This section shares benchmarking and related research information from inside Starkey Hearing Technologies, including detailed methodology used in data collection.
Added Stable Gain (ASG)
Added stable gain (ASG) is the difference in a hearing aid's maximum stable gain (MSG) with its feedback cancellation algorithm turned off and its feedback cancellation algorithm turned on. In other words, it is the additional gain available to a hearing aid user when the hearing aid's feedback cancellation algorithm is activated. MSG is the highest amount of gain that can be provided in a hearing aid without audible feedback or degraded sound quality due to feedback oscillation (Johnson et al., 2007; Ricketts et al., 2008). For an in-depth definition and walk-through on measurement of MSG visit the link below.
Traditionally, feedback management systems minimize feedback through gain reduction in specific frequency channels. A key disadvantage to the use of such feedback management systems is the possible compromise to speech audibility that may arise from gain reduction. Feedback cancellation (FBC), on the other hand, minimizes feedback by subtracting a signal that mimics the feedback signal from the input to the hearing aid. One advantage to FBC is that these algorithms can provide additional gain to the user without compromising speech audibility (Ricketts, 2008). Merks et al. (2006) provides an overview of current feedback cancellation technology. The author concludes that not all FBC algorithms are alike. One of several characteristics that can vary significantly from one FBC algorithm to another is ASG. In order for hearing healthcare professionals to make evidence-based decisions when recommending specific products to their patients, accurate characterization of FBC algorithms should be available. Objective measurement methods that are reliable and valid must be utilized to quantify benefit from FBC algorithms.
An example of a clinically viable method of measuring ASG involves gradually raising the broadband gain of a hearing aid until the MSG (with FBC off) is obtained (Johnson et al., 2007; Ricketts et al., 2008). A real-ear aided response is recorded. Then the FBC algorithm is activated and the overall gain is increased until the MSG has been reached, and a second measurement is recorded. The difference in gain between the two measurements is the ASG. This relationship is shown in Figure 1, where maximum stable real-ear aided gain is plotted as a function of frequency for MSG with the feedback canceller activated and feedback canceller deactivated.
A disadvantage to the clinical evaluation of ASG is that MSG with FBC on is not always reached before the maximum output of the hearing aid is reached, resulting in a ceiling effect. A second problem is that it does not use an objective criterion to determine the MSG. In other words, the presence of feedback is subjectively observed by either the hearing aid wearer or the clinician. A third problem with this clinical method is that it requires extensive gain manipulation by the experimenter, which is time consuming and can compromise reproducibility. A fourth problem is that patients may be subjected to annoying feedback and/or excessive gain.
The ASG benchmarking protocol described below was developed in order to address shortcomings of traditional or clinical ASG measurement methods. The benchmarking technique described in this section has the following advantages over traditional methods: 1) it is free from ceiling effects, 2) it can be conducted on a KEMAR, 3) it is mostly automatic, and 4) it does not require measurements at gain settings close to audible feedback. When single number indices of ASG are discussed in literature from Starkey, the largest difference among the collected data, defined as the peak ASG, will be reported.
A KEMAR is placed one meter away from a loudspeaker. The loudspeaker should be at 0 degrees azimuth relative to the KEMAR, and its height should be positioned to be level with the test ear. The KEMAR's artificial ear contains an IEC 711 coupler used for simulation of an average human external ear canal. This coupler uses a medially placed microphone to record the signal in the KEMAR's ear canal. Figure 2 shows a basic schematic of the ASG benchmarking set-up. A computer controls the loudspeaker (which will present stimuli to the hearing aid), hearing aid, and KEMAR microphone. A professional sound card and an external amplifier are used for signal generation and recordings.
In order to simulate a more realistic test environment than an acoustically treated laboratory, the KEMAR is placed in the center of a moderately reverberant room. The KEMAR includes a complete upper torso. It is expected that room reverberation and reflective surfaces such as KEMAR's shoulders will impact the effectiveness of any feedback canceller (Kates, 2008).
The hearing aid is set to linear gain settings that are equal at all frequencies. All other algorithms such as expansion, occlusion manager, directional microphones, noise reduction, etc. are turned off, and the maximum power output (MPO) is set to its maximum setting. If the FBC algorithm requires initialization, it should be initialized while the hearing aid is in the test position on the KEMAR, prior to starting the ASG measurements. The hearing aid is connected to a programmer for making adjustments to the settings without moving the device. The hearing aids are coupled to KEMAR's ear in a manner that allows for MSG with FBC off to fall between 20-30 dB before eliciting feedback. A skeleton earmold with a 3.5mm vent is recommended. Regardless, the same earmold should be used for all measurements with all test hearing aids to avoid variance in slit leakage and earmold acoustics.
The description below details a procedure for the laboratory evaluation of ASG. Additional information regarding this methodology can be found in the patent application: 20070217638 (Merks, 2006). This procedure consists of an impulse response measurement for each of five different hearing aid gain settings, using a white noise stimulus. The impulse response of a hearing aid is its output to a very brief input signal, an impulse. This approach effectively characterizes the gain and output characteristics of the hearing aid as a function of frequency. The five settings are: 1) mute, 2) FBC off at high gain, 3) FBC off at low gain, 4) FBC on at high gain, 5) FBC on at low gain. High gain is defined as a level that is within 10 dB of the MSG for the FBC on test condition. Low gain is at least 10 dB below the high gain value for the FBC on condition.
