Effectiveness of directional technology in open-canal hearing instruments

After years of advanced algorithm development in an effort to bring more consumers to the hearing aid table, it just might be that the re-emergence of the “open” fit will do more for hearing aid sales than any other attempt of the decade. Open fittings, perse, aren't new. The IROS (ipsilateral routing of signal) style has a 40-plus-year history. That era of opening the canal using a non-occluding earmold, or simply inserting #13 tubing into the ear canal, served to provide some high-frequency gain without the concomitant negative occlusion effects resulting from a closed earmold. The benefit provided was limited, however, because of the unavoidable feedback. The current era provides for adjustable gain in multiple channels, unoccluded sound quality, better cosmetic design, and, perhaps most significantly, good feedback-management schemes (see Mueller in this issue for a review of potential benefits1). One question posed by many practicing clinicians about these open-canal (OC) fittings is whether or not they can provide directional microphone benefit. In this article, we will identify the factors involved in providing directional benefit with an OC fitting. DIRECTIONAL BENEFIT Much has been published over the past 20 years about the benefit that can be expected from a hearing aid employing directional technology. Polar response patterns (or polar plots) are often used to describe the spatial attenuation characteristics of different microphone schemes. To review, a polar pattern displays the angles (typically shown in the horizontal plane) from which sound arrives at the microphone. The concentric circles indicate the amount of sensitivity reduction (in dB) for sounds from any angle. Zero degrees (0°) is represented on the top and indicates the direction that the microphone front port (and presumably the head of the listener) is pointed; 180° is represented on the bottom. While many polar response patterns are obtained with the microphone/hearing aid alone, a better understanding of the nulls (angles of least sensitivity) and lobes (angles of most sensitivity) can be seen when the measures are obtained with the microphone/hearing aid mounted on a head. There are some reasons to question if the directionality of the OC product is similar to that of other BTEs, which we will discuss later. Figure 1 shows frequency-specific polar plots for both the omnidirectional and directional modes of a single OC product, as measured on a Knowles Electronics Manikin for Acoustic Research (KEMAR) in our anechoic chamber. It is apparent that there is no advantage to the directional mode in the lower frequencies, since equal sensitivity is shown for both omnidirectional and directional modes of operation. However, there appears to be considerable advantage in the higher frequencies. Thus, one might conclude that directional benefit, or advantage, is available in an OC fitting, at least when measured in an anechoic chamber.Figure 1: Polar plots obtained at four frequencies on KEMAR in an anechoic chamber. The solid line indicates the omnidirectional mode, the dashed line the directional modeHowever, over 20 years ago, we published evidence that reverberation has a negative impact on this directional benefit.2 We also know that benefit measured in a laboratory is dependent upon the number and orientation of speakers, hearing aid style, baseline comparison, etc.3–5 It also has been shown that degree and configuration of hearing loss are poorly correlated to benefit,6 and that the orientation of the microphone ports and/or talker can have a negative impact.7 Finally, all clinicians know (or should know) that the biggest culprit in directional performance may be the presence of debris in the microphone ports, which can transform a directional polar response pattern into an omnidirectional polar response pattern more quickly than any of the prevailing microphone drift concerns. In all, we have sufficient evidence of the efficacy of the design, or the capability of the microphone system to work in ideal situations. However, effectiveness data are harder to find (refer to Bentler, 20058). Effectiveness refers to real benefit in real environments. Through a series of reports from the Walter Reed group, we know that laboratory data re: benefit don't relate very well to self-report of benefit.9 As we begin to better understand those environments in which this technology provides additional benefit to the hearing aid user, we find that a new level of complexity accompanies the OC style with directional microphones. It must be understood, however, that directivity is in the microphone technology. Whether or not that directivity is realized by the hearing aid user depends on a number of factors, including gain. Loss of low-frequency gain One aspect of OC fittings that is obvious is the