The recent proliferation of directional hearing aids has created a need for improved objective evaluation techniques to assess and compare their effectiveness. The ANSI S3.35–1985 standard (“Methods of Measurement of Performance Characteristics of Hearing Aids under in situ Working Conditions”) specified a method of characterizing directionality that is insufficient to meet the current needs of hearing aid designers, clinicians, dispensers, and researchers for predicting benefit derived from directional hearing aids.
This original ANSI standard specifies a two-dimensional (2D) directivity characterization, using a horizontal polar planar measurement obtained with the hearing aid mounted on a KEMAR manikin while the manikin is rotated. Among the parameters reported from these measurements are Manikin Unoccluded Ear Directional Response, Simulated Insertion Directional Response, and Simulated Aided (in-situ) Directional Response. These 2D directional measurements are considered to be only an approximation of actual or three-dimensional (3D) measurement results since they assume vertical plane rotational polar pattern symmetry, which is usually not achieved with hearing aids mounted on KEMAR or a wearer.1 More accurate predictions of directional benefit can be made with 3D measurements, as will be discussed later.
The Directivity Index (DI) is an electroacoustical directional performance benchmark that has been used in the telecom industry since the 1940s,2 and adapted by the hearing aid industry in the 1985 version of the standard. Unfortunately, DI measurement procedures were not specified in this original version of the ANSI S3.35 standard.
THREE-DIMENSIONAL DI DEVELOPED
Consequently, a proposal was made in 1999 within the Acoustical Society of America (ASA) S3-WG48 committee for hearing aid measurement standards to add Directivity Index (DI) measures to the S3.35 standard. To test whether or not the DI could be reliably measured in different laboratories, Robert Schulein, an S3-WG48 member, organized a round robin experiment in 2000 at the facilities of seven S3-WG48 members.
Planar horizontal 2D polar pattern measures were made on a first-order directional BTE hearing aid, first in isolation in an anechoic chamber and then mounted on KEMAR. The resulting planar 2D-DI calculations were surprisingly consistent. Across the seven laboratories there was only a 0.7-dB total range of 2D-DI averaged across a frequency range of 200–6300 Hz, with the largest deviation across all labs being only 2.2 dB at 200 Hz.
However, since hearing aid wearers listen to sounds in a 3D acoustical environment, calculating a 2D-DI does not tell the entire story. It is not logical to assume that the same nulls that occur in the horizontal plane around a hearing aid wearer will also occur in a vertical plane. Notwithstanding the appeal of a more accurate measure of real-world performance, the greater difficulty of making 3D measurements has led to the use of only 2D polar patterns and DIs up until now.
Figure 1 illustrates figuratively the difference in what is involved in making 2D and 3D polar pattern measurements of directional hearing aid performance on KEMAR. To accomplish a 2D polar pattern measurement (Figure 1a), it is quite straightforward simply to rotate KEMAR in a horizontal plane while leaving a single loudspeaker in a fixed position.
The data in Figure 2 illustrate the differences between 3D-DI and 3D-DI calculations for the unoccluded KEMAR manikin at the laboratory of one S3-WG48 member (and co-author). In the left graph of Figure 2, the blue curve is the 3D-DI and the red curve is the 2D-DI of KEMAR unaided. The single graph on the right is the difference between the two curves in the left panel.3 In general, the 3D-DI is higher than the 2D-DI over most of the frequency range.
Likewise, for an ITE hearing aid with a first-order gradient microphone, the left panel in Figure 3 shows that the 3D-DI is higher generally than the 2D-DI, except for a few small frequency regions across the entire range.3
Thus, to calculate more accurately the actual DI provided by a directional hearing aid, a 3D polar pattern measurement is needed. Early work in the S3WG48 committee for updating S3.35–1985 to include 3D measures began with a proposal from Elmer Carlson, a working group member, to incorporate measures at locations defined by an icosahedron.4 The working group agreed later that the icosahedron had too few sound source locations to achieve the desired measurement resolution.
Oleg Saltykov, another S3-WG48 member, prepared several initial drafts of a standard in the period 2001–2003 that were reviewed and updated by the committee, which meets for only half a day twice a year. After much deliberation about the number and location of 3D measurement points and whether and how to average the DI over frequency, a draft was circulated in 2003 within the working group recommending as a good compromise calculating DI from 48 generally uniformly distributed measurement points on an imaginary spherical surface (Figure 1b). This resulted in a sound source location array of 12 sources at 0° and +30°and 6 sources at +60°, using a single moving loudspeaker and/or multiple loudspeakers and KEMAR. Also, the standard recommends a linear-weighted average over frequency of DI, but other averaging methods, including AI-DI and SII, are specified in Annex C.
