ESTIMATING PRESSURE AT EARDRUM WITH ... - Semantic Scholar
2009 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics
October 18-21, 2009, New Paltz, NY
ESTIMATING PRESSURE AT EARDRUM WITH PRESSURE-VELOCITY MEASUREMENT FROM EAR CANAL ENTRANCE Marko Hiipakka, Matti Karjalainen, and Ville Pulkki Helsinki University of Technology, Department of Signal Processing and Acoustics, Espoo, Finland Source
ABSTRACT It is important to know the sound pressure signal at the eardrum in headphone reproduction and in audiological applications. Unfortunately, it is difficult to conduct direct measurement safely. Estimating pressure signals at the eardrum based on measurements done elsewhere in the ear canal is sensitive to positioning errors. This is particularly the case at the canal entrance. In addition, not knowing the acoustic properties of the ear canal makes estimation difficult. This study shows that when both sound pressure and velocity are measured at the canal entrance using a pressure-velocity probe, the pressure signal at the eardrum can be estimated with much higher accuracy than from the pressure-only measurement. The method is demonstrated and validated by using physical simulators and computational modeling.
Rs
Pe
Zw,γ
Eardrum Qd Pd Zd
Ps
L Figure 1: Circuit model for the open ear canal.
pressure-and-velocity measurements are used to calculate the pressure at the eardrum of ear canal simulators accurately. Measurements are realized with the Microflown PU Match Probe [2], a custom-made ear canal simulator and a dummy head. 2. THEORY AND COMPUTATIONAL MODELING Acoustically the external ear contains the head and torso, the pinna and concha, and the ear canal up to the eardrum (tympanic membrane) [3]. The ear canal transmits sound signals and creates resonances due to standing waves between the entrance and the eardrum. In normal listening conditions, the entrance is open and the eardrum appears as a termination of relatively high acoustic impedance. Sound transmission through the canal is independent of sound arrival angle to the entrance point. Fig. 1 shows an equivalent circuit for sound transmission where an external sound source is feeding a signal through radiation impedance1 to the canal, and this acoustic transmission line is terminated by the eardrum impedance Zd . In an idealized model the radiation impedance Zs consists of a parallel connection of inductance Ls and resistance Rs [4] and the ear canal can be approximated as a cylindrical hard-walled tube with wave impedance Zw and negligible losses. The eardrum can be approximated by a relatively simple impedance model, such as a high resistance, although in reality it may be a quite complex function of frequency. An acoustic transmission line with a constant cross-sectional area S has a wave impedance Zw = ρc/S, where ρ is the density of air and c is the speed of sound. For a tube section, represented as a two-port, the signal variables P for pressure and Q for volume velocity as Laplace or Fourier transforms at ports 1 and 2 follow the equation » – » –» – cosh(γL) Zw sinh(γL) P1 P2 = 1 (1) sinh(γL) cosh(γL) Q Q1 2 Zw
Index Terms— eardrum, pressure, velocity, ear canal, Microflown, HRTF, audiology. 1. INTRODUCTION There are many applications where it is important to know the sound pressure at the eardrum. Unfortunately, direct measurement is possible only in special studies where extreme care can be taken not to harm the fragile membrane or other parts of the ear canal. For example, in audiological applications and particularly in binaural sound reproduction such direct measurements are out of the question in most cases. Therefore, sound signals have to be measured at other points of the ear canal, or often at the open or blocked canal entrance. Then the problem is to use such measurements to estimate sound pressure at the eardrum, because it is the basic reference point for head-related transfer functions, binaural headphone reproduction, and binaural recordings. There have been several efforts to estimate the pressure at the eardrum. Estimations are normally based on pressure measurements along the ear canal and on computational models. The standing waves inside the ear canal result in different pressure amplitudes at different parts of the canal, and the position of the microphone affects the sound pressure measurement values. In 1998 Hudde et al [1] presented the reflectance-phase method for estimating the pressure at the eardrum. It was shown that a good estimation can be achieved by using the minima of the pressure at the ear-canal entrance. When using Hudde’s approach, the physical parameters of individual ear canals as well as the volume velocity have to be determined. A micromachined particle velocity transducer that makes volume velocity measurements possible has recently become available. We are suggesting the use of this device for ear canal measurements as a new approach to estimate the eardrum pressure. In this article a computational model of the unblocked ear canal is first presented in order to shed some light on the physics of the ear canal and ear canal simulators. The model together with
978-1-4244-3679-8/09/$25.00 ©2009 IEEE
Ls
Ear canal Qe
where L is the length of tube section and γ is the wave propagation constant, γ = α + jβ with α being attenuation coefficient and β phase coefficient β = ω/c = 2πf /c. In Fig. 1 the port variables at the canal entrance are P1 = Pe and Q1 = Qe and at the eardrum P2 = Pd and Q2 = Qd . These equations together with Zs and Zd can be used to compute the transfer function from external source pressure Ps to the pressure Pe at the entrance as Hs,e = Pe /Ps and to the pressure Pd at the eardrum as Hs,d = Pd /Ps . Fig. 2 1 Due
289
to reciprocity principle [4].
