By Timothy D. Trine, PhD
IHS offers a diversity of options for obtaining continuing education credit: seminars and classroom training, institutional courses, online studies and distance learning programs. This article represents yet another opportunity. Upon successful completion of the accompanying test you will earn one CEU.
(Please check with your licensing board to verify acceptance of continuing education credit for this activity.)
The hearing aid industry has made significant strides in the fidelity and sophistication of signal processing and fitting strategies utilized in current state-of-the-art hearing aids. Nonetheless, understanding speech-in-noise continues to be the chief complaint of people who wear hearing aids.
Directional processing remains the only proven signal processing approach to improve speech recognition in noise once audibility has been maximized. Moreover, many of the obstacles to using directional hearing aids have been solved recently: cosmetic concerns, size, directional performance, noise performance and stability over time. Given these facts, it is somewhat surprising that directional hearing aid sales have not grown more significantly over the past several years. Why is it that many hearing healthcare professionals have failed to take advantage of the potential benefits afforded to their patients by directional processing? Is the often-repeated assertion that the directivity of the outer ear provides as much or more directivity than a directional hearing aid and as such CIC fittings represent the best approach? Why does it appear that the real-world performance of directional hearing aids fails to live up to the expectations set in the laboratory?
I believe the answers to these questions are multifaceted but can be trimmed down to several technical and clinical components:
By understanding these issues, clinicians can take a more informed approach to selecting hearing aid candidates who are likely to get the maximum benefit from directional processing and, equally important, to set the patient’s expectations for benefit in their unique environments.
The source of the advantage afforded from directional processing is not hard to understand. Directional processing creates a differential sensitivity favoring stimuli that are coming from directly in front of the wearer and attenuating stimuli that come from the sides and rear. This selective sensitivity is most often characterized by measuring the polar response as seen in Figure 1.
Figure 1. Schematic polar response patterns around head. Omnidirectional
pattern is shown on the left, directional on the right.
The two most important design considerations for a directional system are the in situ directivity as a function of frequency (which is most appropriately quantified by the Directivity Index (DI) on a KEMAR) and the sensitivity loss. The in situ directivity is an important design consideration because it correlates well with the real-world speech-in-noise improvements that can be achieved for typical speech-in-noise environments where the signal of interest is in front of the listener and the noise is fairly diffuse. In other words, the higher the directivity, the better the performance will be when listening to a dinner companion at a crowded restaurant. Because of the relatively limited real estate available in a custom hearing aid, however, a competing design consideration is the sensitivity loss for signals arising from zero degrees azimuth (or directly in front of the listener).
Figure 2 illustrates the fundamental dilemma. For traditional directional designs utilizing two microphones—or two microphone ports with a single directional microphone—there is a direct relationship between the low-frequency sensitivity loss and the microphone separation, such that the closer two microphones are placed the more low-frequency attenuation will be observed.
Figure 2. Relative sensitivity for two directional systems with different port spacings relative to an omnidirectional microphone. From Thompson (2003).
This, of course, can become problematic for building smaller hearing aids because the attenuation from the sensitivity loss will require additional gain from the hearing aid circuit to compensate for the hearing loss of the wearer. Moreover, because this sensitivity loss occurs at the front end of the hearing aid, the noise floor will be increased by the amount of gain required to achieve the compensation. For the very small port separations required on a true, in-the-canal (ITC) hearing aid, the compensation for all or part of the sensitivity loss can lead to complaints of “circuit noise” in even moderate background noise. To complicate the design trade-off between directivity and sensitivity loss, higher directivity in the high frequencies can be achieved with closer port spacing.
One further complication arises for custom directional hearing aid design when one considers using matched microphone pairs rather than a mono-mic design. The issue is microphone drift and, until recently, the industry has been fairly quiet about the implications from this phenomenon.
Microphone drift simply refers to the change over time in the relative sensitivity and phase response of a pair of microphones. Achieving high directivity with today’s delay-and-sum directional designs in a custom hearing aid requires exquisite matching of the microphone pair. For example, if we would like to maintain a directional response at 500 Hz (the dominate peak for most noise sources in the real world), the relative sensitivity between the two microphones cannot deviate by more than .15 dB (assuming a 5 mm port separation and a DI tolerance of 1 dB). In other words, if a pair of microphones drifts apart by more than 1/5 of a dB, the polar response shown on the right side of Figure 1 will start to look like the polar response on the left. The key question is whether or not it is reasonable to believe that a pair of microphones can maintain this level of matching over time. The short answer is no.
