by Christopher Schweitzer, Ph.D.
Background – From Real Ear to Real Hear!
In two prior posts,1,2 the argument was put forth that, despite good intentions, real ear measures (REMs) have important short comings. Essentially, it was proposed that there is a nontrivial disconnect between the quite acceptable precision of both audiometric and electroacoustic measures and their relation to the intended accuracy for ‘verifying’ the successful fitting of hearing aids. The limitations noted included (briefly):
- The collection of acoustical measures via a probe tube in a hearing aid user’s independent ear canals provides only a murky estimate of that listener’s binaural preferred listening levels.
- Audiometric signals, unlike many speech elements, are bereft of Time/Intensity (durational) properties, ignoring the well-known sensory integration property whereby brief signals require greater intensity than those longer than the 0.2 to 0.5 seconds used in audiometry.
- The “target gain” assumptions based on pure tone (PT) thresholds for steady state sinusoids, also obtained in deliberately isolated ears, further exchange ‘precision’ for ‘accuracy,’ when applied to the more complex signal management processes of everyday audition.
- The fact that frequency modulations are important acoustical components of speech understanding (and noting that some neurons are specifically ‘tuned’ to fire on Frequency Modulated (FM) signals), adds further uncertainty in the conventional approach’s attempt to draw a line from the PT audiogram to a desired amplification pattern, or “target.”
This wobbly path from “Inference” to “Evidence” is illustrated in Figure 1 as a review of one of the prior pieces2.
While acknowledging the contribution of REM (real-ear measures) procedures in closing prior well-known electroacoustic gaps in clinical information from prior artificial coupler measures, it was argued that they do not go far enough to be considered a singular reference, or exemplar, of hearing aid fitting success. The full argument is comprehensive and only partially reviewed here, but the relevant question is if there are important limitations of the current method, what are the alternatives to the standard protocol?
Evolving Tablet Approach to Successful Hearing Aid Fitting
An evolving approach described here has been recently implemented into a tablet application from principles used in a private practice for several years. The method imposes very little burden on clinical operations, but provides refreshing confidence-building outcomes. The original technique, as used for several years in a private practice in Colorado, was to place a small speaker approximately one meter in the front of the hearing aid client. That speaker, connected to an audiometer in the fitting room, was accompanied by a small cohort of 2 to 4 additional ‘dummy’ speakers, as shown in Figure 2.
In the technique dubbed Real HearTM, the client’s hearing aids are connected in the usual wired or wireless method for fittings and adjustments. Several low, mid, and high- frequency warbled tones (e.g., 500, 1k, 1.5k, 2k, 3k Hz) are then presented in the unaided condition. While adjusting the attenuator for the various signals, the listener simply reports the ‘borderline audible’ levels to note a quasi-threshold. The attenuator level for the unaided condition is recorded, and a second series of minimally audible levels are obtained with the aids on in the initial proposed fitting in this highly simplified functional gain measure. A simple example of findings for a client in the Unaided (U) and an Aided result (A) is shown in Figure 3.
The procedure does not require precise SPL (sound pressure level) calibration as the results are relative improvements, i.e. functional gain, although linearity of the attenuator is obviously desirable. The U (unaided) and A (Aided 1, or 2 or more) recordings are shared with the client and modifications to the fittings are easily rendered since the procedure is done in the ‘fitting room’.
One of the most important outcomes, a quick ‘Binaural Balance’ test is run to provide the clinician important evidence of the correctness of the fitting not included on any REM tests. Binaural balance has long been known to be important, not just for localization accuracy, but as a fundamental mechanism of noise management (e.g., release from masking, auditory object tracking) in the complex binaural auditory system3,4. Hence, when hearing aids are even just slightly asymmetric, perhaps producing 3 to 5 dB more to one ear than the other, it forces the binaural processing system to adapt and compensate. A few moments of clinical time to validate binaural balance is hardly a superfluous waste of time in a professional fitting. Indeed, it is difficult to understand why such a measure is not fundamental to best practices! Trusting the assumptions of fitting screen software based on the inputted thresholds may be a significant omission of professional stewardship.
To accomplish the simplified balance test one or more of the pulsed warbled tonal signals is presented at a comfortable, or moderately soft, level. In addition to the ‘live speaker’ in the front of patient, the several ‘dummy’ speakers located in the general forward direction, but arrayed some +/- 45 to 75 degrees to the sides provide visual foils. The listener is simply asked to close his/her eyes and point to where the sound seemed to be located. A few decibels of asymmetrical delivered sound in the Left-Right fitting generally moves the sound’s location to some position off the actual source. This can, and often does, occur despite the target gain proposals on the fitting screen due to numerous technical and/or psychoacoustic variables that may proceed from the threshold audiogram to the ‘target proposals’ and numerous other small technical factors, as suggested in Figure 1.
The dummy speakers allow the listener to more reliably indicate their auditory experience, rather than be influenced by the belief that it must be from a single known speaker source. In other words, perceptual bias to an expected source may permit a denial of actual imbalances. The visual prospect of alternative sound sources permits a wider allowance for purely auditory perceptual localization. Small adjustments in the hearing aid gain in the general frequency region of the test signal are easily made to the hearing aids which remain connected to the fitting software. In many cases ‘tweaks’ of 2 to 4 dB in gain reductions to the ‘leaning towards’ side, or increases in the ‘leaning away from’ ear side rapidly move the signal to the desired for center position. An example is shown in Figure 4. The client can either mark, on an illustration similar to Figure 4, their perceived location or simply point for the clinician to take note. The fitting professional can then promptly adjust the appropriate parameters to bring the location to the correct point in space. This usually requires only a few moments to accomplish.