2009 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics
P P ZQ
Figure 4: The Microflown PU Match Probe consists of a pressure microphone (above) and a particle velocity transducer (below). The diameter of the pressure microphone is 2.5 mm and the total diameter of the tip of the PU probe is 3.5 mm. [2]
Figure 2: Modeled pressure responses at the eardrum (Pd ) and at the canal entrance (Pe ). Modeled volume velocity response scaled by the wave impedance (Zw Qe ) at the canal entrance. Canal length is 25 mm.
October 18-21, 2009, New Paltz, NY
‘Ear canal’
Microphone
P
Neck
P
Movable piston
UNDAMPED
Cable
Microphone ‘Ear canal’
Movable pistons
DAMPED
Eardrum piston
ZQ
Microphone
Neck
Microphone
Cavity
Cable
Figure 5: A diagram of ADECS with undamped eardrum (upper) and damped eardrum (lower).
‘Ear canal’
Eardrum piston
based on the time delay between the incident and reflected waves. Furthermore, the impedance of the eardrum can be estimated by computing the ratio Pd /Qd , where both the pressure and volume velocity at the eardrum can be solved from Eq. (1). Also energy- and intensity-based computation becomes possible when both pressure and volume velocity are available. In the next section, based on this P-Q formulation and a specific device for measurement, the approach is validated by practical measurements done using a physical simulator of the ear canal and a dummy head.
Figure 3: Measured pressure responses at the eardrum (Pd ) and at the canal entrance (Pe ). Measured volume velocity response scaled by the wave impedance (Zw Qe ) at the canal entrance. Canal length of undamped ADECS is 25 mm. plots the result of numerical computation of magnitude responses. Fig. 3 shows the corresponding magnitude responses as measured with the ADECS physical simulator described in Section 3.2. As can be seen, the simple computational model fits the measured data well. If the sound pressure is measured at a point in the ear canal, including the entrance, and if the acoustic properties of the canal and eardrum are known, it is possible to estimate the pressure at the eardrum using Eq. (1). There are two problems in practice, however. In addition to not knowing the canal and eardrum parameters, another inherent problem arises: the sensitivity of measurements especially at certain frequency ranges. Pressure responses at the canal entrance |Hs,e | in Figs. 2 and 3 show antiresonances at frequencies just above the resonances at the eardrum. This means that the measured pressure has minimum values at these frequencies. As a consequence, measuring the entrance pressure around these frequencies is very sensitive to measurement error. In Fig. 2, the volume velocity response at the canal entrance scaled by the wave impedance, i.e., Zw Qe , is also plotted. It can be seen that this information is most reliable where the pressure is unreliable. Therefore, by combining measured values of pressure Pe and volume velocity Qe , pressure Pd at the eardrum can be computed using Eq. (1) if good estimates of Zw and L are available. One more advantage of the P/Q-measurement at the canal entrance is achieved. The pressure Pe can be decomposed into (pressure) wave components √ (2) Pe+ = (Pe + Zw Qe )/ 2 √ − Pe = (Pe − Zw Qe )/ 2 (3)
3. MEASUREMENTS 3.1. The Microflown sensor (PU probe) The Microflown particle velocity transducer is an acoustic sensor for measuring particle velocity instead of sound pressure. It is a micromachined device that makes use of the fact that if two closely spaced heated wires are exposed to airflow then the upstream wire will be cooled more by the air than the downstream wire. Temperature changes affect the resistance of the wires, and the resulting resistance difference is measured with a bridge circuit that provides a signal proportional to the particle velocity [5] [6]. The “PU Match Probe” consisting of a particle velocity transducer and a miniature pressure microphone is shown in Fig. 4. The probe used in our studies measures the amplitude of the particle velocity in the direction tangential to the probe axis (i.e. the frontal direction). The Microflown signal conditioner used together with the transducer is calibrated so as to produce the same output voltage both for the pressure and the velocity sensor for a plane wave. 3.2. Ear canal simulator (ADECS) A custom-made ear canal simulator was used in the measurements for this study. The ADjustable Ear Canal Simulator (ADECS), which is depicted in Fig. 5, is made of a hard plastic tube with a diameter of 8.5 mm. The canal entrance is simply an open round hole. The ‘eardrum’ is made of a movable plastic piston so that the canal length of the simulator can be adjusted from 0 mm to 39 mm.
where Pe+ is the incident wave and Pe− the reflected wave. The reflectance Re = Pe− /Pe+ is useful in estimating the canal length
290
2009 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics
20
Measured Estim. Eq. (1) Estim. Eq. (4) Level [dB]
Level [dB]
20
October 18-21, 2009, New Paltz, NY
10
10
0 0 1k
Frequency [Hz]
10k
20k
1k
Figure 6: Pressure frequency response at the undamped eardrum of ADECS with canal length of 25 mm in free field conditions compared to estimation from measurements with the PU probe. The estimations are computed using Eq. (1) and Eq. (4).