Considering this question we examined the factors most likely to cause microphones to drift in the field. The primary culprit is changes in humidity. As relative humidity increases, the diaphragm of the microphone absorbs the moisture, causing a change in the sensitivity and phase response. Likewise, as relative humidity decreases, the diaphragm dries out, causing changes in the opposite direction. Are these changes significant enough to cause microphones to drift beyond our performance tolerance? The data in Figure 3 answer this question. Shown are the changes in sensitivity measured after removing and allowing to dry perfectly matched pairs of microphones from a humidity chamber with relative humidity of 90% at 40 degrees centigrade. We refer to this test as accelerated aging and it is designed to simulate a year of normal wear and tear for a microphone. Clearly, most of the microphone pairs fall outside our tolerance for drift of .15 dB for a custom hearing aid. Fortunately, on the other hand, all of the data fall within the tolerance for a 12 mm spacing which is typical for BTEs.
Figure 3. Change in sensitivity at 500 Hz after accelerated aging for 25 matched pairs of microphones.
Given these results, it is clear that using two microphones to achieve excellent directivity in custom hearing aids is risky. The exquisite matching required to maintain directional performance over time with two microphones requires active signal processing and assumptions that have not been proven. These data might partly explain some of the poor reports of custom directional performance in the real world despite specification sheets showing good performance. Consequently, one approach is to attack the problem directly and eliminate the need for microphone matching by using a single-microphone design for our custom products referred to as Precision Directional Imaging (PDI). It should be emphasized that with perfect microphone matching, both approaches (combining two omnidirectional microphones or utilizing a single microphone) can yield comparably excellent directivity performance as measured by the DI on KEMAR in the lab, but maintaining excellent performance in the field is the design objective.
In addition to solving the drift problem, PDI solves both the cosmetic and noise performance issues mentioned above. By reducing the port separation to less than 4 mm, a true ITC sized instrument can be built. That brings us to the question of low-frequency sensitivity loss. Although the solution is proprietary, the evidence that the problem has, in fact, been elegantly addressed is demonstrated in Figure 4 which shows the sensitivity of three different directional systems relative to the sensitivity of an omnidirectional system (the red line at 0 dB). Notice that the sensitivity for the PDI directional system is comparable to the sensitivity for a directional system designed with a matched pair of microphones with a 12.5 mm spacing. In other words, we have achieved the noise performance of a directional system based on a 12.5 mm port spacing in a package with less than 4 mm port spacing.
Figure 4. Sensitivity as a function of frequency for the custom PDI system (green), and dual omni designs at 5 mm port spacing (blue) and 12.5 mm spacking (black), relative to omnidirectional sensitivity (red).
The preceding section should convincingly demonstrate that the technical obstacles historically observed for high performance directional systems are now ancient history. One of the lingering misconceptions in clinical practice, however, is the relative merit of a CIC microphone placement when compared to the performance available with a directional system. Certainly there are many advantages available with a successful CIC fitting including increased bandwidth, reduced electroacoustic gain required relative to other styles to achieve comparable insertion gain and, of course, improved cosmetics and user acceptance. When the question is the ability to understand speech in background noise, through, well-designed directional microphone systems win with flying colors. The data shown in Figure 5 illustrate this point with the Hearing in Noise Test (HINT) on a group of listeners with normal hearing. These listeners were tested in a laboratory designed to be very challenging and representative of the acoustics of the typically noisy real-world environment.
Specifically, the speech material was presented from a loudspeaker directly in front of the listener at a distance of one meter and seven independent noise sources were used to create a diffuse sound field.
The HINT score is expressed as the average signal-to-noise-ratio (SNR) required to achieve 50% sentence recognition so more negative scores reflect better performance. Importantly, the data shown in Figure 5 were obtained from the same subjects with hearing aid fittings where the real-ear aided responses (REAR) were meticulously matched across all aids so that the performance differences can only be attributed to the directional processing. It is readily apparent that directional performance swamps the improvements seen with changes in microphone placement alone.
Figure 5. Average HINT SNRs for five normal hearing subjects wearing well-matched
hearing aids as labeled.
From Olson et al. (2004).