reduction of low-frequency gain; the impact of the microphone's directivity due to that loss is not so well understood. Since an open fitting is the extreme of the vent effect, it should be clear that any gain below 1500 Hz will be hard to come by without significant power cost (Steve Armstrong, personal communication). Because the gain is “bled out” (ouch!), the directivity of the microphone cannot be realized in that frequency range. Several manufacturers have addressed this loss by enhancing the high-frequency directivity of the microphone by moving the ports closer together. Thus, it seems possible that we may be “pushing” the directivity into a higher frequency range, assuming the hearing aid has gain in that higher frequency region. Loss of high-frequency gain As many (older) clinicians will attest, feedback problems are second only to occlusion complaints on the list of hearing aid fitting dilemmas. Remaking earmolds, plugging vents, rolling off the high-frequency response, and so on have given way to some fairly sophisticated algorithms for feedback management. But even the best of these may be insufficient for use with open fittings for moderate or greater hearing loss. It has even been suggested that some of the current and proprietary algorithms for gain/output still roll off the high-frequency response to ensure consumer satisfaction with these new products. In an effort to understand this high-frequency impact, we looked at OC gain in several ways. Since many manufacturers are using adaptation, acclimatization, or experience settings to help ease the newer users into amplification and many dispensers use these “first fit” rules, we evaluated a number of these products to determine if the high frequencies are rolled off in excess of prescriptive intent. We set the premier products from four different manufacturers to their “first fit” algorithm for the audiogram shown in Figure 2. Using a KEMAR, we took insertion gain measures (see Figure 3) for each of four adjustment settings for one of the products. It is apparent that gain increases across frequencies with the first change in setting; more apparent is the low level of gain in even the most experienced user setting.Figure 2: Audiogram used to program open-canal hearing aidsFigure 3: Insertion gain measured on KEMAR with speech stimulus for four settings of experience as provided by the softwareFor comparison, the insertion gain prescribed by NAL-NL1 is shown in the same figure. A comparison of the “first fit” algorithm results for all four manufacturers' products is shown in Figure 4 for the same audiogram. It is apparent that the high-frequency insertion gain (for the “first fit”) differs significantly across these products. High-frequency gain is not uniformly low, but it does fall significantly below NAL-NL1 targets.Figure 4: Insertion gain measured on KEMAR for “first fit” algorithms across four manufacturersSNR due to direct pathway It has been suggested that directional microphones cannot be effective in open fittings due to the noise entering the unobstructed ear canal. That is a valid concern, but only when we consider the interaction of input spectrum and level, gain, and compression. As shown in Figures 5a–d, the low-level input (in this case, a speech-shaped noise) provides for a signal-to-noise ratio (SNR) that should be favorable to the listener. As the input level is increased, and the compression reduces the amplified gain incrementally, the direct pathway can potentially overtake the aided pathway, resulting in a poorer SNR for the listener.Figure 5: Time waveforms of the output from an open-canal fitting (measured on KEMAR) shown separately as the direct (red) and amplified (black) pathways. Input levels for the four panels are: (a) 55 dBA, (b) 65 dBA, (c) 75 dBA, (d) 85 dBAIn an effort to understand this interaction we looked at both SNR for the listener and change in audibility of the intended signal as a function of input. For these measures, we placed KEMAR in the calibrated center of an IAC booth. Speech stimuli (CST passages)9 were presented from a speaker at 0° azimuth (on-axis) and at ear level. Noise was presented from six speakers located at the top and bottom of the other three corners of the booth and pointed toward the KEMAR (off-axis). Levels from the six speakers were calibrated to be equivalent to each other, and summed to be equivalent to the on-axis output. For the Noise Reduction Index (NRI) measurements, we used the methodology of Hagerman, Olofsson and Nasten10 to extract the spectra of speech and noise at the input and output while the hearing aid was set in both directional and omnidirectional modes. We used speech-shaped noise as background interference that was always set to 2 dB below the speech level (for a constant +2 SNR). Figure 6 shows the difference between the omnidirectional and directional SNR (or NRI11) as a