The following describes the steps for recording the five impulse responses:
- Disable the FBC algorithm
- Increase the overall gain of the hearing aid by 10 dB steps until audible oscillation occurs or until the highest gain of the hearing aid has been reached
- Decrease the gain by 5 dB to stop oscillation
- Mute the hearing aid (note: the mute measurement can be done with any hearing aid settings)
- Record the first impulse response (mute)
- Un-mute the hearing aid
- Record the second impulse response (FBC off, high gain)
- Reduce gain by 10 dB
- Record the third response (FBC off, low gain)
- Activate the FBC algorithm
- Increase the overall gain in 10 dB steps until oscillation occurs or the highest gain is reached
- Decrease gain by 5 dB to stop oscillation
- Record the fourth response (FBC on, high gain)
- Decrease gain by 10 dB
- Record fifth response (FBC on, low gain)
A computer program monitors the signal-to-noise ratio (SNR) of the response in KEMAR's ear canal to make sure it is sufficient for the measurements. The stimulus level is automatically varied to enhance the SNR as well. The stimulus level is adjusted so that the output of the hearing aid is near 90 dB SPL.
The response recorded in KEMAR's ear canal has three components: the hearing aid output, the direct signal path (the signal that enters the ear through the vent or slits around the earmold), and feedback (inaudible except possibly to a trained listener). By performing the measurement at different gain settings of the hearing aid, the feedback component can be distinguished from the other two, and MSG is estimated as a function of frequency. ASG is then calculated as the difference between MSG with FBC off and MSG with FBC on.
Figure 3 shows an example of three recordings out of the five measurements in the protocol. In the first (top) panel, the impulse response is recorded with the hearing aid muted. This recording is the measurement of the direct signal path and characterizes the amount of sound, or acoustic leakage that passes around the earmold. The second panel of Figure 3 is the measurement with the hearing aid turned on and programmed to a low gain setting. In this second panel, three acoustic events are observed. The first is, again, the impulse response as it passes around the hearing aid into the ear canal of KEMAR. The second peak in the recording is the impulse signal amplified by the hearing aid and presented into the ear canal of KEMAR. The third, and for the low gain setting, much smaller response is the first feedback response. The third panel of Figure 3 shows the same response peaks as the previous panels. However, in this recording configuration the hearing aid gains have been set much higher, eliciting a stronger feedback response.
The most valuable component of these data is the relationship between the peaks of the first feedback response observed in Figure 3. The difference in this peak response across low and high gain settings illustrates how effective a feedback cancellation system will be at reducing feedback. If the feedback cancellation algorithm is effective, the peak of the first feedback response under high gain conditions should not be significantly higher than that of the low gain condition. In the example above, the first feedback response is significantly larger when the hearing aid is in the high gain condition compared to the low gain condition. Therefore, the feedback cancellation algorithm is not effective at the higher gain settings. This relationship between the first feedback levels is analyzed for feedback canceller on and off test conditions.
Figure 4 shows the results of the ASG analysis in Starkey S Series hearing aids. The plot shows maximum stable gain as a function of frequency for a behind-the-ear hearing aid. The green dotted line is the hearing aid response at MSG in the FBC off test condition. The blue solid line is the hearing aid response at MSG for the FBC on test condition. The difference between these two curves is the ASG.
ASG Maximum Stable Gain (MSG) (dB) with Feedback Cancellation (FBC) on and off as a function of frequency (kHz) for a Starkey S Series hearing aid.
The methodology reported here summarizes an objective benchmarking protocol for measuring ASG. While clinical evaluation of ASG may show wide variability across patients these measures offer an opportunity for clinicians to evaluate the performance of numerous feedback cancellation algorithms and can be completed in a few short minutes only requiring a comparison between two real-ear measurements of MSG.
Banerjee, S. (2006). Active Feedback Intercept: a state-of-the-art algorithm, Starkey Laboratories, Inc. Hearing Research & Technology white paper.
Kates, J. M. (2008). Digital Hearing Aids. California: Plural Publishing.
Merks, I., Starkey Laboratories, Inc. (2006). Method and Apparatus for Measurement of Gain Margin of a Hearing Assistance Device, U.S. Pat. App. 20070217638.
Merks, I., Banerjee, S., & Trine, T. (2006). Assessing the effectiveness of feedback cancellation in hearing aids. Hearing Review, 13(4), 53-54, 56-57.
Ricketts, T. A., Johnson, E. E., & Federman, J. (2008). Individual differences within and across feedback suppression hearing aids. Journal of the American Academy of Audiology, 19(10), 1-24.
Johnson, E. E., Ricketts, T. A., & Hornsby, B. W. Y. (2007). The effect of digital phase cancellation feedback reduction systems on amplified sound quality. Journal of the American Academy of Audiology, 18(5), 404-416.