During this same time period, individual members of the working group were making many directivity measurements to determine if the recommended procedures in the draft standard were viable. Thereafter, one S3-WG48 member, Daniel Warren, spent many hours over the next year modifying the drafts to make them compatible with ASA and ANSI standards requirements and formats. Ultimately, Warren prepared a final draft that became the revised S3.35 standard, which was published in 2004 after review by members of the ASA S3 committee.
CALCULATING THE DI
Whereas the original ANSI S3.35–1985 standard did not specify any DI calculations, the revision, ANSI S3.35–2004, does, covering both 3D and 2D measures.5 There are two measurement components necessary for calculating the DI. Namely, it is the ratio of the sound energy arriving from straight to the front of the hearing aid (normally on-axis re: the microphone ports) to the average sound energy arriving from all directions—the latter typically referred to as “diffuse” if this energy arrives randomly in time and uniformly from all directions.
Ideally, then, DI should relate to the directional performance in a situation in which someone is talking in front of the hearing aid in an environment of diffuse noise, and should thus be predictive of the Speech Reception Threshold (SRT) in noise. The first measurement of the sound incident from directly to the front of the hearing aid has traditionally been made only in an anechoic chamber, in which there is little reflective sound energy.
For the diffuse sound energy measurement, rather than measuring sound energy from several directions one direction at a time, the measurement time can be reduced by using a diffuse sound field, with the sound incident from all angles. Using a highly reverberant chamber, it is possible to develop a sufficiently diffuse sound field with which to measure directly the randomly incident sound energy from all angles.
A perfectly reverberant chamber can be thought of as a kind of gold standard to which 3D anechoic measures can be compared. Only a single measurement of random energy would be required in a perfectly diffuse sound field, rather than multiple measurements at many angles in an anechoic environment. However, a perfectly diffuse sound field has been very difficult to achieve in practice because it is difficult to make a perfectly reflective room, and the results do not always correlate with measurements performed in an anechoic chamber.6 While no facility can produce a perfectly diffuse sound field, a number of laboratories can produce a sufficiently diffuse field; in architectural acoustics applications, these accredited spaces typically are used for characterizing a material's acoustical absorptive coefficients.
Steven Julstrom, a member of the S3-WG48 committee, compared DI measures at his laboratory on a directional ITE hearing aid averaged at several locations in a reverberant room to 3D measures at 48 points in an anechoic chamber, per ANSI S3.35–2004.7 Figure 4 shows that these two measures were quite close over most of the frequency range. However, to obtain such a good correlation to the 3D anechoic chamber measures, Julstrom averaged together 15 measurements made in the reverberant room at different locations and orientations. Therefore, while measures in a reverberant room on directional hearing aids might theoretically be faster than 48 discrete 3D measurements in an anechoic chamber, they may be impractical to perform routinely.
CAVEAT RE: USE OF DIRECTIVITY INDEX
Although the Directivity Index was the S3-WG48 committee's measure of choice and was incorporated into ANSI S3.35–2004, there are some indications that it may not predict real-world benefit as accurately as is desired. Cord et al. have cited differences between the comparatively small real-world benefit provided by directional hearing aids in some studies and the larger benefit predicted by the reported DI and the speech-recognition-in-noise measures in the laboratory.8 The reason for this disparity is uncertain, but one may speculate that the hearing aid earmold vent reduces directionality relative to ear-molds and earmold simulators used in the laboratory, and actual listening environments are more reverberant than the laboratory conditions in which measurements for the DI calculation and speech recognition were made.
One important factor is the critical distance of the space in which measurements are made. Critical distance (Dc) is defined as the distance from the sound source at which the directly incident sound energy is equal to the reflective sound energy. At source-to-listener distances less than Dc, directly incident sound energy is greater than reflective sound energy. Directivity and speech-recognition measures in the laboratory may often be made at distances less than Dc.
In contrast, in real-world environments, much of the actual listening—to both the desired signal and the interfering noise—may occur at distances greater than Dc, resulting in more reverberant or diffuse sound energy than direct energy. To the extent that the distance between source and listener exceeds Dc, the defined measure that we call DI will fail to represent real-world performance of a directional system, independent of whether 2D or 3D measurements are made.
In a diffuse-field scenario, directional-microphone hearing aids, whether they incorporate adaptive or fixed polar patterns, become less effective farther away from the source than at shorter source-to-listener distances. As a result, patients should be counseled about the importance of listening distance in reverberant situations.
