The quest for the perfect hearing aid has been inextricably linked to the quest for the perfect hearing aid fitting. Despite countless research and development hours, research papers, book chapters, and software products, the achievement of that optimum setting of electroacoustic characteristics yielding the best possible outcome for each patient has been frustratingly elusive.
This paper summarizes the history of prescriptive hearing aid fitting and introduces a new patient-centered hearing aid optimization tool called SoundPoint. The SoundPoint system offers patients the ability to assist in the fine-tuning of their hearing aids. With a simple mouse movement or finger sweep, anybody can navigate through an auditory world of changes to sound quality.
Figure 1 shows a patient using SoundPoint to assist in the fine-tuning of their hearing aids. All adjustments are made to the hearing aids in a smooth and seamless manner, allowing patients to find their preferred subjective sound quality without the need to articulate their own complex auditory perception. A companion paper1 in an upcoming HR provides additional information on the clinical validation and use of SoundPoint.
Early Approaches to Optimizing the Hearing Aid Response
The earliest attempts at optimizing hearing aid fittings were known as “uniform” and “selective amplification” strategies.2 Uniform amplification applied the same amount of amplification to all of the speech frequencies, whereas selective amplification applied varying amounts of amplification to the speech frequencies as determined by audiometric and speech thresholds, most comfortable listening levels, and/or predicted speech “articulation” (ie, intelligibility).
Lybarger3 proposed an early prescriptive formula known as the half-gain rule, which, as its name implies, prescribed gain equal to one-half of the mid to high frequency hearing loss with less low frequency prescribed gain to avoid upward spread of masking. The concept of selective or prescribed amplification was challenged by Davis et al4 and Radley et al5 in the mid-1940s who argued for a common frequency response approach6 that consisted of only two responses: flat and approximately 6 dB per octave.
At about the same time that Davis et al4 and Radley et al5 published their recommendations, Raymond Carhart7 introduced a hearing aid selection approach for returning WWII veterans that made no attempt to prescribe gain or output. Instead, the “Carhart Method” involved comparing clinical performance across several different hearing aids. The specific performance measures included sensitivity to sound, tolerance limits, effectiveness in background noise, and efficiency in distinguishing small sound differences. The instrument that performed “best” among these dimensions was the one that was selected for the patient, after which the veteran attended a comprehensive multi-week aural rehabilitation program.8
The Prescriptive Hearing Aid Fitting
The use of prescribed or selective amplification (providing specific gain as a function of frequency-specific hearing impairment) increased in the 1970s and 1980s, enabled by technological advances that allowed for flexible adjustments of gain across frequency. Hearing care subsequently experienced a proliferation of threshold-based formulas designed for linear amplification. Examples included the Berger method,9 the Libby Method,10 Prescription of Gain and Output (POGO),11 the National Acoustics Laboratories formulas (NAL),12 and the Desired Sensation Level (DSL).13
The introduction of multiband compression technology required advanced fitting rationales that prescribed both frequency- and level-dependent gains, and a number of threshold-based formulas for these non-linear prescriptions were developed during the 1990s, including NAL-NL1,14 DSL [i/o],15 FIG6,16 and CAMEQ.17 Many of these also allowed for suprathreshold measures (MCL, UCL, loudness scales) to be included in the fitting formula calculation. Linear and non-linear prescription examples include the Shapiro,18 MSU,19 and CID methods,20 the Loudness Growth in half-Octave Bands (LGOB),21 and the Independent Hearing Aid Fitting Forum method (IHAFF).22
In addition to the approaches listed above, most major hearing aid manufacturers develop their own formulas that are integrated into their proprietary fitting software. These algorithms are constantly being refined and updated as new products and technologies are introduced.
As a demonstration of how fitting strategies have changed through the decades, David Hawkins23 presented an historical perspective in the November/December 1994 HR, characterizing four eras (post-Lybarger et al) of hearing aid selection and fitting as:
- The Carhart Era;
- The Let the Word Lists Select the Hearing Aid Era;
- The Let the Frequency Response Formula Select the Hearing Aid Era; and
- The Non-Linear Era.
Which Is the Best Prescription?
We now have a variety of fitting methods that use the audiogram and other measures to calculate prescriptions. Which approach, then, is the best? Threshold-based approaches are designed to maximize intelligibility, sometimes at the expense of loudness comfort; suprathreshold approaches may account for comfort but risk sacrificing intelligibility in the process.
Factors other than audibility and comfort may affect the goodness of fit from these prescriptions (eg, spectral balance for different sounds, appropriateness of compression ratios, preservation of binaural cues), and compression parameters (eg, kneepoints and time constants) can also affect the success of fitting prescriptions.24 Unfortunately, there is not a large enough body of evidence to draw any conclusions concerning superiority between or within any of the fitting approaches.
