound source location will be the same for all microphone positions.

The general approach used in this study was to present sounds from a number of loudspeakers located about the subject's head and to record from a miniature microphone positioned in the ear canal. The use of multiple speakers (all calibrated to produce identical spectra), rather than a single movable speaker, reduced the time necessary to sample the desired number of locations to about 6 min. This eliminated

the need to restrain the subject's head with a bite bar or head rest, which might have interfered with the sound field. We used a broadband stimulus, bandpassed to include only the frequencies of interest. This enabled us to test a range of frequencies simultaneously, yet it concentrated energy in a limited band, thus improving the signal-to-noise ratio. Using this approach, we were able to sample sound source locations in all directions around the subject in 10 ø horizontal and vertical increments. The results reveal the types of spatial information that are available from intensive cues over

various ranges of frequency and, within a small subject population, indicate the nature of intersubject variability.

I. METHODS

We measured sound-pressure levels inside the human ear canal as a function of the location of a broadband sound

source presented in a free sound field. Measurements were obtained from six volunteer subjects, three male and three

dght go'.o'

FIG. 1. Diagram of the spatial coordinate system. The locations of sound sources are represented on a unit sphere as if looking in toward the subject from a location 30 ø above the horizontal plane and 30 ø to the subject's right. Locations are given by the angles measured at the center of the subject's interaural axis: Horizontal locations are given by azimuth, the angle to the right ( d- ) or left ( - ) of the vertical midline plane, and vertical locations are given by elevation, the angle above ( d- ) or below ( - ) the horizontal plane. Isoazimuth and isoelevation lines are shown in 20 ø increments. The loudspeakers, represented here by small circles, were separated in azimuth by 10 ø.

female. They were all Caucasian, ranging in age from 25-35 years. Their external ears were completely normal in appear- ance. Their ear canals were free of visible abnormalities or

accumulations of cerumen. Thus this group's external ears and ear canals are probably more homogeneous than one would find in a random sample from the general population.

A. Experimental apparatus

Experiments were conducted in a 2.7-m X 4.3-m X 3.1- m room, the walls of which were covered with 3-in. acousti-

cal foam (Illbruck) to suppress sound reflections. Ambient noise levels were 40 dB SPL (A-weighting scale). For most of the measurements, sound stimuli were presented from a circular array of 36 loudspeakers (Fig. 1 ). The array was 1.2 m in radius, and the loudspeakers were spaced in horizontal intervals of 10 ø. The audio signal could be directed to any one of the loudspeakers by use of a system of computer-controlled relays, thus controlling the horizontal location of the stimulus. The plane of the array of loudspeakers could be rotated about a horizontal axis, thereby providing control of the vertical stimulus location. For one series of measure-

ments, stimuli were presented from a vertical semicircular array of loudspeakers positioned in the vertical midline plane. The loudspeakers in that array were spaced in vertical intervals of 10 ø from -- 40 ø to q- 90 ø.

The orientation of the subject's head during the recordings was monitored using an electromagnetic device (3SPACE Isotrak, Polhemus Navigation Sciences Division, McDonnell Douglas Electronics Company). This device generated an alternating electromagnetic field from a fixed

9O J. Acoust. Soc. Am., Vol. 86, No. 1, July 1989 Middlebrooks eta/.: Ear directionality 90


source, which induced a current in a sensor. The sensor was embedded in a plastic block, with 27 mm as its largest dimen- sion. It was held on the subject's head with a cloth headband. The Isotrak device provided measurements of the azimuth, elevation, and roll of the subject's head with uncertainty less than one-half degree.

B. Spatial c. oordinate system

All sound sources lay on the surface of an imaginary sphere, 1.2 m in radius, concentric with the loudspeaker array (Fig. 1 ). The locations of sound sources are specified in a double-pole coordinate system in Which locations are referred to two planes, the interaural horizontal plane and the vertical midline plane; this coordinate system has been used in several physiological studies of auditory spatial sensitivity (e.g., Knudsen, 1982; Middlebrooks and Knudsen, 1984). The reference planes are defined by structures of the testing chamber and the loudspeaker array although, to the degree that the subject's head was upright and centered in the loudspeaker array, they also correspond to anatomical planes of the subject. The interaural horizontal plane is parallel to the floor of the testing chamber and contains the axis of rotation of the loudspeaker array, which coincides with the subject's interaural axis. With a subject's head in an upright position, his/her orbito-meatal plane (i.e., Frankfort plane) was tipped roughly 10 ø up from the interaural horizontal plane. The vertical midline plane, which coincides with the subject's midsagittal plane, is the vertical plane that is orthogo- nal to the axis of rotation of the loudspeaker array.

Horizontal locations are specified by the azimuth coordinate, which is the angle subtended at the center of the subject's head to the right ( + ) or left (-) of the midline plane. As shown in Fig. 1, lines of constant azimuth form circles that are centered on the interaural axis. Lines of con-

stant azimuth correspond to the paths traversed by the loudspeakers as the loudspeaker array is rotated. They also correspond to the projections onto the coordinate sphere of the hyperboloid "cones of confusion,'" which are often discussed in relation to interaural delays (e.g., Blauert, 1983). All points on the coordinate sphere that lie at a given azimuth lie within a single parasagittal plane. Vertical locations are specified by the elevation coordinate, which is the angle above ( + ) or below ( - ) the horizontal plane. Lines of constant elevation lie in horizontal planes and form circles that are centered on the vertical axis through the subject's head. These isoelevation circles coincide with the circular

loudspeaker array only at 0 ø elevation. The double-pole coordinate system has the advantage

that azimuth and elevation are mutually independent, in that a given angle in azimuth always corresponds to a constant arc length on the coordinate sphere, irrespective of elevation. In this respect, it differs from a more conventional single-pole coordinate system. In a typical single-pole system (e.g., as used by Wightman and Kistler, 1989), locations on the coordinate sphere are specified by two angles that are analogous to geographical longitude and latitude. The horizontal location, analogous to longitude, is the angle around the vertical axis. The vertical location, analogous to latitude and identical to our elevation, is the angle above or

below the horizontal plane. Lines of constant longitude all converge on the vertical axis, so that, for a given increment in longitude, the horizontal angle subtended at the subject's head varies with latitude (elevation). For example, at + 80 ø elevation, two sound source locations separated by 180 ø in longitude would subtend an angle of only 20 ø at the center of the subject's head.

C. Sound stimulation and recording

Sound stimuli were generated digitally, using an IBM PC/AT computer with a 12-bit digital-to-analog converter (Modular Instruments, Inc.) running at a rate of 50 000 samples per second. The audio signal was passed through a computer-controlled attenuator (Wilsonics) and a Macin- tosh power amplifier (model MC40). The loudspeakers were piezoelectric tweeters (Motorola model KSN 1072A, available as Radio Shack Cat. ½/:40-1379).

Measurements of sound pressures were made using a special computer-generated waveform. This special waveform was produced using an inverse fast Fourier transform. The waveform was 10.24 ms in duration and contained fre-

quency components spaced in intervals of 97.7 Hz. The waveform contained energy in the frequency band between 3 and 16 kHz, and we will refer to this frequency band as the "stimulus band." The amplitude of each component in the stimulus band increased by 6 dB per octave to roughly com- pensate for the frequency rolloff that was encountered in recording from the ear canal with miniature microphones. The phase spectrum of the special waveform had all the even components in cosine phase and all the odd components in negative cosine phase. This combination of amplitude and phase spectra produced a clicklike stimulus that had negligi- ble voltage excursions at stimulus onset and offset, thus eliminating the need to shape the envelope of the waveform to avoid onset transients.