Moreover, even BTE directional hearing aids yield significantly better performance in noise when compared to the performance of an omnidirectional CIC. Although these data were collected on subjects with normal hearing (data collection with hearing impaired subjects is in process), the change in performance observed here should generalize to subjects with hearing loss because of the care taken in matching the REAR. In other words, we expect that the relative differences between styles will be the same with hearing impaired subjects, but the absolute performance will likely be poorer due to the residual hearing loss even after being aided.
Having tackled the technical realities that may have contributed to poor directional performance and further demonstrated the performance possible from directional versus omnidirectional processing regardless of hearing aid style, it is time to turn to the most difficult variable in the performance equation: the patient. Ultimate success cannot be achieved without evaluating how an individual patient’s capacity to understand speech in the presence of background noise and the acoustics of their environments interact to impact their performance and ultimately their satisfaction with directional hearing aids.
The fact that many people with hearing loss require a higher SNR than normal hearing listeners to achieve the same level of performance on a speech recognition task has been well recognized for many years (Tillman et al., 1970). Figure 6 shows the average performance of a group of 12 listeners with non-presbycusic hearing loss (dotted lines) relative to a group of normal hearing listeners (solid line) on a speech recognition task as a function of SNR.
Notice that to achieve the same level of performance, the SNR for the unaided hearing impaired subjects would have to be increased by 14 dB.1 This change in SNR is what many have labeled the hearing loss for SNR loss.
Figure 6. Monaural interference functions for unaided and aided listening for 12 subjects with non- presbycusic sensorineural losses. From Tillman et al., 1970.
The data in Figure 7 are the results from word recognition testing (NU-6 word lists) of 511 subjects seen at the Vanderbilt Bill Wilkerson Center in Nashville, Tennessee. All testing was completed with recorded stimuli at a fixed signal level (80 dB HL). As a consequence, many individual datum fall on top of each other; this is indicated in the graphs by the symbol size. The data from these subjects were arbitrarily sorted according to their average hearing loss at 2, 3, and 4 kHz (HFA HL). Those subjects with HFAs less than 50 dB are shown in the left plot and those with HFA HLs greater than 50 are shown on the right.
Fifty dB HL was chosen as the cutoff because it represents a reasonable boundary between those subjects with primarily outer hair cell or strial loss (<50) and those subjects with some degree of inner hair cell or neural loss (>50) based on current models of cochlear function (Moore and Glassberg, 1998). There are two primary differences between the two groups of subjects. The proportion of subjects for which SII accurately predicts performance is much lower for the subjects with more severe hearing loss (36% vs. 72%). In addition, the range of performance for the subjects with more severe hearing loss is large relative to the subjects with less severe high-frequency hearing loss. Note that these results are for speech-in-quiet but the extent by which performance is under-predicted by the SII can be interpreted as an SNR loss. Similar results are seen for speech reception in noise.
Figure 7. Word recognition performance expressed in rationalized acrsine units as a function of SII for two groups of patients sorted according to their HFA PTA (2, 3 and 4 kHz). Symbol size indicates the number of overlapping datum. The two curves represent ± 2 standard deviations of the predicted performance based on SII.
These data lead to two important clinical conclusions: 1) on average, the mean SNR deficit is related, although somewhat crudely, to the severity of the hearing loss and, more importantly, 2) an individual’s SNR deficit cannot be predicted by the pure-tone audiogram alone.2 This conclusion is also supported by several data sets reported by Killion (1997).
11 The 14 dB difference observed here is larger than observed in other studies (e.g. Wilson 2003) due to the severity of the hearing loss and likely due to the fact that the etiology of presbycusis was not included in the subject sample because pure presbycusis is likely dominated by strial loss.
2 For the reader who may not find this conclusion intuitively obvious for the speech-in-quiet data set, recall that the data were all collected at a fixed signal level; consequently, individuals with similar audiograms will have similar audibility.
From a physiological basis, hearing loss for speech-in-noise is likely related to inner hair cell or neural (e.g. spiral ganglion cell) loss. Clinical decision making and counseling would be much easier if we always had access to the data complementing the audiograms in Figure 8. The additional data shown below the audiogram in Figure 8 are collectively referred to as the cytocochleogram.
Figure 8. Audiograms and cytocochleograms from two human subjects. From Schuknecht (1992).