To test if the 3D-DI could be measured with repeatable precision, a round robin was conducted in 2005 at the facilities of eight S3-WG48 members for two different, first-order, ITE directional hearing aids on KEMAR. The two ITE aids were deliberately chosen to have specific, free-field polar patterns: ITE #1 was engineered to have a free-field cardioid polar pattern and ITE #2 was engineered to have a free-field figure eight (dipole) pattern.
The individual 3D-DIs over frequency for the hearing aid with the cardioid pattern on KEMAR at the eight laboratories are shown in the top panel of Figure 5, while the bottom panel shows the averaged DIs with standard deviations over frequency from the round robin measurements. The maximum standard deviation was 0.3 dB below, and 0.7 dB above, 1500 Hz. These results indicate that the newly revised S3.35–2004 ANSI standard provides a repeatable and relatively precise procedure for comparing electroacoustical directional performance of hearing aids.
In Figure 6, a horizontal slice of data was taken from the full 3D data sets of the round robin aids. The data slices were chosen at 500 Hz, a frequency at which the 3D-DI happened to be equal to 4.5 dB for both ITE #1 and ITE #2. It is interesting to note that, although the 3D-DIs are the same, the polar-pattern responses are radically different, one being a cardioid and one being a figure eight or dipole pattern. This reveals that the 3D-DI value alone is insufficient for characterizing the electroacoustical directional response of a device.
Ideally, 3D-DI values should be accompanied by some type of benchmark that describes the strength of the directional pattern's rear lobe. One useful benchmark is the relative front-to-back strength, sometimes reported as a 0°/180° dB ratio. For example, the data of Figure 6 indicate that 0°/180° = 14 dB for ITE #1 whereas 0°/180° = 2 dB for ITE #2. These numbers would be helpful in correlating electroacoustical directional characteristics to perceptual outcomes in clinical studies. It is quite possible, for example, that ITE #1 would perform differently from ITE #2 in a clinical speech-in-noise test.
There are many adaptive directional hearing aids in the marketplace, but no consistent method for assessing their directional performance. ANSI S3.35–2004 specifies that the directivity be assessed with all adaptive systems—such as the directional null locations, automatic directional-omnidirectional switching, automatic gain control, and noise management—disabled so they do not alter the measured sound levels used to calculate the DI. While this approach is good from a standard measurement point of view, the resulting measures do not reflect the performance of adaptive directional hearing aids in real-use conditions.
Consequently, the S3-WG48 committee is investigating ways in which ANSI S3.35–2004 can be extended to assess adaptive directional systems without disabling the adaptive algorithms, which will better predict speech-in-noise performance by hearing aid wearers in real-world use conditions.
The 2004 revision of the ANSI S3.35 standard provides us with additional methods with which to assess the performance of directional hearing aids—the 3D polar pattern measurement and the Directivity Index calculation. Although the DI is a useful metric in comparing the performance of directional hearing aids, the signal-to-noise ratio benefit that it predicts may not be realized by hearing aid wearers in real-world listening situations and with all types of directional-microphone systems.
The authors would like to thank Jerry Yanz and Brent Edwards for reviewing and commenting on an early draft of this article.
1. Saltykov O: 2-D vs. 3-D Directivity Index data provided to the ASA S3-WG48 working group on hearing aid measurement standards at the December 2002 working group meeting.
2. Bauer B: Graphical determination of the random efficiency of microphones. J Acoust Soc Am
1940; 11 (April):477.
3. Burns T: Round robin measurements of 3-D directivity using the ANSI S3.35 protocol. Paper 1pEA14 presented at ASA meeting, Minneapolis, October 2005.
4. Carlson E: Icosahedron concept for a 3-D directional measurement. Memo to ASA S3-WG48 standards committee, 1998.
5. ANSI: Methods of Measurement of Performance Characteristics of Hearing Aids Under Simulated Real-Ear Working Conditions
(S3.35) New York: American National Standards Institute, 2004.
6. Roberts M, Shulein R: Measurement and intelligibility optimization of directional microphones for use in hearing aid devices. Paper 4515 (B-3) presented at 103rd AES convention, New York, September 1997.
7. Julstrom S: Comparison of 48-point 3-D anechoic and reverberation room diffuse sound field measurements of Directivity Index. Paper 1pEA4 presented at ASA meeting, Minneapolis, October 2005.
Copyright © 2007 Wolters Kluwer Health, Inc. All rights reserved.
8. Cord M, Surr R, Walden B, Olsen L: Performance of directional microphone hearing aids in everyday life. JAAA