The loudspeakers were calibrated using a B&K sound level meter with a B&K model 4133 microphone centered in the loudspeaker array; the analog output of the sound level meter was digitized at 50 kHz with 12-bit resolution. A digital output buffer was tailored for each speaker using a computer program that adjusted the amplitude and phase of each frequency component until the amplitudes and phases measured by the B&K microphone matched those of the special calibration waveform to within criteria of +_ 1 dB and +_ 10 ø, respectively. Figure 2(a) shows the time waveform of the stimulus as produced by one of the loudspeakers and measured with the B&K sound level meter. Figure 2(b) shows the corresponding amplitude spectrum. As we changed the orientation of the loudspeaker array, the recorded spectra also changed, since the testing chamber was not perfectly anechoic. These changes were slight, however, in the frequency band used in these measurements. Figure 2 (c) and (d) shows the waveform and amplitude spectrum measured when the loudspeaker array was rotated vertically by 70 ø relative to its position when the speaker was originally calibrated. The illustrated deviations in amplitude, less than _+ 1.5 dB, are representative of the largest that we encoun-

tered.

Sound levels inside the ear canal were measured with

Knowles model EA-1934 microphones (Knowles Electron-

91 J. Acoust. Soc. Am., Vol. 86, No. 1, July 1989 Middlebrooks eta/.' Ear directionality 91


!

(a)

I I , I I I I

Time (ms)

o (b) -lO

-20

•.--50 ......................

0

-50 0 5 10 15 20 25

Frequency (kHz) FIG. 2. Time waveforms and amplitude spectra of the stimulus. These waveforms were produced from the loudspeaker at 0 ø azimuth and recorded with a B&K microphone at a position corresponding to the center of the subject's head in the testing apparatus. The waveform in 2(a) and the spectrum in (b) were recorded with the loudspeaker array oriented in the horizontal plane, the orientation for which all of the loudspeakers were calibrated. The waveform and spectrum in (c) and ( d ) were recorded with the array rotated upward by 70 ø so that the sound source was located at 0 ø azimuth and + 70 ø elevation. Pressures in (b) and (d) are normalized relative to the maximum pressure level in each spectrum.

ics, Inc. ). The physical dimensions of the microphones were 2.3X4.0X5.6 mm. We recorded from two microphones. Normally, one microphone was placed in each ear canal, but, in a few instances, we recorded from two microphones placed in a single canal. The microphone outputs were fil- tered between 1 and 25 kHz, amplified 2000 times with a custom two channel amplifier, and digitized using a 12-bit analog-to-digital converter (Modular Instruments, Inc.) at a sample rate of 50 kHz per channel. Data from each channel were averaged over four samples. Each sample was 10.24 ms in duration, and samples were separated in time by 38 ms; the use of brief time windows effectively excluded any reflections from distant walls or equipment. For each measurement, the level of the loudspeaker output was adjusted auto- matically so that the amplified signal from each microphone was within 12 dB of full scale at the analog-to-digital converter without clipping. Because the stimulus level was adjusted to optimize the recording from each microphone, it was necessary to sample from the two microphones one at a time rather than simultaneously. However, measurements from the two microphones for each sound source location were separated in time by no more than 400 ms. Data were transformed into the frequency domain using a 512 point fast Fourier transform, yielding 257 frequency components spaced 97.7 Hz apart in frequency from 0 to 25 kHz.

D. Signal-to-noise ratio We took several steps to maximize the signal-to-noise

ratio of the ear canal recordings. ( 1 ) We limited the stimulus band to between 3 and 16 kHz so that energy was concentrated in the frequency band of interest. (2) We equalized the output of the loudspeakers by presenting a custom digital output buffer through each loudspeaker. An alternative approach would have been to present the same output buffer through each loudspeaker, then to apply a correction for each loudspeaker to the signal acquired from the microphone. That approach was less satisfactory because the signal-to-noise ratio would have been decreased at frequencies at which the response of a given loudspeaker was depressed. (3) We tilted the spectrum of the stimulus up by 6 dB per octave. This compensated roughly for the fundamental resonance of the ear canal near the low-frequency end of the stimulus band and for the high-frequency rolloff of the miniature microphones. (4) We adjusted the output level of the loudspeaker for each stimulus presentation so that the amplified microphone signal effectively "filled" the 12 bits of the analog-to-digital converter. (5) We averaged the microphone signal over four presentations of the stimulus waveform.

Figure 3 (a) shows an example of a time waveform recorded in an ear canal; the corresponding amplitude spec-



'(a )

I

Time

• --10

• -20

c -30

o

-40

.>

-• -50

i i I I i I• i i i i i i I i

5 10 15 20 25

Frequency (kHz)

i i I i i I i i

(b)! .

.

,=

.

.

,=

.

.

FIG. 3. Time waveform (a) and amplitude spectrum (b) recorded from an ear canal. This signal was recorded from the right ear canal of subject 4 in response to a stimulus presented from right 30 ø, 07.

trum is shown in Fig. 3 (b). In comparison with the stimulus waveform and amplitude spectrum shown in Fig. 2, this recording shows prominent maxima and minima in the amplitude spectrum. These alterations in the amplitude spectrum show the influence of the transfer function from the free-

field source to the position in the ear canal occupied by the microphone and the detailed frequency response of the microphone. The signal-to-noise ratio was about 30 dB on both the high- and the low-frequency ends of the stimulus band, as can be seen by comparing the peak level within the stimulus band to the levels recorded outside the stimulus band.

E. Experimental protocol

In a typical experiment, the subject was seated at the center of the loudspeaker array with his/her interaural axis aligned with the axis of rotation of the array. The subject wore light clothing. An elastic cloth headband held the electromagnetic sensor and supported the microphone leads. The subject was asked to position one of the Knowles microphones in each ear canal as deeply as was comfortable. The position of each microphone was inspected with an otoscope to ensure that the microphone was at least 5 mm deep to the opening of the ear canal. During the recording, the subject was asked to hold his/her head stationary in an upright position and to refrain from moving the jaw or swallowing. A complete series of measurements took about 6 min. The data were rejected if the subject's head moved more than 2 ø in any

direction away from the starting position during that time. The vertical position of the loudspeaker array was con-

trolled manually from outside of the testing chamber. Every other aspect of stimulus presentation and recording was un- der computer control. Stimuli were presented in 10 ø increments of azimuth from - 170 ø to + 180 ø. Because the plane of the loudspeaker array coincided with circles of constant elevation only at 0 ø elevation, it was not convenient to sample in equal increments of elevation. Instead, we stepped the loudspeaker array through 10 ø vertical increments from - 60 ø to + 70 ø and, at each position of the array, we activat-

ed sequentially only the loudspeakers that were necessary to produce a spacing in elevation of 10 ø or less. For example, the loudspeaker at 0 ø azimuth was activated at each position of the loudspeaker array, whereas the loudspeakers at +__ 80 ø and +__ 100 ø were activated only when the array was at - 60 ø, 0 ø, and + 70 ø. A total of 356 loudspeaker locations

were tested in each 6-rain series of measurements.

II. RESULTS

A. Influence of microphone position on measurements of directionality

Because of the pattern of longitudinal standing waves in the ear canal, the recorded spectra varied substantially according to the position of the microphone in the canal. These variations were, essentially, a nuisance variable in this study. The primary interest was the variation in amplitude of single frequency components as a function of sound source location, which we refer to as the "directionality." We determined the extent to which the measured directionality de- pended on the position of the microphone in the ear canal by recording from two microphones placed at two locations ap- proximately 9 mm apart in a single ear canal. Figure 4 shows, for two different frequencies, the levels recorded at each microphone for each of 356 loudspeaker locations. The open symbols represent data at 5.27 kHz, and the crosses represent data at 8.2 kHz. At 5.27 kHz, the absolute levels measured at the distal microphone for most source locations were about 5 dB greater than those measured by the proximal microphone, whereas, at 8.2 kHz, the absolute levels measured by the distal microphone were about 20 dB less than those at the proximal microphone. These differences in absolute level reflect the standing wave pattern present in the ear canal. Despite these substantial differences in the absolute levels, the relative variations in the levels with changing sound source location were nearly the same at the two microphones, especially for sound source locations that produced levels within about 20 to 30 dB of the maximum for each

frequency. Sound sources at some locations behind the head produced lower absolute levels in the ear canal, resulting in reduced signal-to-noise ratio and somewhat less correspondence for those source locations.