Specifically, these plots show the health of the inner hair cells (IHC), the three rows of outer hair cells (OHC), the stria vascularis and the spiral ganglion cells. The height of the black bar indicates the relative health of the relevant structures such that a black bar that fills the box indicates complete destruction and no bar indicates an anatomically normal structure.
For example, the cytocochleogram for the audiogram on the left at 500 Hz shows healthy IHC, healthy OHC, complete destruction of the stria vascularis and approximately 85% of the neural tissue intact. Contrasting these two patients we see that although the magnitude of their pure-tone hearing loss is comparable, the underlying physiological basis is quite different. The data for the audiogram on the left show that hearing loss is largely due to loss of the stria vascularis whereas the hearing loss on the right is largely due to the patches of IHC loss and loss of neural tissue.
If we had access to these data at the time of the hearing aid fitting, we might predict that the patient on the left would perform better than the patient on the right. Unfortunately, these data are only available directly from a post-mortem analysis. Nonetheless, understanding the likely underlying etiology can lead us to better clinical decision making. We see some indication of this behaviorally in the lower speech recognition score of 64% for the subject on the right vs. 100% for the subject on the left. It is important to note, however, that this information alone will not lead to unambiguous decisions regarding candidacy for directional hearing aids. This information coupled with an understanding of the acoustics of their environment can lead to more powerful predictive accuracy.
Figure 9. Word recognition performance in percent correct as a function of SNR. Solid line represents the performance for a patient with a 0 dB hearing loss for speech-in-noise, dotted line represents the performance for a patient with a 10 dB SNR loss.
Although laboratory performance of directional signal processing has been repeatedly demonstrated (Walden et al., 2000), the performance of directional microphones in everyday life has not met the expectations created by the laboratory performance (Cord et al., 2002; Surr et al., 2002).
In part, this observation can be explained by considering the acoustics of everyday life in conjunction with the SNR loss for an individual patient (Ricketts, 2001). Consider the performance of two hypothetical patients as shown in Figure 9. The solid line represents the performance as a function of SNR for a patient with a 0 dB loss for speech-in-noise (as would be expected for the patient represented by the left panel of Figure 8) and the dashed line represents the performance for a patient with a 10 dB SNR loss. Now consider the environmental acoustics for a challenging speech-in-noise situation; say a conversation in a crowded bar or restaurant. It would not be surprising to have an input SNR of –5 dB in such an environment. For a well-designed directional hearing aid, an average SNR improvement of 5–6 dB should be expected. For the patient with a 0 dB SNR loss, a 5 dB improvement in SNR would change their performance from 50 to over 95% correct for conversational speech under these acoustic conditions. Needless to say, the 0 dB SNR patient would experience significant benefit from the directional signal processing. The patient with a 10 dB SNR loss, however, with the same well-designed directional system in the same environment would move from zero to 5% correct and very likely would not notice a change in their performance.
On the other hand, if the environmental acoustics change such that the input SNR was +5 dB, as might be expected at a less crowded restaurant or perhaps even the dinner table at home, the patient with the 10 dB SNR loss would experience a significant improvement in performance (50 to 95%) while the patient with the 0 dB SNR loss would not yield an improvement in performance because his/her performance would already by asymptotic.
These scenarios emphasize the importance of assessing the real-world acoustics for individual patients in order to set realistic expectations, to help the counseling process and perhaps to assess candidacy for directional signal processing. For example, the patient with a 10 dB SNR loss might still represent a good candidate for directional hearing aids if that person rarely found themselves in environments with very poor SNR (0 to –5 dB) or if they were able to actively manage the environment to improve the SNR ratio at the input to the hearing aid microphones.
On the other hand, the patient with a 0 dB SNR loss might not be a good candidate for directional hearing aids if they rarely found themselves in noisy environments. Of course, estimating the environmental acoustics of your patient’s everyday life without directly measuring it is not necessarily an easy task. In the not-too-distant future this will be made easier by data logging hearing aids that can store the level and SNRs over time as the patient wears the hearing aids and these data can then be used for additional optimization and counseling at the first post-fitting visit.
In the meantime, the following assumptions about the acoustics of everyday life can help:
Figure 10. Audiometric profile of patient who would likely be a good candidate for directional processing.