There is compelling evidence that, when given a choice, individuals prefer different gain than prescribed, requiring multiple visits to adjust hearing aid settings to a preferred final fit. In a review of 5 separate studies involving 189 subjects, Keidser and Dillon25 compared the NAL-NL1 prescribed gain to the preferred gain for a 65 dB input speech signal. While half (49%) of the subjects preferred the NAL-NL1 prescribed gain (±2dB), 46% of the subjects preferred less gain.
Preferred gain appears to be influenced by hearing aid experience and gender25 and degree of hearing loss.26 The development of trainable hearing aids (where the start-up gain changes in response to VC adjustments made by the user) also provides some insight into patient preferences27,28 and suggests that patients prefer the “trained” setting to the initially prescribed setting.
What is clear is that fitting formulas only provide a starting point toward optimizing the fit for any individual. As evidenced by the repeat clinical visits necessary to obtain a satisfactory fitting,the initial fit is rarely the best fit. Why is it that individuals with the same hearing loss may prefer different amounts of gain and that audiogram-based fitting algorithms merely provide starting points for the fitting? The reasons can be explained, in part, by the following:
- Different mechanisms of sensorineural damage can result in the same audiogram but quite different changes to sound transduction in the cochlea, requiring different gain and compression ratios for optimal fitting.
- Different people have different preferences for loudness comfort based on the way that their auditory system has developed over their lives, something we currently attribute to “personal preference.”
- Hearing aids do not perfectly compensate for hearing loss, and their fitting is a compromise across multiple aspects of sound quality and intelligibility. Different people experience different sound environments and have different preferences for those trade-offs, resulting in different preferences for the hearing aid fitting.
There is little doubt that most patients require fine-tuning of their initial fit—regardless of which prescriptive formula is used. A challenge to this fine-tuning process is that the dispensing professional can only assume what the patient hears through a question-and-answer approach. Often patients have a difficult time describing their amplified experience, and the clinician has to guess not only what the patient is hearing but also how to adjust the gain and compression parameters to optimize the fitting. Given all of the parameters available for adjustment, this can be a difficult task and can be compared to a chef creating a meal without being able to taste the food.
An additional difficulty with achieving an optimal fit is that fine-tuning is often complaint-driven only. If a fitting is “pretty good,” the patient may not request an adjustment to the initial fit; even if an adjustment is desired, the dispensing professional will have a difficult time determining how to improve the fitting. In most cases, there is the potential for a better set of electroacoustic characteristics than the patient’s initially fit parameters, but neither the patient nor the audiologist is capable of knowing what that ideal setting is and how to achieve it given current fitting practices.
Clearly, there is a need for a new approach, one that directly involves the patient in optimizing the hearing aid fitting.
Origins of SoundPoint
SoundPoint evolved from technology that has been under development by UC Berkeley Professor David Wessel for over 30 years.29,30 This long-standing research has produced computer interface tools using spatial arrangements of musical material to provide expressive control of musical instrument synthesis in live performance. We have adapted this technology into a simple and intuitive self-tuning system for controlling multiband compression in hearing aids.
Spatial arrangement is a natural and intuitive tool for organizing information. Experimental psychology has greatly enriched our understanding of the mechanisms of auditory perception through the use of techniques such as multidimensional scaling that rely on spatial representations. In the early 1970s, Wessel demonstrated that the spatial arrangements of musical timbres from multidimensional scaling could function as a palette of musical material that could be used by performers and composers to specify timbre relationships and transformations. In the first application of this technique to a large-scale composition, Jean-Claude Risset used a timbre space created by multidimensional scaling to compose passages of his 1978 work Mirages for the Ensemble Intercontemporain in Paris.
SoundPoint presents a simple, square control space on a computer screen. Each pixel in the space corresponds to a different compressor setting. As the cursor is moved around the control space, the compressor parameters in the hearing aid change smoothly and instantaneously. (In Wessel’s application, moving the cursor smoothly and instantaneously changed the parameters of a musical instrument synthesizer.)
While this approach seems simple, the underlying technology is complex, and there are numerous technical challenges to be overcome to make it work well. Wessel’s research addressed two major challenges:
1) Wessel developed a method of perceptual multidimensional scaling to map spatial relationships in the control space onto the numerous parameters of a complex system (here, multiband compression) in a way that has perceptual consistency (ie, hearing aid settings that sound similar are located close together in the control space, and hearing aid settings that sound different are located far apart in the control space).
2) Wessel employed a neural network interpolator to affect a perceptually smooth and continuous transition between hearing aid settings as the cursor moves between the corresponding locations in the control space.