Figure 5 (a) and (b) shows the amplitude spectra averaged across all 356 sound source locations for the proximal and distal microphones. Differences in the frequencies of peaks and nulls are readily apparent. Figure 5 (c) shows the difference (in decibels) between the two spectra. This difference spectrum is consistent with that estimated for two microphones separated by 9 mm in a straight tube. Absolute



-10

-20

-3o

-40 -

-50 -

-60 -

-70

X x x• x x

I I

-70 -60 -50

o x 0 0

C• o ø

o 5.27 kHz, r--.992 x 8.20 kHz, r=.g70

I I I

-40 -,30 -20

Distal Mic (dB)

I

-10

FIG. 4. Pressure levels at two frequencies recorded from two microphones positioned in the same ear canal. Two mi-

crophones were fastened 9 mm apart and positioned in the ear canal so that the more distal microphone (i.e., the one furthest from the tympanic membrane) was at least 5 mm deep to the canal entrance. Each symbol represents the levels recorded by the two microphones for a single sound source location. Circles represent the level at the 5.27-kHz frequency component, and crosses represent the level at the 8.20-kHz component.

amplitude differences of as much as 20 dB presumably are due to differences in the positions of the microphones relative to the pattern of standing waves within the ear canal. Figure 5 (d) shows as a function of frequency the standard deviation of the difference distribution computed over the 356 sound source locations. Standard deviations range from less than 1 dB at frequencies below 7.5 kHz to around 3 dB near 16 kHz. This indicates that we can measure direction-

ality, independent of microphone position in the ear canal, with standard deviation of about 1 dB in the lower half of our

stimulus band and up to 3 dB across the entire band. We also explored the range of microphone positions that

would satisfy the criterion of being "within the ear canal" and hence would produce comparable measurements of directionality. In the experiment represented in Fig. 6, we recorded simultaneously from one microphone placed about 10 mm deep to the entrance of the canal and a second microphone placed further distal: ( 1 ) at the canal entrance [Fig. 6 (a) ]; (2) outside of the canal, on the floor of the cavum concha [Fig. 6(b) ]; or (3) entirely free of the external ear, suspended opposite the canal entrance, 10 mm lateral to the tragus [ Fig. 6 (c) ]. The figures show the standard deviations of the difference distributions for the distal compared to the proximal recording and can be compared with the standard deviations shown in Fig. 5 (d), for which the distal microphone was 5 mm deep to the canal entrance. The standard deviations measured when the distal microphone was positioned at the canal entrance [Fig. 6(a) ] are comparable in

magnitude to those measured with both microphones inside the canal [ Fig. 5 (d) ]. This suggests that comparable measurements of directionality could be made for any microphone position at the canal entrance or deeper. Above about 4 kHz for more distal microphone positions [Fig. 6(b) and (c) ], the standard deviations were much larger, nearly 10 dB, demonstrating that measurements of directionality are sensitive to the position of the microphone relative to the external ear at those frequencies. The standard deviations at the lowest frequencies in the stimulus band, however, were uniformly low for all microphone positions, even those outside the external ear. This is consistent with the idea that the

head is the major acoustical element contributing directionality below about 4 kHz.

In summary, measurements of directionality from positions outside of the ear canal can differ substantially from those made inside the canal, but equivalent measures of directionality can be made from any position within the canal. Thus we assume that our measurements of directionality at the middle of the canal are representative of the directionality of sound-pressure levels immediately in front of the tympanic membrane.

B. Directionality of sound-pressure levels

The sound-pressure level measured at any frequency for any sound location is the product of the longitudinal resonance of the ear canal and a directional term. To provide a definition of directionality that was independent of absolute

94 J. Acoust. Soc. Am., Vol. 86, No. 1, July 1989 Middlebrooks ot a/.' Ear directionality 94


0 ' ' "(a) Proxirr;al Mic

•,-12 '

m -10 .> '*" --20

rY -•0

-40

-5O

' (b) Dist•l Mic

I i i , i

20 (c) D•fference

15

10

I I I I

5 ' ' (d) Stc•ndard De•/iation .

- I

5 10 15 20 25

Frequency (kHz)

FIG. 5. Amplitude spectra and amplitude difference spectrum for recordings from two microphones in the same ear canal. Data are from the same recordings as those described for Fig. 4. (a) and (b) show the amplitude spectra recorded by the proximal and distal microphones, respectively, averaged across all 356 sound source locations. These spectra are normalized with respect to the greatest level that was recorded by either microphone, i.e., the level around 3.5-kHz recorded by the proximal microphone. (c) is a difference spectrum showing the average (across 356 source locations) of the level at the proximal microphone minus the level at the distal microphone. (d) shows the standard deviations (across 356 source locations) of the differences between the proximal and distal recordings. In (c) and (d), the extent of the graphs in the frequency domain is limited to the 3- to 16-kHz stimulus band.

sound-pressure level, we normalized the sound-pressure levels at any given frequency relative to the greatest level that was measured at that frequency for any sound source location:

D( az, el, f ) = max,,z,e/L ( az, el, f ,x ) -- L ( az, el, sf,,x ) , ( 1 ) where L (in decibels) is the level measured in the ear canal,

(a).

10 I i i

(b). .

.,

.

.

..

.

,,.

I I I

10

o o 2o

(c).

,=

,=

,=

I I I 5 10 15

Frequency (kHz)

FIG. 6. Standard deviations of the difference distributions for recordings from two microphones in the same ear. In the three recording configura- tions shown in (a)-(c), one microphone was positioned about 10 mm deep to the entrance of the ear canal, and a second microphone was positioned either: at the entrance to the ea• canal (a); outside of the canal, on the floor of the carurn concha (b); or entirely free of the external ear, suspended opposite the canal entrance, 10 mm lateral to the tragus (c). The data plotted are the standard deviations across 356 sound source locations of the

differences between the levels recorded at the proximal and distal microphones. The extent of the graphs in the frequency domain is limited to the 3- to 16-kHz stimulus band.

at position x and at frequency f, for a sound source location given by az (azimuth) and el (elevation); max,,z, et L is the maximum level obtained at that microphone position and frequency for any sound source location; D (in decibels) is the directionality at the frequency for a given sound source location. Values of D can be regarded in the sense of attenu- ation, for which a value of zero represents the greatest level at any frequency and the largest values represent the small- est sound levels. Note that D is independent of microphone position x.

95 J. Acoust. Soc. Am., Vol. 86, No. 1, July 1989 Middlebrooks et al.' Ear directionality 95


,I. kHz

4- 4-

4-

4-

+4-

0.0

6kHz

++

4.

0.0

8kHz 4.+

+;-

0,0

10kHz 4.+

++

+

4-

+

-I.

4.

-I.

+4.

0,0

12kHz + ++

+++ • +

+

+ +

:+ 0,0

14kHz + 4.+

-I. +

-I. .I..i.• +4-

+ - +

4.

4- -I.' +

+

+ + -I.

+'H.

0.0

FIG. 7. Directionality of sound pressures in the ear canal. Recordings were made with one microphone in the right ear canal of subject 1. The spatial coordinate system is the same as that illustrated in Fig. 1. In each of these six plots, the axis labeled 0ø,0 ø is directly in front of the subject, and the axis extended to the left of each figure coincides with the subject's interaural axis on his right side. Lines are drawn to represent the vertical midline plane and the interaural horizontal plane. Data are shown for frequency components at 4, 6, 8, 10, 12, and 14 kHz. In the plot for each frequency, contour lines represent the value of D(az, el, f ) [ Eq. ( 1 ) ], which is the sound pressure normalized relative to maximum level that was measured at that frequency for any sound source location. Contour lines are plotted in 5-dB increments of normalized sound pressure. The 5-dB contours are filled with stippling to indicate the areas from which sound sources produced levels that were within 5 dB of the maximum at each frequency. The small crosses indicate the sound source locations for which measurements were made.