0
Frequency [Hz]
10k
20k
Figure 7: Pressure frequency response at the damped eardrum of ADECS with canal length of 25 mm in free field conditions compared to estimations from measurements with the PU probe. A simulator with a rigid (or stiff) eardrum has a frequency response that is slightly different from that of a human ear. Resonance frequency peaks and antiresonance notches are sharp, whereas with real ears they are smoother. The ADECS simulator comes with exchangeable rigid and damped eardrums. With the damped eardrum, the frequency response at the eardrum and at the canal entrance are almost identical to average responses from human ear canals. A miniature microphone is fitted at the center of the eardrum piston. In the undamped eardrum the position of the eardrum microphone is adjustable by hand. In the damped eardrum some of the sound energy inside the ear canal is dissipated in a Helmholtz resonator. The volume of the resonator’s cavity can be changed by sliding the back wall of the cavity. Absorbing material inside the resonator’s cavity spreads the damping effect to a wider frequency range. The pressure frequency response at the eardrum of the undamped ADECS was measured in free field conditions in an anechoic chamber. The simulator’s canal entrance was pointed towards a loudspeaker 2 meters away. Several different canal lengths from 20 mm to 28 mm were used. The response with the 25 mm canal is shown in Fig. 6. The tip of the PU probe was placed in the center of the ear canal entrance pointing towards the eardrum. The pressure and velocity frequency responses at the entrance were measured directly after the eardrum pressure had been measured (the PU probe was put in place before the eardrum pressure response measurement). The velocity and pressure frequency responses obtained and calculation based on Eq. (1) were used for estimates of the eardrum pressure of the ADECS simulator with different canal lengths. The best results were obtained when the influence of the phase was first removed by using the formula: Pd,est = |Pe cosh(γL)| + |Zw Qe sinh(γL)|,
10k
20k
where L is the length of the tube section. In a lossless transmission line the propagation coefficient is given by γ = jω/c. The value for Zw Qe was given by the PU signal conditioner output. The estimated and measured pressure frequency responses at the eardrum agree well. The results with different canal lengths were similar, and as a typical result, the estimations with the 25 mm canal length are depicted in Fig. 6. In order to test the estimation method with the damped eardrum of ADECS, too, some additional measurements were made. In similar conditions to those above, the length of the ear canal was adjusted while the responses were measured at the eardrum (pressure) and at the entrance (pressure and velocity). The impedance of the damped eardrum has the effect of making the resonance frequency peaks clearly attenuated. The eardrum pressure response estimations, however, with different canal lengths were once again successful. The results for the 25 mm case are shown in Fig. 7.
Measured Estim. Eq. (1) Estim. Eq. (4)
1k
Frequency [Hz]
Figure 8: Pressure frequency response at the damped eardrum of DADEC with canal length of 25 mm in free field conditions compared to estimations. Azimuth angle is 90◦ .
Level [dB]
10
Measured Estim. Eq. (4) Estim. Eq. (5)
3.3. Dummy head (DADEC) The goal of the study was to investigate the possibilities of accurately estimating the pressure at the real eardrum based on measurements from the entrance of the ear canal using the PU probe. A custom-made acoustic dummy head that closely imitates the acoustics of the whole outer ear was used to test this idea. The Dummy with ADjustable Ear Canals (DADEC) [7] was used for this purpose. The dummy head (and torso) is of normal adult male size and it comes with exchangeable pinnae and ear canals. The DADEC was placed in an anechoic chamber two meters from a loudspeaker with an azimuth angle of 90◦ . Hence, the loudspeaker pointed towards the left ear of the head. Once again the length of the ear canal was altered while the responses at the eardrum and the canal entrance were measured. The eardrum microphone of the ear canal simulator was used for measuring the pressure response at the eardrum. The PU probe was carefully placed in the center of the concha for the pressure and velocity response measurements. A good position for the sensor was difficult to find, since it is difficult to determine the center of the entrance of the ear canal inside the concha. The effect of the point of measurement is left as a subject for future research. When the responses from the canal entrance had been captured, an estimation for the eardrum pressure was calculated using Eq. (4). The estimated and measured eardrum pressures with a total canal length of approximately 25 mm are depicted in Fig. 8. Compared to the responses from the ear canal simulator in Fig. 6 and 7, the DADEC responses in Fig. 8 are more complex due to the effects of the head, shoulders, and the pinna. Nevertheless, the estimated and measured pressure frequency responses at the eardrum are in close agreement.