In a perfect world, the clinician would know the underlying physiology for each individual’s hearing loss and have a clear picture of that individual’s acoustical environments. Armed with this information, one could make informed decisions regarding the efficacy of various signal processing algorithms and make more confident hearing aid recommendations. Unfortunately, we don’t live in a perfect world but it is possible to infer quite a bit of information about both of these important predictors of directional hearing aid success through routine clinical measurements. Figure 10 represents a reasonable starting point for identifying low-risk directional hearing aid candidates; that is, those candidates who have little hearing loss for speech-in-noise once full bandwidth audibility is provided.
In other words, an audiogram that falls within the shaded region coupled with open-set word recognition in quiet that is well-predicted by the Speech Intelligibility Index (SII) (greater than or equal to 68% for PB-max), represents a patient who is likely to have little IHC or neural damage and, consequently, a low SNR loss (less than 3 dB). Recall that an individual with a small SNR loss should realize significant benefit from directional processing even in the most challenging circumstances.
As illustrated in Figures 6 and 7, however, estimating the SNR loss from the pure-tone audiogram and speech-in-quiet performance alone is at best an estimated guess. A better approach can be achieved without too much additional valuable clinical time. Several relatively quick and direct measures of hearing loss for speech-in-noise are now readily available in many languages. Two of the most popular in the U.S. are the HINT and the QuickSIN. Both of these measurements can be performed in less than ten minutes and can provide a direct estimate of the hearing loss for speech-in-noise conveniently expressed in dB.
An alternative “better” approach would be to do an assessment of the environmental acoustics of those situations that are most important for the individual patient. Again, this requires a bit of guesswork regarding the SNR that the individual is likely to encounter, but the guidelines provided above should help. More importantly, this dialogue with the patient opens the door for counseling regarding realistic expectations and about the importance of actively controlling the acoustics of their environment to optimize their performance.
Finally, the best approach to directional candidacy is probably represented by measuring the SNR loss in the clinic and completing an assessment of the important everyday acoustics for the individual. The underlying assumption behind each of these candidacy approaches is that, armed with enough information about the individual’s hearing loss for speech-in-noise and the acoustics of their environments, the savvy clinician can make an informed recommendation.
It is important to point out that the assessment process outlined here need not be limited to the decision between omnidirectional and directional processing. Clearly, there are many patients who have such severe hearing loss for speech- in-noise that they would be better served by considering alternative means of improving the SNR such as FM systems or other assistive listening devices. THP
Cord, M. T., Surr, R. K., Walden, B. E., and Olson, L. (2002). Performance of directional microphone hearing aids in everyday life. Journal American Academy of Audiology 13: 295–307.
Killion, M. C. (2004). Myths about hearing in noise and directional microphones. The Hearing Review February 2004: 14, 16, 18–19, 72–73. Ricketts, T. (2000).
Kochkin, S. (2002). 10-Year Customer Satisfaction Trends in the US Hearing Instrument Market. The Hearing Review 9(10): 14–25, 46.
Moore, B. C. J., and Glassberg, B. R. (1998). Use of a loudness model for hearing aid fitting. I. Linear hearing aids. British Journal of Audiology 32, 301–319.
Ricketts, T. A. (2001). “Directional hearing aids.” Trends in Amplification 5(4): 139–176.
Ricketts, T. A. and Mueller, H. G (2000). Predicting directional hearing aid benefit for individual listeners. Journal of the American Academy of Audiology 11: 561–569.
Ricketts, T. A. and Hornsby B. W. Y. (2003). Distance and reverberation effects on directional benefit. Ear and Hearing 24(6): 472–484.
Schuknecht, H. F. (1992). Disorders of Aging in Pathology of the Ear. Cambridge, Harvard University Press.
Smoorenburg, G. F. (1992). Speech reception in quiet and in noisy conditions by individuals with noise-induced hearing loss in relation to their tone audiogram. Journal of the Acoustical Society of America 91: 421–437.
Surr R. K., Walden, B. E., Cord, M. T., and Olson, L. (2002). Influence of environmental factors on hearing aid microphone preference. Journal American Academy of Audiology 13: 308–322.
Walden, B. E., Surr, R. K., Cord, M. T. and Olson, L. (2000). Comparison of benefits provided by different hearing aid technologies. Journal American Academy of Audiology 11: 540–560.
Wilson, R. H. (2003). Development of a speech-in-multitalker-babble paradigm to assess word recognition performance. Journal American Academy of Audiology 14: 453–470. The Hearing Lions