96 J. Acoust. Soc. Am., Vol. 86, No. 1, July 1989 Middlebrooks eta/.: Ear directionality 96


Directionality data at constant frequencies are represented by isolevel contours drawn on the coordinate sphere. The viewpoint for every contour plot is the same as that shown in Fig. 1: Coordinates are shown as if looking in toward the subject from a position 30 ø above and 30 ø to the subject's right. The origin of the coordinate system (the axis labeled 0 ø, 0 ø in Fig. 1 ) is directly in front of the subject, and the interaural axis on the subject's right side is drawn on the left side of the coordinate sphere (the axis labeled right 90 ø, 0 ø in Fig. 1 ). Figure 7 shows directionality plots for one subject's right ear canal at six frequency components within the stimulus band. The small crosses indicate the sound source

locations that were tested. Isolevel contour lines in each plot are drawn in increments of 5 dB. The stippled area in each plot indicates the area from which sound sources produced levels within 5 dB of the maximum at each frequency. The pattern of directionality is qualitatively different at different frequencies. For example, the areas from which sounds produced maximal amplitudes occupied one (e.g., 4, 6, 10, and 14 kHz) or two (8 and 12 kHz) discrete areas. At some frequencies (e.g., 14 kHz), the 5-dB contour crossed the frontal midline, while, at other frequencies ( 10 kHz), it en- circled the interaural axis. When there were two areas from

which the greatest amplitudes were produced, the two areas could be displaced either horizontally (e.g., 12 kHz) or vertically (8 kHz) from each other.

The general patterns of directionality measured for a given ear canal were reproducible between testing sessions. This is demonstrated in Fig. 8(a) and (b), which shows the directionality for 12 kHz measured in one ear canal on two different days. Close inspection shows only minor differences in the two plots. Figure 8 (c) shows a scatter plot of the amplitudes at 12 kHz measured for 356 source locations in the two testing sessions. The standard deviation of the difference distributions between the two sets of measurements was

2.5 dB. We presume that some of the scatter in this plot is due to slight differences in the way that the subject positioned his head in the two sessions, since the standard deviation due to differences in microphone position should have been less than 2 dB at this frequency (Fig. 5).

C. Comparisons between subjects

The directionality of sound-pressure levels measured for all subjects was qualitatively similar to those of subject 1 illustrated in Fig. 7. The similar patterns of directionality, however, occurred at slightly different frequencies for different subjects. Figure 9 shows directionality contour plots at 8 kHz for subject 1 along with data for two other subjects. The pattern for subject 1 [Fig. 9 (c) ] shows two discrete maxima, one above the horizontal plane and one below. The directional patterns most closely resembling the 8-kHz pattern for subject 1 were found at 6.9 kHz for subject 2 [Fig. 9(a) ] and at 8.9 kHz for subject 5 [Fig. 9(d) ]. The patterns at 8 kHz for subject 2 [ Fig. 9 (b) ] and subject 5 [ Fig. 9 (e) ] differ from the patterns for subject 1 in that they each show only a single area from which maximal amplitudes were produced. The patterns at 8 kHz for subjects 2 and 5 more closely resemble the patterns at 10 and 6 kHz, respectively, for subject 1 (Fig. 7).

(a) , +4-

++•,.

0.0

(b) +

_+•M.•--,"• + +++ •--• + +\-

+ ,?+ + 4-

0.0

(c) o

m -10

:•u1-30

-40

I I ß

50 -40 -30 -20 - 0 0 Relative SPL (dB)

Session 1

FIG. 8. Replication of directionality in the same ear canal. Measurements were made in the right ear canal of subject 1 in separate testing sessions. (a) and (b) show directionality for the 12-kHz frequency component measured in testing sessions 1 and 2, respectively. Conventions of these contour plots are described in Fig. 7. (c) shows the sound pressures at 12 kHz measured for 356 sound source locations in the two testing sessions. Each symbol represents the level recorded for one sound source location in the two sessions.

The dashed line indicates equality between the levels measured in the two sessions. The symbols distribute around a line parallel to, but above, the dashed line, indicating that the levels recorded in the two testing sessions differ by a constant.



(a)

.I- -I.

.I-

-I-

.i.'1- -I..I-

6.9kHz

Subject 2

.I-

-I.

-I.

.I-

,I-

.I-

-I.

.I-

-I.

.I-

-I.

0,0

(b)

.i..i.

4.-I,

.i.-I,

8kHz

Subject 2

-I.

-I,

-I.

.H-

0.0

(c) .i..I.

.I-

8kHz 0,0 Subject 1

(d) +

+.1.-I-

.i..I.

8.9kHz 0.0 Subject 5

(e) .I-

.i..i,

+++

8kHz

Subject 5

.I-

.I-

0.0

FIG. 9. Comparisons between subjects. Directionality data are shown for one microphone in the right ear canals of subject 2 [ (a) and (b) ], subject 1 [ (c) ], and subject 5 [ (d) and (e) ]. Conventions of these contour plots are described in Fig. 7. Data at 8 kHz are shown for subject 1. The directionality at 8 kHz for subjects 2 and 5 [(b) and (e)] differed conspicuously in the numbers and locations of sound-pressure maxima and in the general disposition of isolevel contours. In regard to these characteristics, the directionality at 6.9 kHz for subject 2 [ (a) ] and at 8.9 kHz for subject 5 [ (d) ] was more similar to that at 8 kHz for subject 1.

98 J. Acoust. Soc. Am., Vol. 86, No. 1, July 1989 Middlebrooks ot a/.: Ear directionality 98


10.0

5.0

0.0

12kHz

• • I I I

t-

10.0

5.0

0.0

10.0 ..... . 1

5.0 ,.

8kHz

0.0 '

10.0 ....

.

Frequency Retio (Subject 5 / Subject 1)

FIG. 10. Standard deviations of the distribution of differences in levels mea-

sured for subjects 1 and 5 at 356 sound locations at various frequencies. For each panel of this figure, a reference frequency (i.e., 6, 8, 10, or 12 kHz) was held fixed for subject 1 and a comparison frequency was varied systematically in 97.7-Hz steps for subject 5. The standard deviations of the difference distributions are plotted as a function of the comparison frequency, which is expressed as the ratio of the comparison frequency to the reference frequency.

We used standard deviations of difference distributions

to quantify the similarity between directional patterns for pairs of subjects. To compare a subject S with a reference subject R, we used the following procedure: (1) Select a

reference frequency for subject R and a (variable) frequency for subject S; (2) compute the difference in pressure levels measured for the two subjects at each of the 356 loudspeaker locations; (3) compute the standard deviation of the difference distribution; (4) repeat this procedure, systematically varying the frequency for subject S in increments of 97.7 Hz (i.e., the spacing between individual Fourier components); and (5) plot the standard deviations as a function of the variable frequency for subject S divided by the reference frequency for subject R. The minimum value of the standard deviation identifies the directional patterns that are most similar, and hence we can determine the frequency ratio that produces the most similar directional hearing pattern. Fig- ure 10 shows the standard deviations for subject 5 compared to subject 1 for four reference frequencies. For this pair of subjects, the minimum standard deviations all fell at frequency ratios greater than 1.

We noted a general correlation between the frequencies

._.16

E12

E

i i I t i i

ß ,.•-I-." ß Subject 2 / "Subject 4

o Subject 5 [] Subject 6 i • [ • 6 8 10 12 14

Reference Frequency (kHz, Subject 1)

FIG. 11. Frequencies at which directionality data for subjects 2-6 showed the minimum standard deviations with that for subject 1. Measurements were made in the right ear canal of subjects 1-6. Data for subject 1 were tested at fixed reference frequencies of 4 to 15 kHz in steps of 500 Hz. At each reference frequency for subject 1, we computed the standard deviation of the difference distribution as a function of a varied frequency for each of subjects 2-6. This procedure produced plots of standard deviation versus frequency ratio like those shown in Fig. 10. From each plot, we determined the frequency at which the minimum standard deviation was found. We computed that frequency by fitting a second-order polynomial to the lower 1 dB of the correlation plots, and by finding the frequency that corre- sponded to the minimum of the fitted polynomial. The heights of the subjects, in order of increasing height, were: subject 5, 152 cm; subject 4, 157 cm; subject 6, 173 cm; subject 3, 175 cm; subject 1, 183 cm; subject 2, 193 cm. Data for the reference frequency of 14 kHz are shown only for subjects 2 and 6 compared to subject 1 because a clear minimum standard deviation could not be determined within the stimulus bandwidth (i.e., below 16 kHz) using the procedure described above.