(4)
291
2009 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics
Level [dB]
20
4.2. Audiological measurements
Eq (1) Eq (5)
One area of applications where knowing the ear canal acoustics is important is technical audiology. The estimation of the parameters of the eardrum and ear canal is of interest in audiological measurements, such as tympanometry for obtaining the mobility of the eardrum or measuring the impedance at the ear canal entrance. This information is also useful in the fitting of hearing aids. The method described above can be used in several such tasks. An estimate of the eardrum impedance is obtained from the reflectance computed from the pressure and velocity measured at the canal entrance when using an external sound source and the PU probe without even touching the ear if the probe is a few millimeters away from the entrance. The length of the canal can also be estimated from the reflectance data. The details of such measurements remain an interesting topic for future research.
10
0 1k
October 18-21, 2009, New Paltz, NY
Frequency [Hz]
10k
20k
Figure 9: Pressure frequency response at the undamped eardrum of ADECS (25 mm canal) modeled using Eq. (1) and Eq. (5). In addition to estimating the pressure using Eq. (4) a powerbased approach was also tested. When computing power flow through the ear canal entrance, a pressure equivalent is obtained by p Pd,est = |Pe |2 + |Zw Qe |2 , (5)
5. CONCLUSIONS A method for estimating the pressure at the eardrum from measurements at the ear canal entrance is presented and evaluated in practice. The method is based on measuring not only the pressure but also the volume velocity using a miniature temperaturegradient sensor. Different estimation approaches are revised. The approaches are validated using computational modeling, and real measurements from physical ear canal simulators and from a dummy head. It is shown that using volume velocity measurements, too, produces better estimates of the eardrum pressure, when compared with pressure-only measurements, also in cases when the ear canal parameters are not used in the estimation. The most prominent future applications are assumed to be measurements of HRTFs and audiological parameters.
where Pe is the measured pressure at the canal entrance and Zw Qe is the volume velocity response at the canal entrance scaled by the wave impedance. The free-field-calibrated and equalized velocity output signal of the Microflown signal conditioner is equal to Zw Qe . This estimation method was used for the DADEC dummy head with different canal lengths. The result with a 25 mm long ear canal is depicted in Fig. 8. With the case of the dummy head with the damped ear canal, both estimation methods yield almost exactly the same result. Estimations with ear canal simulators and different canal lengths showed similar results when Eq. (5) was applied for the estimation. Eq. (5) was also used together with the computational model described in Sec. 2. The two different equations give almost the same results as depicted in Fig. 9. Further analysis of alternative estimation methods based on PU measurements such as the one presented will be made in the future.
6. ACKNOWLEDGMENT This work is supported by the Academy of Finland (project no. 121252). We wish to thank Seppo Uosukainen and Henri Penttinen for their comments.
4. APPLICATIONS
7. REFERENCES
4.1. Head-Related Transfer Function (HRTF) measurements
[1] H. Hudde, A. Engel, and A. Lodwig, “Methods for estimating the sound pressure at the eardrum,” J. Acoust. Soc. Am., vol. 106, pp. 1977–1992, Oct. 1999.
Individual HRTF measurements are needed for binaural reproduction when accurate localization of sound objects is called for. HRTFs are typically measured with a microphone at the entrance of a blocked or unblocked ear canal or inside the ear canal. With a blocked ear canal, no information on the behavior of the individual ear canal is obtained, but the measurement will not be as sensitive to the position of the microphone [8]. For accurate HRTF measurements, the pressure would have to be measured at the eardrum. The effect of the ear canal and the eardrum would be included in the measurement and no post-processing of the transfer functions would be needed. Binaural reproduction which ignores the effects of the ear canal and the eardrum produces errors in timbre and localization [9]. The pressure-velocity measurement can be used for calculating the length and the impedance of the ear canal. These parameters combined with blocked ear canal measurements would yield the correct HRTFs at the eardrum. In addition, an alternative estimation of the eardrum pressure such as Eq. (5), based on the pressure-velocity measurement, would give more accurate HRTFs than the traditional techniques. Implementing the PU probe for HRTF measurements remains a task for the future.
[2] Microflown Technologies, www.microflown.com, May. 2009. [3] J. Blauert, Spatial Hearing: The Psychophysics of Human Sound Localization, revised ed. The MIT Press, 1997. [4] L. L. Beranek, Acoustics, 3rd ed. America, 1988.