From Perceptual Multidimensional Scaling to SoundPoint
Wessel’s control structure solves one of the critical issues in allowing patients to fine-tune their hearing aids themselves: how to provide a user interface that is simple to use and understand,
and that allows someone to make adjustments to dozens of parameters simultaneously in a way that has perceptual consistency.
In Wessel’s system, graphical objects in the user interface represent individual musical timbres, and musicians arrange these objects in the control space. Similar or related timbres are placed close together, while dissimilar or unrelated timbres are kept apart, forming a perceptually intuitive network. The organization or layout of these timbre elements governs the behavior of the system and the nature of the timbral transformations that are available to the performer using the control space.
In SoundPoint, the adjustment of gain and compression ratios in 32 channels simultaneously is as simple as moving a cursor around a computer screen (Figure 1), making it easy for the patient to find their preferred fitting—and easy return to the original fitting if nothing preferred is found. By applying perceptual multidimensional scaling to associate positions in the SoundPoint control space with fine-tuning adjustments in the hearing aid, and by implementing an interpolator as a means of “filling in the gaps” between these adjustment points, we provide an interface that allows a patient to select from over 1 million possible fine-tuning adjustments to 64 parameters in a quick, easy-to-use, and easy-to-understand interface.
Early Research on System
Early work on the development of SoundPoint focused on the arrangement of parameter sets, termed presets, into a perceptually intuitive network or layout. The organization of the presets in the layout governs the behavior of the system and is experienced by the patient as the control space.
In order for listeners to be able to use this control space to adjust the hearing aid parameters, the space must have perceptual consistency—meaning that parameter sets that sound similar are placed close together and those that sound different are far apart. With a seamless organization of the control space, listeners are able to consistently navigate through this network to preferred settings that are perceptually similar.
A series of experiments was conducted to validate these claims and understand the use of this tool as a means of identifying optimal or preferred configurations of hearing aid processing parameters. There were three main objectives in validating this system as a hearing aid fitting tool:
1) In regard to the layout of the control space, we set out to understand if a naïve listener can arrange the presets in a repeatable, perceptually logical way, creating a smooth interpolation within the control space.
2) Similarly, we were interested in determining if, during the navigation task, listeners were able to consistently navigate to preferred parameter sets that were perceptually similar and preferred over the presets that initially comprised the layout.
3) We wanted to understand the influence the layout exerts on the navigated settings and particularly on the validity and optimality of those navigated settings.
These experiments involved both listeners with normal and impaired hearing who performed the layout and navigation tasks under headphones listening to either speech in quiet or speech in noise. During the layout task, participants were asked to judge the perceived similarity of two or more presets by dragging presets into clusters that were judged to sound similar, while those that had different sound qualities were moved farther apart. The process of creating personalized layouts was used to derive an optimized layout that would provide the most intuitive navigation of the control space for the individual with hearing impairment. Participants would then navigate through their personalized control space or a predetermined control space. Similarity and sound quality preference were assessed in a paired-comparison task in which the presets and the navigated parameter sets were compared.
Analysis of the results of these experiments showed that most subjects were consistently able to navigate to similar fitting settings from trial to trial, suggesting that desirable sound quality preferences could be achieved via this type of system. Figure 2 shows the average preference rating for the presets and the preferred settings average across listeners. The first 9 bars represent the presets that were used to construct the control, and the final 8 bars represent the listeners’ preferred settings obtained via the navigation task. Participants overwhelmingly preferred hearing aid settings achieved via the navigation task over the predetermined parameter sets that were used to construct the control space.
Data indicated that listeners do not need to organize the control space themselves so long as they are provided with a well-organized layout. Creating a well-organized control space is difficult to do when there are 64 different parameters to consider. If the layout does not provide perceptual consistency throughout the control space, then it will be difficult for the patient to successfully navigate throughout the space and make sense of the sound quality changes they are hearing. Accordingly, in order to streamline the SoundPoint process, an optimized layout was derived that could be applied to hearing aid fittings for a wide range of audiograms.
Our desire to create a fitting method that optimizes the hearing aid experience has paralleled changes in hearing instrument technology. While hearing aid technology and prescriptive fitting approaches have increasingly improved the clinicians’ ability to meet the audibility and intelligibility requirements of our patients, we are beginning to appreciate the effect that individual difference plays on satisfaction and outcome effectiveness.
SoundPoint represents an empirically based, intuitive, logical, and effective method of hearing aid optimization that recognizes and leverages individual variability while allowing the patient to be involved in their own fitting—a true patient-centered approach to hearing aid fitting.