99 J. Acoust. Soc. Am., Vol. 86, No. 1, July 1989 Middlebrooks et aL' Ear directionality 99


4kHz ++ +• + /.•,.-." ,-..:'.-.., ...'.•...:.::..'•_ •

+ + •i::.:',•:! ':':':'--',- '-'..'. :'. ::..";:•.• ß •:!8 ::"-': :*:':':':::i. :::. '.'::;•-•. ß + +•;::'•: ::::::::::-:..%'::::::•.:-.::,.

ß ' ,•':':':':':'-",1,'-', "'-'-'-"'""'- ;'"-" • + rzi ,-.-.-.:..-•:::...,,:.: '.<.:.:.:•..:.:.:.:.:.:.:. •, •1 :::::::::'.'.::i:!:,: •::: :::::::::::::::::::::::

++ :::::::::--:::: :. :':•::::.:_ :'e'.':_::.:;•!:i: I•..:::f::.: -. .• .-...-. ......... •':-::::::: •:: ::•-':::::k::::::::;:::::: %:'.'.• P,',

8kHz + +4-

+ + 4-

4' 4' 4' 4- 4-

4-

4-

-I-.N.

0,0

6kHz

+4-

4-

4-

4-

10kHz + 4-4-

+•+•"• + 4- 4-

ß I. 4-

4- '•,•,.-I- 4- 4- 4- 4-

4-

4- 4- 4-

4-

.+ 0.0

4- 4-

4- 4-

4-

4-

4-

4- 4- ß ,,. 4- 4-

4-4-

0,0

,4-

12kHz +

++

4- 4-

4-

.I-4- 4-

4- 4-

4-

4- 4-

4- 4-

4-

+ 4-

4-

-I-.N.

0.0

14kHz

+'P"%,,.,• 2o + + +

4- 4-4- 4-

4- 4- 4-

4-

4- 4- c)

4- 4-

4-

_•+ ++

4- .I-'v 471;. -+4- + 0,0

FIG. 12. Directionality of sound pressures in the right ear canal averaged across subjects. At each nominal frequency (i.e., 4, 6, 8, 10, 12, and 14 kHz), we averaged data across subjects 1 to 6 using the frequencies plotted in Fig. 13, which were selected to minimize the standard deviations between subjects; the plot for 14 kHz represents the average of data only from subjects 1, 2, and 6. The nominal frequencies are the frequencies at which the data from subject 1 were taken. Other details of these contour plots are described in Fig. 7.

1 O0 J. Acoust. Soc. Am., Vol. 86, No. 1, July 1989 Middlebrooks eta/.' Ear directionality 100


at which particular patterns were observed and the physical sizes of the subjects. Specifically, smaller subjects tended to exhibit a given pattern of directionality at higher frequencies than larger subjects. Figure 11 represents comparisons of five subjects with subject 1. The frequency at which the minimum standard deviation was observed is plotted for each subject against a range of reference frequencies; the dashed line has unity slope and indicates perfect correspondence between the varying and the reference frequencies. The heights of subjects provide a simple measure of their overall physical sizes. Subject 1, whose data form the reference in Fig. 11, is 183 cm tall. Subjects 3 and 6 are somewhat shorter (175 and 173 cm tall, respectively), and their data fall at frequencies equal to or slightly higher than those for subject 1. The data for our two shortest subjects (subjects 4 and 5, 157 and 152 cm tall, respectively) all fall at the highest frequencies. Conversely, the data for the tallest subject (subject 2, 193 cm tall) all lie at frequencies below those for the reference. If a single scaling factor accounted for all the differences in the directional data at all frequencies, then the data in Fig. 11 for each subject would fall on a straight line and the slope of that line would equal the scaling factor. The fact that the data scatter about a straight line indicates that no single scaling factor accounts for all the differences between subjects. This suggests that the many physical features of the head and ears, differing in magnitude among subjects, make different acoustical contributions at different frequencies.

We used this frequency scaling procedure to derive an average directionality plot that is representative of all our subjects. We combined directionality data over different subjects from sets of frequencies that minimized the standard deviations among subjects. The average directionality plots at six different frequencies are shown in Fig. 12. The nominal frequency for each plot is the frequency of the data set that was used from subject 1. Compared to the plots shown for a single subject (subject 1 ) in Fig. 7, the average plots are remarkably similar. Naturally, the averaged data are somewhat smoother, but the similarity with respect to the number and location of sound-pressure maxima and the general disposition of isolevel contours is striking. Differ- ences in detail are most evident in areas where absolute pressure levels were low and the signal-to-noise ratio was corre- spondingly reduced.

D. Interaural level differences

We computed interaural level differences (ILDs) by comparing the normalized sound levels measured in the two ear canals. Because the amplitudes in each canal were known only relative to other amplitudes measured in that canal, it was impossible to determine absolute interaural differences. Therefore, we made the simplifying assumption that the average of all ILDs in the frontal half of the vertical midline at any given frequency was equal to zero, and we normalized the ILD values accordingly. Errors introduced by this assumption should amount only to constant shifts in ILD values and should not influence the general spatial distribution of ILDs. We expressed ILDs as the sound level (in decibels) in the right ear canal minus that in the left.

Figure 13 shows in contour plots the spatial patterns of

ILDs measured at six frequencies for subject 1. The contour lines indicate ILDs in increments of 5 dB, and the stippling indicates the stimulus locations that produced maximal ILDs at each frequency. In general, ILDs increased roughly monotonically with increasing azimuth, and maximum ILDs were produced by sound sources located near the interaural axis (i.e., ñ 90 ø azimuth). In the spatial coordinate system that was used, lines of constant azimuth lie parallel to the vertical midline. For frequencies below about 8 kHz, iso- ILD contours roughly paralleled the midline throughout most of the frontal hemifield and, thus, were relatively constant with changing elevation at a fixed azimuth. In contrast, the iso-ILD contours for frequencies of 8 kHz and higher cut across the lines of constant azimuth, indicating that ILD varied both with azimuth and with elevation. The variation

in ILD with elevation was especially pronounced at frequencies for which multiple discrete maxima were observed in monaural directionality plots, such as the plots for 8 and 12 kHz shown in Fig. 7.

For several of the frequencies represented in Fig. 13, the 0-dB ILD contours deviated substantially from the vertical midline, suggesting that the directionality measured in the two ear canals is asymmetrical. We explored this issue further by making repeated series of measurements of ILD for stimuli located on the vertical midline. In each block of

measurements, we tested elevations from -- 40 ø to -3- 90 ø in 10 ø steps, repeating each series of elevations for a total of ten measurements at each elevation; an entire block of ten measurements at each of 12 elevations took about 2 min. The

means and standard deviations of these measurements for

one subject are shown in Fig. 14. In some cases, the largest ILDs also had the largest standard deviations, which, to some extent, reflects the fact that the amplitude in one ear was of low level and was contaminated with noise. However, at several elevations, the ILD in the vertical midline plane was as great as 10 dB and was more than two standard deviations away from zero. The pattern of ILDs as a function of elevation for midline sound sources varied substantially among subjects. However, nonzero ILDs roughly of the magnitude illustrated were observed for all subjects.

One possible artifact that might have produced nonzero ILDs for midline stimulus locations would be a slight tilt of the subject's head away from a vertical position. To test this possibility, we measured the ILDs produced by midline sound sources with the subject's head held upright and with the head inclined 15 ø to the right and 15 ø to the left. These data are plotted in Fig. 15. There are slight changes in the overall slope of the ILD versus elevation curves that can be attributed to changes in the inclination of the head. Never- theless, there are substantial peaks in the curves that were largely invariant across these 15 ø tips of the head. Thus these departures of measured ILDs away from zero cannot be due entirely to the position of the subject's head and apparently reflect actual asymmetries between the two ears.

III. COMPARISON TO PREVIOUS STUDIES

A central premise of the current study was that measurements of directional sensitivity are independent of the position of the recording microphone within the ear canal.



0.0

4 ++ 4.

4- .!-

4.

0.0

8kHz +4-

+• •+

4. 4' 4.