function of input. It is apparent that the NRI difference decreases as a function of input level, a direct result of the direct pathway increasing in a linear fashion while the hearing aid pathway increases in a non-linear manner as compression has more impact on the higher-level inputs.Figure 6: Directional benefit calculated as the difference in Noise Reduction Index11 for omnidirectional and directional modes on a single model of open-canal style. Measures were taken on KEMAR and are shown as a function of speech level (noise set at −2 dB re: speech)Figure 7 shows the impact of using different types of noise on audibility, as measured with the Speech Intelligibility Index (SII).12 The spectrum of the speech signal was measured in both the omni and directional modes, as were the spectra for random noise in four shapes: flat, +6-dB/octave, −6-dB/octave, and speech-shaped. The figure shows the audibility difference between the omnidirectional and directional modes as a function of level. It is obvious that regardless of the noise shape, the directional benefit (calculated as SII difference between the omnidirectional mode and the directional mode) increases for moderate inputs and then decreases again (as with Figure 6).Figure 7: Directional benefit calculated as the difference in Speech Intelligibility Index (SII) for omnidirectional and directional modes on a single model of open-canal style. Measures were taken on KEMAR and are shown as a function of speech level (noise set at −2 dB re: speech)The apparent discrepancy between high benefit (for NRI) and low benefit (for SII) at low-level inputs can be explained based on audibility alone. For the SII calculations, there is little difference in the two microphone modes for the softer inputs. This likely occurs because the audibility of the speech signal for the given hearing loss is low and the difference between the two microphone modes is not distinguishable. For the NRI calculations, the low-level input is amplified above the direct pathway input; the mid- and high-level inputs are subject to compression effects whereby the direct pathway grows in a linear fashion, while the amplified pathway does not. We used still another approach to quantify the impact of using a directional microphone with an open fitting.13 Four unvoiced consonants (s, f, sh, th) were paired with four vowels (/i/, /ai/, /a/, /u/) resulting in final stimuli of /isi/, /ifi/, etc. Six loudspeakers were positioned at ear level and placed at a distance of 1.25 m from a KEMAR, with a 60° difference in azimuth angle between any two adjacent speakers. The central axis of the loudspeaker was oriented to directly face KEMAR. The speech stimuli were presented at 65 dB SPL from the primary speaker (at 30° azimuth relative to KEMAR's nose). Jet noise was played from the other five speakers with the combined level at 65 dB SPL. Recordings were obtained on KEMAR through the hearing aid in both omni and directional modes. The SNR of the formants (for the four vowels) and spectral peaks (for the four consonants) relative to the noise was determined using Adobe Audition (v 1.0) as follows: The root-mean-square (RMS) levels at these frequency regions were measured for the signals recorded from the hearing aid on KEMAR. The RMS levels of the frequency regions adjacent to these formants/ spectral peaks, defined as the noise, were measured. The effective SNR for the formant/ spectral peak was defined as the difference in dB between the RMS level of formants or spectral peaks and the RMS level of noise. Statistical analysis (using Mann-Whitney U test) suggested that there were no significant differences for the SNRs measured for these two microphone conditions. However, a trend for improved SNRs in fixed-directional mode over omnidirectional mode was apparent. The measures have been made on one product to date. This trend may prove significant for other products, especially those including enhanced directional-microphone technology. SUMMARY We still have a lot to learn about optimizing the open-canal style, especially with respect to directional-microphone use. To date, we do know: Directivity is in the microphone, and the open fit reduces the effectiveness of the directional-microphone scheme in the low frequencies, due to the loss in gain. High-frequency directivity is accessible as long as gain is available. Because of the direct input pathway, the SNR at the ear of the listener is related to (some combination of) noise type and level, speech level, and compression characteristics. It appears that maximum directional benefit occurs for low and moderate inputs. As clinicians, we also know that many patients are reacting very positively to this new design. Obtaining a clearer understanding of the potential benefits through continued research can only enhance the acceptance of both the technological advantages and limitations.