One important addition in the App version is the use of filtered Speech Tokens (STs) for the Low/Mid/High-pitch listening targets. This makes use of standardized and highly researched speech materials developed at the Audio Quality Research Lab of the Institute for Telecommunication Sciences, in Boulder, Colorado.5 Mixed gender spoken monosyllables were concatenated into an uninterrupted string and then octave band filtered into the Low/Mid/High pitch regions shown on Figure 5 screen shot. The sliders on the touch screen allow either the clinician or the hearing aid user to adjust the levels to some agreed level such as “I can understand 50% of the words” or “It’s at a Comfortably Clear Level.”
In the App version, a wireless Blue Tooth speaker is typically used as the delivery source to remove the need for a wired connection. An example of the outcome from a recent fitting using the Speech Tokens on the App is shown in Figure 5. The three pitches show lower (better) levels in the Aided condition than in the Unaided. A Broadband Speech Noise signal used on the right side of the screen shows a similar reduction in level to achieve a ‘barely audible’ (quasi-threshold) report. The noise used is a phase-controlled, multi-tone noise shaped to mimic the long-term average of speech, and interrupted to prevent noise reduction circuitry from activation in the hearing aids.
It is worth reiterating that this is a relative gain measure, where observable improvement of aided versus unaided is the goal. In the absence of hard reference levels, given the many variables of room acoustics, speakers, ambient noise, etc. clinicians or family members can make their own observation of audibility improvements by, for example, comparing the softer aided levels to their own perceptual experience.
Consumer Confidence in the Fitting
This approach is believed to exchange the patina of “precision” for the prospect of greater ‘accuracy’ in terms of confidence in the relevance of the fitting’s ‘verification’. Many clinicians will be uncomfortable with that trade off and may wish to anchor the measure against some known normative standard. That is entirely reasonable, but it is also arguable that the Real Hear approach provides a clear indication to the professional and the client that a ‘walk around’ improvement of audibility has been obtained in the office. Obviously, other routine checks for the consumer’s tolerance of ambient noise, loud transients, and so on still require proper attention, as they would with REMs. However, the fundamental assumption of the Real Hear method is there is a higher level of confidence of binaural balance and delivered benefit than the inferential assumption of a threshold-to-gain-in-ear-canal which underlies REMs.
This is part of an ongoing project to explore ways to extend beyond the organic coupler measures of REMs and predictive assumptions from classical audiometric measures. It is expected that numerous changes and enhancements will follow as clinical experiences from professionals are collected and implemented into the App. The audiometric version described in Figure 2 was valuable in working out an alternative approach that simply recruited a used or retired audiometer, and it may be unsuitable for many fitting rooms. But it also has the advantage of readily available speaker outputs of warble tones, and familiar attenuator and presentation controls. The present version of the Real Hear iPad™ app only incorporates the Speech Tokens (filtered speech segments in 3 filter bands). Taking advantage of the software flexibility of a smart tablet, it is expected that an enlarged library of sound options, such as FM signals (warble tones) controlled and produced by the App will follow. Hence, an audiometer as shown in Figure 2 is unlikely to be required to verify both improvements in speech type signals and to obtain confirmation of binaural balance, but it is a viable clinical option.
The introduction of REMs was an important and commendable step towards closing the gaps on hearing aid fittings and listener preferences. Their becoming the presumed gold standard of best practices6 should not go without imagining that perhaps there could be something more relevant to the consumer experience. As with any widely-accepted and institutionalized practice, and in some cases legally mandated, inertia can discourage the consideration of improvements. Hence, the assumption that proper verification of hearing aid fittings demands REMs, merits critical discussion as to the completeness of information to affirm a ‘successful’ fitting of a consumer’s binaural amplification system. Any proposed supplementary or alternative approach needs proven clinical efficacy and a rationale that alleges to close some of the remaining gaps in fitting accuracy as obtained by REMs. Clinical experience with this alternative audiological procedure, with the trademark name, Real Hear, has led to the introduction of a workable tablet App. The approach incorporates direct aided hearing experience of the user, rather than an inferential verification based on electroacoustic data. The intention is not to promote a specific product or procedure, but to encourage fellow hearing professionals to exercise their creative energies in the continuous advance of professional skills presented to consumers in the customization of hearing aid products. Gathering of experience in multiple locations and the data-based cycle of feedback and continuous improvement is in the early stages of an ongoing project.
Much of the content of this and the previous cited HHTM blogs will be presented as a Learning Module at AudiologyNow in Indianapolis, LM219 – Moving Beyond Real-Ear Measures for Hearing Aid Validation and Verification Thursday, April 6, 2017
- Schweitzer, HC. hearinghealthmatters.org/waynesworld/2017/all-ears-are-real-ears-but-thats-not-enough
- Schweitzer, HC. hearinghealthmatters.org/waynesworld/2017/evolution-electroacoustic-measures-hearing-aids
- Moore, BCJ. Introduction to the Psychology of Hearing 5th 2003. Academic press.
- Bregman A. Auditory Scene Analysis: The Perceptual Organization of Sound. Cambridge, Mass: MIT Press;1990.
- Voran, S, Cattelier, A. Speech Codec Intelligibility Testing in Support of Mission-Critical Voice Applications for LTE. NTIA Report 15-520. 2015.
- Amlani, A., Pumford, J., Gessling, E. Improving patient perception of clinical services through real ear measurements. Hearing Rev 23(12) 2016.
Christopher Schweitzer, Ph.D. is a longtime industry consultant and clinical researcher in Boulder, Colorado where he and his wife have operated private practice offices for over 31 years. He serves as Technical Audiology consultant for IMHear Corp of Downers Grove, Illinois, which sponsored the development of the Real Hear app. He can be reached at firstname.lastname@example.org.