+4-

0,0

10kHz +4.

+4.

0,0

12kHz + ++

0.0

14kJ-Iz + 4.-I.

.I- +

+ + +

+ +

+

+ + + +

+ +

+ +•+•,• t'.1 , ++

+ +-I,i. +

0.0

FIG. 13. Directionality of interaural level differences (ILDs). Data were collected from subject 1 with one microphone in each ear canal. The ILDs were normalized at each frequency so that the average of the ILDs along the frontal midline was equal to zero. The contours are labeled in decibels as the level in the right canal minus the level in the left canal. The stippling indicates the areas from which sound sources produced maximum ILDs. Other details of the contour plots are as described in Fig. 7.

102 J. Acoust. Soc. Am., Vol. 86, No. 1, July 1989 Middlebrooks ot a/.: Ear directionality 102


i i ! i i i i i ! i i i

1 • 14kHz i .

i i i i i i i i i ! i i i i i

• -lo L• I I I I I I I I I I I I I I I

6kHz

i _ •__•.___ i i i ! i

i i i I

•kl-lz

-10-• . , • , -60 -30

i

i I i i i I i i I i i

0 30 60 90

Elev=tion (deg)

FIG. 14. Interaural level differences for sound sources in the vertical mid-

line plane. The ILDs were normalized as described for Fig. 13. Data were collected from subject 2 with one microphone in each ear canal. The values plotted are the averages and standard deviations of ten sound presentations at each location.

This issue has been addressed in several previous studies. Wiener and Ross (1946) plotted the ratio of the sound pressures measured at the ear canal entrance to that measured at

the midpoint of the canal or the tympanic membrane for sound sources at azimuths of 0 ø, 45 ø, and 90 ø. At frequencies up to 8 kHz, the ratios were independent of source location to within a few decibels. Thus, at any given frequency, a change in recording position resulted only in a change in the absolute level that was recorded, not a change in the directional dependence. Shaw and Teranishi (1968) showed that blocking the ear canal had little effect on the sensitivity to sound source azimuth measured at the ear canal entrance up to at least 12 kHz. This indicates that the longitudinal resonance of the ear canal contributes little to the directional

dependence. Mehrgardt and Mellert (1977) measured the transfer functions from various points in the ear canal to the

tympanic membrane for ten azimuths. Up to 15 kHz, these measurements varied less than 5 dB with source location, again indicating that the longitudinal resonance of the ear canal can be largely separated from the directionality. In the current study, we found that measurements of directionality were constant to within standard deviations of 3 dB for any microphone position that we tested in the ear canal; the maximum standard deviation was as small as 1 dB if we excluded

source locations behind the head for which the signal-to- noise ratio was low. It should be noted that in the current

study and in most previous studies, level measurements were made within narrow frequency bands, using either tonal stimuli or individual Fourier components of broadband stimuli./These narrow-band measurements would have

tended to accentuate any local variability in spectra that were introduced by changing experimental conditions. One would expect that levels measured within the tuning curve of an auditory-nerve fiber or within a psychophysical critical band would show substantial smoothing in the frequency domain. Thus the changes in levels within narrow bands resulting from changes in microphone position probably represent worst case estimates. Given the limited influence of recording position anywhere in the ear canal on measured directionality, we infer that recordings of directionality made within the canal are representative of the intensive cues available to the auditory system for sound localization.

There have been several previous studies of the directionality of sound pressures in the human ear canal (e.g., Wiener and Ross, 1946; Robinson and Whittle, 1960; Schirmer, 1963; Shaw, 1966; Blauert• 1969/70; Searle et al., 1975; Mehrgardt and Mellert, 1977). In those studies, recordings were made at the entrance of the canal, at various points along the length of the canal, or near the tympanic membrane. In each of those studies, the sound source was restricted to the horizontal plane, the vertical midline plane, or the plane defined by the vertical axis and the interaural axis. Our data from those three planes are largely consistent with published directionality data. In Fig. 16, we have re- plotted the appropriate data from our study along with data from two published reports (Shaw, 1974, Fig. 11; Mehrgardt and Mellert, 1977, Fig. 18). The directionality plots presented by Shaw (1974) were derived from the data of 12 previous studies by several groups of investigators representing results from 100 subjects. The data presented by Mehrgardt and Mellert (1977) were averaged across several measurements in each of 20 subjects. As in the current study, Mehr- gardt and Mellert scaled their data in the frequency domain before averaging. Their scaling procedure differed from ours, however, in that for each sound source location, they selected a single scalar to apply at all frequencies. In contrast, we selected for each reference frequency a single scalar that was applied to data for all sound source locations.

Figure 16 shows data for sound sources in the horizontal plane. In each panel, the upper curve represents data from Shaw (1974), and the middle curve represents data from Mehrgardt and Mellert (1977). The lower curve represents the mean and standard deviation from our six subjects. Prior to averaging, we scaled our data in the frequency domain using the procedure described in Sec. II. The data are in good



20-

10-

0-

-10-

20-

10-

0-

-10-

-20-

20

lO

-1

-(a)

8kHz ,

! ! ! ! ! ! ! ! ! ! ! ! !

! ! ! i ! ! ! ! w ! i ! ! !

(b)

.

.

.

. .

. .

. .

. .

. .

.

.

I i i I i i i i I i i i i i

(c)

!

(d)

12kHz ! i i ! ! ! ! ! ! ! [ ! !

I I I ! I ' ' ' ' ' ' I I I I

(e)

(f)

-so -3o o 3o so -so -3o o 3o so

Elevation (deg)

FIG. 15. Influence of head position on ILDs produced by midline sound sources. Data were collected from subject 2 with one microphone in each ear canal. The values plotted are the averages and standard deviations often sound presentations at each location. ILDs at 8 and 12 kHz are shown in the left and right columns, respectively. Initially, the data were collected with the subject's head in an upright position (middle row). Then, the head was tipped by 15 ø to the subject's right (top row) or left (bottom row); the head orientation was monitored electromagnetically.

agreement, particularly at frequencies up to 8 kHz. At 10 kHz and higher, the total variation in level for our data is considerably greater than in the previous reports. Part of that might be due to the effects of averaging across subjects, given that our averaging procedure will minimize the vari- ance between subjects. Another source of differences between studies may be attributed to differences in the align- ment of the subject's head relative to the horizontal plane. One can see from Fig. 7 that, particularly at high frequencies, the profile of amplitude versus azimuth depends strongly on the elevation that one chooses to define the horizontal plane.

Several aspects of the directionality of the ear can be related to the modal characteristics of the concha, as de-

scribed by Shaw (Shaw and Teranishi, 1968; Shaw, 1975; Shaw, 1980). Specifically, responses near 8 kHz are domi- nated by the second and third concha modes at which there is a horizontally oriented pressure node near the crus helias. This node could account for the vertical bipolar sensitivity that we have measured around 8 kHz (Figs. 7 and 12). Simi- larly, the horizontal bipolar sensitivity that we have measured around 12 kHz can be related to the vertical nodal

surfaces found at the fourth concha mode. It is noteworthy that Shaw and Teranishi (1968) found that the frequencies



0

-10

-20

r- ::3 o [(d) C•k• ' ' ' ' ' ' ' ' [ ß '1 z

- 180 -90 0 90 180 - 180

(e) 1

, I , • I , ,

-90 0 90

Azimuth (deg) 180 - 180 -90 0 90 180

FIG. 16. Sound pressure as a function of sound source azimuth compared to previous reports. In each panel, the upper trace is redrawn from Shaw ( 1974, Fig. 11 ), and the middle trace is redrawn from Mehrgardt and Mellert ( 1977, Fig. 18). The lower trace shows the means and standard deviations for our six subjects (only subjects 1, 2, and 6 at 14 kHz). The data from our study were scaled in frequency according to the procedure described in the text.

of individual modes could not be matched between individ-

ual subjects by the application of a single scaling factor. This is consistent with our observation that no single frequency scalar could account for differences in the directionality among individual subjects.