Acoustical Society of
[5] F. Jacobsen and H.-E. de Bree, “Measurement of sound intensity: pu probes versus pp probes,” subm. to NOVEM, Jan. 2005. [6] H.-E. de Bree, “An overview of microflown technologies,” Acta Acustica united with Acustica, Jan. 2003. [7] M. Hiipakka, M. Tikander, and M. Karjalainen, “Modeling of external ear acoustics for insert headphone usage,” In proc. AES 126th Convention, Munich, Germany, 2009, May 2009. [8] S. Muller and P. Massarani, “Transfer-function measurement with sweeps,” J. Audio Eng. Soc., Jan 2001. [9] D. Griesinger, “Frequency response adaptation in binaural hearing,” In proc. AES 126th Convention, Munich, Germany, 2009, May 2009. 292
October 18-21, 2009, New Paltz, NY
ESTIMATING PRESSURE AT EARDRUM WITH PRESSURE-VELOCITY MEASUREMENT FROM EAR CANAL ENTRANCE Marko Hiipakka, Matti Karjalainen, and Ville Pulkki Helsinki University of Technology, Department of Signal Processing and Acoustics, Espoo, Finland Source
ABSTRACT It is important to know the sound pressure signal at the eardrum in headphone reproduction and in audiological applications. Unfortunately, it is difficult to conduct direct measurement safely. Estimating pressure signals at the eardrum based on measurements done elsewhere in the ear canal is sensitive to positioning errors. This is particularly the case at the canal entrance. In addition, not knowing the acoustic properties of the ear canal makes estimation difficult. This study shows that when both sound pressure and velocity are measured at the canal entrance using a pressure-velocity probe, the pressure signal at the eardrum can be estimated with much higher accuracy than from the pressure-only measurement. The method is demonstrated and validated by using physical simulators and computational modeling.
Rs
Pe
Zw,γ
Eardrum Qd Pd Zd
Ps
L Figure 1: Circuit model for the open ear canal.
pressure-and-velocity measurements are used to calculate the pressure at the eardrum of ear canal simulators accurately. Measurements are realized with the Microflown PU Match Probe [2], a custom-made ear canal simulator and a dummy head. 2. THEORY AND COMPUTATIONAL MODELING Acoustically the external ear contains the head and torso, the pinna and concha, and the ear canal up to the eardrum (tympanic membrane) [3]. The ear canal transmits sound signals and creates resonances due to standing waves between the entrance and the eardrum. In normal listening conditions, the entrance is open and the eardrum appears as a termination of relatively high acoustic impedance. Sound transmission through the canal is independent of sound arrival angle to the entrance point. Fig. 1 shows an equivalent circuit for sound transmission where an external sound source is feeding a signal through radiation impedance1 to the canal, and this acoustic transmission line is terminated by the eardrum impedance Zd . In an idealized model the radiation impedance Zs consists of a parallel connection of inductance Ls and resistance Rs [4] and the ear canal can be approximated as a cylindrical hard-walled tube with wave impedance Zw and negligible losses. The eardrum can be approximated by a relatively simple impedance model, such as a high resistance, although in reality it may be a quite complex function of frequency. An acoustic transmission line with a constant cross-sectional area S has a wave impedance Zw = ρc/S, where ρ is the density of air and c is the speed of sound. For a tube section, represented as a two-port, the signal variables P for pressure and Q for volume velocity as Laplace or Fourier transforms at ports 1 and 2 follow the equation » – » –» – cosh(γL) Zw sinh(γL) P1 P2 = 1 (1) sinh(γL) cosh(γL) Q Q1 2 Zw
Index Terms— eardrum, pressure, velocity, ear canal, Microflown, HRTF, audiology. 1. INTRODUCTION There are many applications where it is important to know the sound pressure at the eardrum. Unfortunately, direct measurement is possible only in special studies where extreme care can be taken not to harm the fragile membrane or other parts of the ear canal. For example, in audiological applications and particularly in binaural sound reproduction such direct measurements are out of the question in most cases. Therefore, sound signals have to be measured at other points of the ear canal, or often at the open or blocked canal entrance. Then the problem is to use such measurements to estimate sound pressure at the eardrum, because it is the basic reference point for head-related transfer functions, binaural headphone reproduction, and binaural recordings. There have been several efforts to estimate the pressure at the eardrum. Estimations are normally based on pressure measurements along the ear canal and on computational models. The standing waves inside the ear canal result in different pressure amplitudes at different parts of the canal, and the position of the microphone affects the sound pressure measurement values. In 1998 Hudde et al [1] presented the reflectance-phase method for estimating the pressure at the eardrum. It was shown that a good estimation can be achieved by using the minima of the pressure at the ear-canal entrance. When using Hudde’s approach, the physical parameters of individual ear canals as well as the volume velocity have to be determined. A micromachined particle velocity transducer that makes volume velocity measurements possible has recently become available. We are suggesting the use of this device for ear canal measurements as a new approach to estimate the eardrum pressure. In this article a computational model of the unblocked ear canal is first presented in order to shed some light on the physics of the ear canal and ear canal simulators. The model together with
978-1-4244-3679-8/09/$25.00 ©2009 IEEE
Ls
Ear canal Qe
where L is the length of tube section and γ is the wave propagation constant, γ = α + jβ with α being attenuation coefficient and β phase coefficient β = ω/c = 2πf /c. In Fig. 1 the port variables at the canal entrance are P1 = Pe and Q1 = Qe and at the eardrum P2 = Pd and Q2 = Qd . These equations together with Zs and Zd can be used to compute the transfer function from external source pressure Ps to the pressure Pe at the entrance as Hs,e = Pe /Ps and to the pressure Pd at the eardrum as Hs,d = Pd /Ps . Fig. 2 1 Due
289
to reciprocity principle [4].