- Valentine S, Dundas JA, Fitz K. Evidence for the use of a patient-centered fitting tool. Hearing Review. 2011. In press.
- Watson NA, Knudsen VD. Selective amplification in hearing aids. J Acoust Soc Am. 1940;11:406-419.
- Lybarger S. US Patent Application SN 532, 278; 1944.
- Davis H, Hudgins CV, Marquis RJ, et al. The selection of hearing aids. Laryngoscope. 1946;56:85-115,135-162.
- Radley WG, Bragg WL, Dadson RS, et al. Hearing Aids and Audiometers. Report on the Committee on Electroacoustics. London: His Majesty’s Stationery Office; 1947.
- Sammeth CA, Levitt H. Hearing aid selection and fitting in adults: history and evolution. In: Valente M, Hosford-Dunn H, Roeser R, eds. Audiology Treatment. New York: Thieme; 2000:213-259.
- Carhart R. Selection of hearing aids. Arch Otolaryngol. 1946;44:1-18.
- Ross M, Plant G. Aural rehabilitation: Some personal & professional reflections. Hearing Review. 2001;8(9):62-67.
- Berger K, Hagberg E, Rane R. Determination of hearing aid gain. Hear Instrum. 1979;30:26-34, 44.
- Libby ER. The 1/3-2/3 insertion gain hearing aid selection guide. Hear Instrum. 1986;37:27-28.
- McCandless G, Lyregaard P. Prescription of gain and output (POGO) for hearing aids. Hear Instrum. 1983;35:16-21.
- Byrne D, Tonisson W. Selecting the gain of hearing aids for persons with sensorineural hearing impairments. Scand Audiol. 1976;5:1-59.
- Seewald RC, Ross M. Amplification for young hearing impaired children. In: Pollack M, ed. Amplification for the Hearing Impaired. New York: Grune & Stratton Inc; 1988:213-267.
- Dillon H. NAL-NL1: A new prescriptive fitting procedure for non-linear hearing aids. Hear Jour. 1999;52(4):10-16.
- Cornelisse LE, Seewald RC, Jamieson DG. The input/output formula: a theoretical approach to the fitting of personal amplification devices. J Acoust Soc Am. 1995;97:1854-1864.
- Etymotic Research. Fig6. Hearing Aid Fitting Protocol. Operating Manual. Elk Grove Village, Ill: Etymotic Research; 1996.
- Moore B, Glasberg B, Stone M. Use of a loudness model for hearing aid fitting. III. A general method for deriving initial fittings for hearing aids with multi-channel compression. Br J Audiol. 1999;33:241-258.
- Shapiro I. Hearing aid fitting by prescription. Audiology. 1976;15:63-173.
- Cox R. The MSU hearing instrument prescription procedure. Hear Instrum. 1988;39(1):6-10.
- Skinner M, Pascoe D, Miller J, Popelka G. Measurements to determine the optimal placement of speech energy. In: Studebaker G, Bess F, eds. The Vanderbilt Hearing Aid Report. Upper Darby, Pa: Monographs in Contemporary Audiology; 1982:161-169.
- Allen JB, Hall JL, Jeng PS. Loudness growth in 1/2-octave bands (LGOB)—a procedure for the assessment of loudness. J Acoust Soc Am. 1990;88(2):745-53.
- Van Vliet D. A comprehensive hearing aid fitting protocol. Audiology Today. 1995;7:11-13.
- Hawkins D. A historical perspective on hearing aid selection. Hearing Review. 1994;1(10):9-12.
- Kuk F, Ludvigsen C. Variables affecting the use of prescriptive formulae to fit modern nonlinear hearing aids. J Am Acad Audiol. 1999;10:458-465.
- Keidser G, Dillon H. What’s new in prescriptive fittings down under? In: Palmer C, Seewald R, eds. Hearing Care for Adults 2006: Proceedings of the First International Adult Conference. Chicago: Phonak AG; 2006:133-142.
- Keidser G, O’Brien A, Carter L, McLelland M, Yeend I. Variation in preferred gain with experience for hearing-aid users. Int J Audiol. 2008;47:621-635.
- Zakis J, Dillon H, McDermott J. The design and evaluation of a hearing aid with trainable amplification parameters. Ear Hear. 2007;28:812-830.
- Mueller H, Hornsby W, Weber J. Using trainable hearing aids to examine real-world preferred gain. J Am Acad Audiol. 2008;19:758-773.
- Wessel DL. Low dimensional control of musical timbre. In: Proceedings of the Audio Engineering Society Convention; 1978;59.
- Momeni A, Wessel D. Characterizing and controlling musical material intuitively with geometric models. In: Proceedings of the Conference on New Interfaces for Musical Expression; National University of Singapore; 2003.