The presence of nonzero ILDs produced by sound sources in the vertical midline plane has been reported pre- viously by Searle et al. (1975). Generally, the magnitudes of deviations away from zero that we observed are comparable to the previous results, although the profiles of ILD as a function of sound source elevation differ in detail. The latter

differences are to be expected, since prominent differences between subjects were encountered both in the study by Searle et al. and in the current study. The standard devia-

tions of amplitudes are considerably smaller in our • data than in the previous study. Presumably, this occurred because our ten series of measurements were made in a single 2-min block, whereas, in the previous study, data were averaged over multiple test sessions.

IV. GENERAL DISCUSSION: ACOUSTIC CUES FOR SOUND LOCALIZATION

Measurements of the directionality of sound levels in the ear canal provide some insights into the acoustical cues that are available for sound localization. The current results

do not bear on the issue of temporal cues for localization, which certainly are important at frequencies below 1.5 kHz (i.e., Rayleigh, 1907; Mills, 1958) and, in the form of inter-



aural envelope cues, might also contribute spatial information at high frequencies (Henning, 1974; McFadden and Pa- sanen, 1976; Nuetzel and Hafter, 1976). However, we are able to evaluate the spatial information that would be available from characteristics of spectral shape and interaural level differences (ILDs). The amplitude spectrum produced at the tympanic membrane by a broadband sound source in a free field contains the amplitude spectrum of the sound source shaped by the longitudinal resonance of the ear canal and the •ar's directionality. In presenting our directionality data, we effectively eliminated the influence of the source spectrum and the canal resonances by normalizing sound levels with respect to the maximum level at each frequency. The question remains whether the nervous system, in inter- preting the sound levels at the tympanic membrane, is able to make valid estimates of the source spectrum and to compen- sate for ear canal resonance in order to gain access to spatial information available from characteristics of spectral shape.

A body of behavioral evidence suggests that human subjects do utilize spectral shape information in localizing a sound source. For example, spectral cues to sound location have been invoked to account for the accuracy of localization by monaural listeners (Harris and Sergeant, 1971; But- ler and Flannery, 1980; Musicant and Butler, 1984; Oldfield and Parker, 1986). Moreover, distortions of the shape of the external ear, which modify its transfer function, result in substantial deterioration in sound localization accuracy (Gardner and Gardner, 1973; Oldfield and Parker, 1984; Humanski and Butler, 1988). Several studies have shown that alterations of the amplitude spectra of broadband stimuli, such as the insertion of bandpass or bandstop filters, can result in characteristic errors in localization by normal binaural listeners (e.g., Blauert, 1969/70; Hebrank and Wright, 1974; Butler and Helwig, 1983). This suggests that, in the absence of a priori knowledge of the source spectrum, a listener assumes that the source spectrum of a natural sound will be relatively flat. Studies of localization of narrow-band sounds are most readily comparable to our measurements of directionality within single-frequency components. Blauert (1969/70) presented narrow-band sounds in the vertical midline plane and asked subjects to report whether the apparent sound source was in front, above, or behind. The reported location was influenced more strongly by the center frequency of the stimulus than by the actual location of the source. The reported location for a given frequency tended to correspond to the location from which a sound of that frequency produced its greatest level in the ear canal. Butler and colleagues have reported similar observations for stimuli presented on the vertical midline (Butler and Helwig, 1983) or in the horizontal plane (Butler and Flannery, 1980; Musicant and Butler, 1984). Butler refers to a band of frequency that is present at its greatest level when presented from a particular source location as a "covert peak." A possible explanation of the narrow-band localization results is that the nervous system interprets the spectrum of a narrow- band sound source as a broadband spectrum containing a covert peak and associates that stimulus with a location from which a broadband source would produce a corresponding covert peak.

We find that at any given frequency there are one or sometimes two areas of space from which a sound source produces a maximum sound pressure in the ear canal. The locations of these areas vary with frequency in both azimuth and elevation. In the studies of narrow-band localization

cited above, subjects were required to constrain their reports of apparent sound locations to the horizontal or the vertical midline plane. However, our acoustical data predict that the spectral peak mimicked by a narrow-band stimulus would signal both azimuth and elevation components of a source location. Indeed, our preliminary behavioral observations indicate that the apparent location of a narrow-band stimulus can deviate from its actual location in both azimuth and

elevation. Thus acoustical and behavioral results suggest that characteristics of spectral shape provide cues both for azimuth and for elevation and that, within any narrow band of frequency, these cues are not mutually independent.

The contribution of spectral shape cues to localization has been emphasized in studies of vertical localization in the midline plane (Gardner, 1973; Hebrank and Wright, 1974; Butler and Belendiuk, 1977). Behavioral studies have shown that stimuli can be localized accurately in the vertical midline only if they contain frequencies above about 6 kHz (Roffler and Butler, 1968). Consistent with this finding, our acoustical data show that sound levels below about 8 kHz are

relatively constant with changing elevation along the vertical midline, whereas isolevel contours for higher frequencies tend to cross the vertical midline. Moreover, we would ex-

pect that vertical localization of sounds below 8 kHz would be poor even for locations off the midline, given the roughly vertical orientation ofisolevel contours at lower frequencies.

Interaural level differences (ILDs) traditionally have been regarded as cues only for the azimuth of a sound source. Indeed, our measurements indicate that ILDs increase roughly monotonically with increasing azimuth and, at frequencies below 8 kHz, ILDs are roughly constant with changing elevation. However, at frequencies of about 8 kHz and higher, we also see considerable variation in ILD as a function of elevation. This indicates that ILDs in high frequencies could also provide cues to sound source elevation. Moreover, the departure of iso-ILD contours from vertical at high frequencies indicates that the ILD cues within any narrow band of frequencies will tend to confound the elevation of a sound source with its azimuth. Thus, like spectral cues, ILDs at high frequencies cannot be regarded as cues specific for azimuth or specific for elevation.

Our measurements of ILDs produced by sound sources on the vertical midline confirm the earlier reports by Searle et al. (1975). There is some controversy among previous behavioral studies as to whether ILDs contribute substan-

tially to the perceived elevation of a midline sound source (Searle et al., 1975; Morimoto and Nomachi, 1982). The current results contribute no insight into the salience of midline ILD cues relative to other cues for elevation. However,

we can say that the ILDs that we observed would be of a magnitude that should be readily detectable by a listener (Mills, 1960).

Comparisons of ear canal directionality between subjects indicate that the patterns of directionality are remark-



ably similar among subjects, especially when scaled in frequency according to the physical size of the subject. The variation in directionality among subjects, however, suggests that subjects who differ in physical size would perform differently in tests of localization of narrow-band signals. For example, the directionality plots in Fig. 9 predict that, when presented with a narrow-band noise centered on 8 kHz, subject 2 would localize the source to the side and down, whereas subject 5 would localize the same source to the side and up. The similarities and differences among our subjects are consistent with what data are available from experiments in which subjects attempted to "localize" prere- corded signals (Butler and Belendiuk, 1977; Morimoto and Ando, 1980). In those experiments, subjects performed best when listening to the signals recorded from their own ear canals, although they could also achieve some success at localizing recordings made from the ear canals of different subjects. The partial success in "localization" using recordings from another's ear canals is consistent with the general similarity in directionality among subjects. Conversely, the fact that most subjects localized best using recordings from their own ears testifies to differences in directionality between subjects of different physical size.

It is interesting to compare the directional sensitivity of the human ear canal with that measured for the cat, given the large body of neurophysiological data available for the later species. In man, the areas from which sources produce maximum sound-pressure levels range in location throughout most of the frontal half of space, whereas maximum ILDs are produced by sound sources located around 90 ø azimuth. Nearly the opposite situation is seen in the cat. In the cat, the area from which maximum sound levels are produced is relatively invariant with frequency, whereas the sources produc- ing maximum ILDs vary in location from 90 ø to as far frontal as about 20 ø azimuth (Middlebrooks and Pettigrew, 1981; Middlebrooks and Knudsen, 1987). Based on these differences in acoustic cues, one would expect that the cat would obtain more useful spatial information from interaural differences than from characteristics of monaural spectral shape. Indeed, a neural map of auditory space has been described in the cat's superior colliculus (Middlebrooks and Knudsen, 1984), and a simple model based on sensitivity to ILDs can account for much of the spatial sensitivity of neurons in that structure (Middlebrooks, 1987). In contrast, the behavioral data in man emphasize the relative salience of spectral cues that are present at each ear and suggest that at least some spatial information is extracted prior to or inde- pendently of an interaura! comparison.