2009 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics
P P ZQ
Figure 4: The Microflown PU Match Probe consists of a pressure microphone (above) and a particle velocity transducer (below). The diameter of the pressure microphone is 2.5 mm and the total diameter of the tip of the PU probe is 3.5 mm. [2]
Figure 2: Modeled pressure responses at the eardrum (Pd ) and at the canal entrance (Pe ). Modeled volume velocity response scaled by the wave impedance (Zw Qe ) at the canal entrance. Canal length is 25 mm.
October 18-21, 2009, New Paltz, NY
‘Ear canal’
Microphone
P
Neck
P
Movable piston
UNDAMPED
Cable
Microphone ‘Ear canal’
Movable pistons
DAMPED
Eardrum piston
ZQ
Microphone
Neck
Microphone
Cavity
Cable
Figure 5: A diagram of ADECS with undamped eardrum (upper) and damped eardrum (lower).
‘Ear canal’
Eardrum piston
based on the time delay between the incident and reflected waves. Furthermore, the impedance of the eardrum can be estimated by computing the ratio Pd /Qd , where both the pressure and volume velocity at the eardrum can be solved from Eq. (1). Also energy- and intensity-based computation becomes possible when both pressure and volume velocity are available. In the next section, based on this P-Q formulation and a specific device for measurement, the approach is validated by practical measurements done using a physical simulator of the ear canal and a dummy head.
Figure 3: Measured pressure responses at the eardrum (Pd ) and at the canal entrance (Pe ). Measured volume velocity response scaled by the wave impedance (Zw Qe ) at the canal entrance. Canal length of undamped ADECS is 25 mm. plots the result of numerical computation of magnitude responses. Fig. 3 shows the corresponding magnitude responses as measured with the ADECS physical simulator described in Section 3.2. As can be seen, the simple computational model fits the measured data well. If the sound pressure is measured at a point in the ear canal, including the entrance, and if the acoustic properties of the canal and eardrum are known, it is possible to estimate the pressure at the eardrum using Eq. (1). There are two problems in practice, however. In addition to not knowing the canal and eardrum parameters, another inherent problem arises: the sensitivity of measurements especially at certain frequency ranges. Pressure responses at the canal entrance |Hs,e | in Figs. 2 and 3 show antiresonances at frequencies just above the resonances at the eardrum. This means that the measured pressure has minimum values at these frequencies. As a consequence, measuring the entrance pressure around these frequencies is very sensitive to measurement error. In Fig. 2, the volume velocity response at the canal entrance scaled by the wave impedance, i.e., Zw Qe , is also plotted. It can be seen that this information is most reliable where the pressure is unreliable. Therefore, by combining measured values of pressure Pe and volume velocity Qe , pressure Pd at the eardrum can be computed using Eq. (1) if good estimates of Zw and L are available. One more advantage of the P/Q-measurement at the canal entrance is achieved. The pressure Pe can be decomposed into (pressure) wave components √ (2) Pe+ = (Pe + Zw Qe )/ 2 √ − Pe = (Pe − Zw Qe )/ 2 (3)
3. MEASUREMENTS 3.1. The Microflown sensor (PU probe) The Microflown particle velocity transducer is an acoustic sensor for measuring particle velocity instead of sound pressure. It is a micromachined device that makes use of the fact that if two closely spaced heated wires are exposed to airflow then the upstream wire will be cooled more by the air than the downstream wire. Temperature changes affect the resistance of the wires, and the resulting resistance difference is measured with a bridge circuit that provides a signal proportional to the particle velocity [5] [6]. The “PU Match Probe” consisting of a particle velocity transducer and a miniature pressure microphone is shown in Fig. 4. The probe used in our studies measures the amplitude of the particle velocity in the direction tangential to the probe axis (i.e. the frontal direction). The Microflown signal conditioner used together with the transducer is calibrated so as to produce the same output voltage both for the pressure and the velocity sensor for a plane wave. 3.2. Ear canal simulator (ADECS) A custom-made ear canal simulator was used in the measurements for this study. The ADjustable Ear Canal Simulator (ADECS), which is depicted in Fig. 5, is made of a hard plastic tube with a diameter of 8.5 mm. The canal entrance is simply an open round hole. The ‘eardrum’ is made of a movable plastic piston so that the canal length of the simulator can be adjusted from 0 mm to 39 mm.
where Pe+ is the incident wave and Pe− the reflected wave. The reflectance Re = Pe− /Pe+ is useful in estimating the canal length
290
2009 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics
20
Measured Estim. Eq. (1) Estim. Eq. (4) Level [dB]
Level [dB]
20
October 18-21, 2009, New Paltz, NY
10
10
0 0 1k
Frequency [Hz]
10k
20k
1k
Figure 6: Pressure frequency response at the undamped eardrum of ADECS with canal length of 25 mm in free field conditions compared to estimation from measurements with the PU probe. The estimations are computed using Eq. (1) and Eq. (4).