ACKNOWLEDGMENTS

We thank Bradley Bessant for preparing the illustra- tions. The miniature microphones were supplied by Knowles Electronics, Inc. This research was supported by UF Award DSR-D-54, NIH-NINCDS Grant R29 NS25022, and ONR Grant N00014-89-J- 1427.

Blauert, J. (1969/70). "Sound localization in the median plane," Acoustica 22, 205-213.

Blauert, J. (1983). Spatial Hearing (MIT, Cambridge, MA). Butler, R. A., and Helwig, C. C. (1983). "The spatial attributes of stimulus

frequency in the median sagittal plane and their role in sound localization," Am. J. Otolaryngol. 4, 165-173.

Butler, R. A., and Belendiuk, K. (1977). "Spectral cues utilized in the localization of sound in the median sagittal plane," J. Acoust. Soc. Am. 61, 1264-1269.

Butler, R. A., and Flannery, R. (1980). "The spatial attributes of stimulus frequency and their role in monaural localization of sound in the horizontal plane," Percept. Psychophys. 28, 449-457.

Gardner, M. B. (1973). "Some monaural and binaural facets of median plane localization," J. Acoust. Soc. Am. 54, 1489-1495.

Gardner, M. B., and Gardner, R. S. (1973). "Problem of localization in the median plane: effect of pinnae cavity occlusion," J. Acoust. Soc. Am. 53, 400-408.

Harris, J. D., and Sergeant, R. L. (1971). "Monaural/binaural minimum audible angles for a moving sound source," J. Speech Hear. Res. 14, 618-629.

Hebrank, J., and Wright, D. (1974). "Spectral cues used in the localization of sound sources on the median plane," J. Acoust. Soc. Am. 56, 1829-1834.

Henning, G. B. (1974). "Detectability ofinteraural delay in high-frequency complex waveforms," J. Acoust. Soc. Am. 55, 84-90.

Henning, G. B. (1980). "Some observations on the lateralization of complex waveforms," J. Acoust. Soc. Am. 68, 446-454.

Humanski, R. A., and Butler, R. A. (1988). "The contribution of the near and far ear toward localization of sound in the sagittal plane," J. Acoust. Soc. Am. 83, 2300-2310.

Knudsen, E. I. (1982). "Auditory and visual maps of space in the optic tectum of the owl," J. Neurosci. 2, 1177-1194.

Kuhn, G. F. (1987). "Physical acoustics and measurements pertaining to directional hearing," in Directional Hearing, edited by W. Yost and G. Gourevitch (Academic, New York), pp. 3-25.

McFadden, D., and Pasanen, E.G. (1976). "Lateralization at high fre- quenc. ies based on interaural time differences," J. Acoust. Soc. Am. 59, 634-639.

Mehrgardt, S., and Mellert, V. (1977). "Transformation characteristics of the external human ear," J. Acoust. Soc. Am. 61, 1567-1576.

Middlebrooks, J. C. (1987). "Binaural mechanisms of spatial tuning in the cat's superior colliculus," J. Neurophysiol. 57, 688-701.

Middlebrooks, J. C., and Knudsen, E. I. (1984). "A neural code for auditory space in the cat's superior colliculus," J. Neurosci. 4, 2621-2634.

Middlebrooks, J. C., and Knudsen, E. I. (1987). "Changes in external ear position modify the spatial tuning of auditory units in the cat's superior colliculus," J. Neurophysiol. 57, 672-687.

Middlebrooks, J. C., and Pettigrew, J. D. (1981). "Functional classes of neurons in primary auditory cortex of the cat distinguished by sensitivity to sound location," J. Neurosci. 1, 107-120.

Mills, A. W. (1958). "On the minimum audible angle," J. Acoust. Soc. Am. 30, 237-246.

Mills, A. W. (1960). "Lateralization of high-frequency tones," J. Acoust. Soc. Am. 32, 132-134.

Morimoto, M., and Ando, Y. (1980). "On the simulation of sound localization," J. Acoust. Soc. Jpn. (E) 1, 167-174.

Morimoto, M., and Nomachi, K. (1982). "Binaural disparity cues in median-plane localization," J. Acoust. Soc. Jpn. (E) 3, 99-103.

Morse, P.M., and Ingard, K. U. (1968). Theoretical Acoustics (McGraw- Hill, New York).

Musicant, A.D., and Butler, R. A. (1984). "The psychophysical basis of monaural localization," Hear. Res. 14, 185-190.

Nuetzel, J. M., and Hafter, E. R. (1976). "Lateralization of complex waveforms: Effects of fine structure, amplitude and duration," J. Acoust. Soc. Am. 60, 1346-1399.

Oldfield, S. R., and Parker, P. A. (1984). "Acuity of sound localization: a topography of auditory space. II. Pinna cues absent," Perception 13, 601-617.

Oldfield, S. R., and Parker, P. A. (1986). "Acuity of sound localization: a topography of auditory space. III. Monaural hearing conditions," Per- ception 15, 67-81.

Rabbitt, R. D., and Holmes, M. H. (1988). "Three-dimensional acoustic waves in the ear canal and their interaction with the tympanic membrane," J. Acoust. Soc. Am. 83, 1064-1080.

Rayleigh, Lord (J. W. Strutt) (1907). "On our perception of sound direction," Philos. Mag. 13, 214-232.

Robinson, D. W., and Whittle, L. S. (1960). "The loudness of directional

107 J. Acoust. Soc. Am., Vol. 86, No. 1, July 1989 Middlebrooks et aL: Ear directionality 107


sound fields," Acoustica 10, 74-80. Roffier, S. K., and Butler, R. A. (1968). "Factors that influence the local-

ization of sound in the vertical plane," J. Acoust. Soc. Am. 43, 1255-1259.

Schirmer, W. (1963). "Die Richtcharakteristik des Ohres," Hochfre- quenztech. u. Electroakust. 72, 39--48.

Searle, C. L., Braida, L. D., Cuddy, D. R., and Davis, M. F. (1975). "Bin- aural pinna disparity: Another auditory localization cue," J. Acoust. Soc. Am. 57, 448--455.

Shaw, E. A. G. (1966). "Earcanal pressure generated by a free sound field," J. Acoust. Soc. Am. 39, 465--470.

Shaw, E. A. G. (1974). "Transformation of sound pressure level from the free field to the eardrum in the horizontal plane," J. Acoust. Soc. Am. 56, 1848-1861.

Shaw, E. A. G. (1975). "The external ear: New knowledge," in Earmolds

and Associated Problems: Proc. Seventh Danavox Symposium, edited by S. C. Dalsgaard (Scandinavian Audiology, Suppl. 5), pp. 24-50.

Shaw, E. A. G. (1980). "The acoustics of the external ear," in Acoustical Factors Affecting Hearing Aid Performance, edited by G. A. Studebaker and I. Hochberg (University Park, Baltimore, MD), pp. 109-125.

Shaw E. A. G., and Teranishi, R. (1968). "Sound pressure generated in an external-ear replica and real human ears by a nearby point source," J. Acoust. Soc. Am. 44, 240-249.

Wiener, F. M., and Ross, D. A. (1946). "The pressure distribution in the auditory canal in a progressive sound field," J. Acoust. Soc. Am. 18, 401-408.

Wightman, F. L., and Kistler, D. J. (1989). "Headphone simulation of free- field listening. II: Psychophysical validation," J. Acoust. Soc. Am. 85, 868-878.

108 J. Acoust. Soc. Am., Vol. 86, No. 1, July 1989 Middlebrooks et al.: Ear directionality 108


Directional Sensitivity of Sound Pressure Levels in the Human Ear Canal

Documents

suspended opposite thecanalentrance

pole coordinate system

sound level meter

356sound source locations

vertical midline plane

pressure levels measured

khz stimulus band

coordinate sphere