0
Frequency [Hz]
10k
20k
Figure 7: Pressure frequency response at the damped eardrum of ADECS with canal length of 25 mm in free field conditions compared to estimations from measurements with the PU probe. A simulator with a rigid (or stiff) eardrum has a frequency response that is slightly different from that of a human ear. Resonance frequency peaks and antiresonance notches are sharp, whereas with real ears they are smoother. The ADECS simulator comes with exchangeable rigid and damped eardrums. With the damped eardrum, the frequency response at the eardrum and at the canal entrance are almost identical to average responses from human ear canals. A miniature microphone is fitted at the center of the eardrum piston. In the undamped eardrum the position of the eardrum microphone is adjustable by hand. In the damped eardrum some of the sound energy inside the ear canal is dissipated in a Helmholtz resonator. The volume of the resonator’s cavity can be changed by sliding the back wall of the cavity. Absorbing material inside the resonator’s cavity spreads the damping effect to a wider frequency range. The pressure frequency response at the eardrum of the undamped ADECS was measured in free field conditions in an anechoic chamber. The simulator’s canal entrance was pointed towards a loudspeaker 2 meters away. Several different canal lengths from 20 mm to 28 mm were used. The response with the 25 mm canal is shown in Fig. 6. The tip of the PU probe was placed in the center of the ear canal entrance pointing towards the eardrum. The pressure and velocity frequency responses at the entrance were measured directly after the eardrum pressure had been measured (the PU probe was put in place before the eardrum pressure response measurement). The velocity and pressure frequency responses obtained and calculation based on Eq. (1) were used for estimates of the eardrum pressure of the ADECS simulator with different canal lengths. The best results were obtained when the influence of the phase was first removed by using the formula: Pd,est = |Pe cosh(γL)| + |Zw Qe sinh(γL)|,
10k
20k
where L is the length of the tube section. In a lossless transmission line the propagation coefficient is given by γ = jω/c. The value for Zw Qe was given by the PU signal conditioner output. The estimated and measured pressure frequency responses at the eardrum agree well. The results with different canal lengths were similar, and as a typical result, the estimations with the 25 mm canal length are depicted in Fig. 6. In order to test the estimation method with the damped eardrum of ADECS, too, some additional measurements were made. In similar conditions to those above, the length of the ear canal was adjusted while the responses were measured at the eardrum (pressure) and at the entrance (pressure and velocity). The impedance of the damped eardrum has the effect of making the resonance frequency peaks clearly attenuated. The eardrum pressure response estimations, however, with different canal lengths were once again successful. The results for the 25 mm case are shown in Fig. 7.
Measured Estim. Eq. (1) Estim. Eq. (4)
1k
Frequency [Hz]
Figure 8: Pressure frequency response at the damped eardrum of DADEC with canal length of 25 mm in free field conditions compared to estimations. Azimuth angle is 90◦ .
Level [dB]
10
Measured Estim. Eq. (4) Estim. Eq. (5)
3.3. Dummy head (DADEC) The goal of the study was to investigate the possibilities of accurately estimating the pressure at the real eardrum based on measurements from the entrance of the ear canal using the PU probe. A custom-made acoustic dummy head that closely imitates the acoustics of the whole outer ear was used to test this idea. The Dummy with ADjustable Ear Canals (DADEC) [7] was used for this purpose. The dummy head (and torso) is of normal adult male size and it comes with exchangeable pinnae and ear canals. The DADEC was placed in an anechoic chamber two meters from a loudspeaker with an azimuth angle of 90◦ . Hence, the loudspeaker pointed towards the left ear of the head. Once again the length of the ear canal was altered while the responses at the eardrum and the canal entrance were measured. The eardrum microphone of the ear canal simulator was used for measuring the pressure response at the eardrum. The PU probe was carefully placed in the center of the concha for the pressure and velocity response measurements. A good position for the sensor was difficult to find, since it is difficult to determine the center of the entrance of the ear canal inside the concha. The effect of the point of measurement is left as a subject for future research. When the responses from the canal entrance had been captured, an estimation for the eardrum pressure was calculated using Eq. (4). The estimated and measured eardrum pressures with a total canal length of approximately 25 mm are depicted in Fig. 8. Compared to the responses from the ear canal simulator in Fig. 6 and 7, the DADEC responses in Fig. 8 are more complex due to the effects of the head, shoulders, and the pinna. Nevertheless, the estimated and measured pressure frequency responses at the eardrum are in close agreement.
(4)
291
2009 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics
Level [dB]
20
4.2. Audiological measurements
Eq (1) Eq (5)
One area of applications where knowing the ear canal acoustics is important is technical audiology. The estimation of the parameters of the eardrum and ear canal is of interest in audiological measurements, such as tympanometry for obtaining the mobility of the eardrum or measuring the impedance at the ear canal entrance. This information is also useful in the fitting of hearing aids. The method described above can be used in several such tasks. An estimate of the eardrum impedance is obtained from the reflectance computed from the pressure and velocity measured at the canal entrance when using an external sound source and the PU probe without even touching the ear if the probe is a few millimeters away from the entrance. The length of the canal can also be estimated from the reflectance data. The details of such measurements remain an interesting topic for future research.
10
0 1k
October 18-21, 2009, New Paltz, NY
Frequency [Hz]
10k
20k
Figure 9: Pressure frequency response at the undamped eardrum of ADECS (25 mm canal) modeled using Eq. (1) and Eq. (5). In addition to estimating the pressure using Eq. (4) a powerbased approach was also tested. When computing power flow through the ear canal entrance, a pressure equivalent is obtained by p Pd,est = |Pe |2 + |Zw Qe |2 , (5)
5. CONCLUSIONS A method for estimating the pressure at the eardrum from measurements at the ear canal entrance is presented and evaluated in practice. The method is based on measuring not only the pressure but also the volume velocity using a miniature temperaturegradient sensor. Different estimation approaches are revised. The approaches are validated using computational modeling, and real measurements from physical ear canal simulators and from a dummy head. It is shown that using volume velocity measurements, too, produces better estimates of the eardrum pressure, when compared with pressure-only measurements, also in cases when the ear canal parameters are not used in the estimation. The most prominent future applications are assumed to be measurements of HRTFs and audiological parameters.
where Pe is the measured pressure at the canal entrance and Zw Qe is the volume velocity response at the canal entrance scaled by the wave impedance. The free-field-calibrated and equalized velocity output signal of the Microflown signal conditioner is equal to Zw Qe . This estimation method was used for the DADEC dummy head with different canal lengths. The result with a 25 mm long ear canal is depicted in Fig. 8. With the case of the dummy head with the damped ear canal, both estimation methods yield almost exactly the same result. Estimations with ear canal simulators and different canal lengths showed similar results when Eq. (5) was applied for the estimation. Eq. (5) was also used together with the computational model described in Sec. 2. The two different equations give almost the same results as depicted in Fig. 9. Further analysis of alternative estimation methods based on PU measurements such as the one presented will be made in the future.
6. ACKNOWLEDGMENT This work is supported by the Academy of Finland (project no. 121252). We wish to thank Seppo Uosukainen and Henri Penttinen for their comments.
4. APPLICATIONS
7. REFERENCES
4.1. Head-Related Transfer Function (HRTF) measurements
[1] H. Hudde, A. Engel, and A. Lodwig, “Methods for estimating the sound pressure at the eardrum,” J. Acoust. Soc. Am., vol. 106, pp. 1977–1992, Oct. 1999.
Individual HRTF measurements are needed for binaural reproduction when accurate localization of sound objects is called for. HRTFs are typically measured with a microphone at the entrance of a blocked or unblocked ear canal or inside the ear canal. With a blocked ear canal, no information on the behavior of the individual ear canal is obtained, but the measurement will not be as sensitive to the position of the microphone [8]. For accurate HRTF measurements, the pressure would have to be measured at the eardrum. The effect of the ear canal and the eardrum would be included in the measurement and no post-processing of the transfer functions would be needed. Binaural reproduction which ignores the effects of the ear canal and the eardrum produces errors in timbre and localization [9]. The pressure-velocity measurement can be used for calculating the length and the impedance of the ear canal. These parameters combined with blocked ear canal measurements would yield the correct HRTFs at the eardrum. In addition, an alternative estimation of the eardrum pressure such as Eq. (5), based on the pressure-velocity measurement, would give more accurate HRTFs than the traditional techniques. Implementing the PU probe for HRTF measurements remains a task for the future.
[2] Microflown Technologies, www.microflown.com, May. 2009. [3] J. Blauert, Spatial Hearing: The Psychophysics of Human Sound Localization, revised ed. The MIT Press, 1997. [4] L. L. Beranek, Acoustics, 3rd ed. America, 1988.
Acoustical Society of
[5] F. Jacobsen and H.-E. de Bree, “Measurement of sound intensity: pu probes versus pp probes,” subm. to NOVEM, Jan. 2005. [6] H.-E. de Bree, “An overview of microflown technologies,” Acta Acustica united with Acustica, Jan. 2003. [7] M. Hiipakka, M. Tikander, and M. Karjalainen, “Modeling of external ear acoustics for insert headphone usage,” In proc. AES 126th Convention, Munich, Germany, 2009, May 2009. [8] S. Muller and P. Massarani, “Transfer-function measurement with sweeps,” J. Audio Eng. Soc., Jan 2001. [9] D. Griesinger, “Frequency response adaptation in binaural hearing,” In proc. AES 126th Convention, Munich, Germany, 2009, May 2009. 292
