The Real-World Benefits and Limitations of Active Digital Feedback Cancellation

Design considerations of an active feedback cancellation system that may result in the least amount of undesirable processing artifacts were recently published in The Hearing Review.¹ Such a mechanism allows a hearing instrument wearer to potentially receive approximately 10 dB more usable gain from a digital instrument compared to devices without such a mechanism under the same wearer conditions (eg, earmold/shell leakage). One benefit of this type of system is the ability to meet a prescriptive gain target more easily across more frequencies, resulting in more consistent audibility and a fuller, richer sound picture under the same or larger earmold/venting condition. Another outcome from being able to use a larger vent is a reduction of the perception of occlusion and an improvement in perceived sound quality.²

Furthermore, intermittent feedback from leakage of the hearing instrument/earmold during talking, chewing, or proximity to reflective surfaces (eg, a telephone placed close to the ear) may be minimized. Such a mechanism may be especially helpful to some elderly individuals whose hearing instruments work their way out of the ear because of their soft skin texture and their straight ear canals. It may also be helpful in pediatric fittings where the growing ear gradually loosens the fit between the earmold and the child’s ear canal. These potential benefits could increase the satisfaction of hearing instrument wearers with the devices.

Despite such theoretical benefits, the effectiveness of a feedback cancellation system also depends on the listening conditions and on several parameters related to the individual fitting. Because the real-world practical benefits provided by a feedback cancellation system are limited to an average of 10 dB, feedback may still occur for some wearers, and some still may not enjoy a close match to their desired gain targets.

On the opposite end, some wearers may not notice any real-life advantages provided by this 10 dB increase in available gain. Rather than attributing such observations to the failure of the feedback cancellation system, the unmet expectations may lie with characteristics of the wearers, eg, degree of hearing loss, ear canal characteristics, the listening environments, and a lack of complete understanding of the feedback cancellation action.

This article provides several examples, using the Widex Senso Diva digital instrument, to illustrate for whom a digital feedback cancellation algorithm will provide benefit and when its actions may be unnecessary and/or insufficient.

Origins of Feedback
As mentioned in the previous article,¹ feedback occurs when the insertion gain of the hearing instrument exceeds the available gain before feedback (or simply available gain) of the instrument. This means that, at any given frequency, there is a gain limit above which whistling occurs.

Figure 1 reviews the stages involved in the occurrence of feedback. Although the feedback path is defined by static measures, such as the amount of venting and openness of the hearing aid/earmold, the feedback path also changes dynamically during daily activities, like chewing or moving close to reflective surfaces. These occasional changes in the feedback path lead to changes in the available gain in one or more frequencies, and may cause intermittent feedback during many activities.

Figure 2 shows the attenuation characteristics of a BTE hearing instrument with (light line) and without (dark line) a telephone handset held close to the ear. Note the dramatic changes in the feedback path characteristics between the two conditions. A versatile feedback cancellation system should account for both the static requirement and dynamic changes to the feedback path.

Example of Feedback Cancellation
In the case of the Diva Feedback Cancellation (FBC) system, there are two parts: a slower component and a fast component. The slower feedback path simulator (FPS) continuously estimates the feedback path characteristics (frequency, magnitude, and phase) in order to cancel the feedback signal. This estimation process may take as long as 10 seconds if a sudden change in the feedback path occurs. The exact time for this estimation depends on how closely the new feedback signal matches the previous estimation of the feedback signal. A fast component, the dynamic cancellation optimizer (DCO), limits the gain of the hearing instrument while the FPS is estimating the new feedback path characteristics. The DCO does not, however, reduce the applied gain. Rather, the two components correct for the rapid and slow changes in the feedback path characteristics and increase the available gain by approximately 10 dB.

The sequence of actions is summarized in Figure 3. Figure 3a shows what may happen when the available gain suddenly becomes lower than the desired insertion gain. This would have led to feedback. Figure 3b shows that the DCO provisionally limits the gain of the hearing instrument at the feedback frequencies to stop the feedback. Meanwhile, the FPS adapts to the new feedback path in order to cancel the feedback. The net result is shown in Figure 3c, where the available gain is now 10 dB more than the case where a feedback cancellation system is unavailable. Because the available gain is now more than the insertion gain, feedback is eliminated.

Figure 3 shows that feedback may still occur if the action of the FBC mechanism fails to keep the available gain before feedback curve above the insertion gain curve. This situation may occur if the desired gain exceeds the available gain by more than the 10 dB offered by the feedback cancellation mechanism. This situation could arise when the required insertion gain is relatively high. Furthermore, it should be noted that feedback might still occur if the feedback path changes faster than the corrective action of the feedback cancellation system, or if the magnitude of the changes in the feedback path requires more than the 10 dB increase in available gain. The following examples illustrate some of these cases.

What Feedback Cancellation Can and Cannot Do
Table 1 includes numeric examples to show how the degree of hearing loss of the wearer may interact with the hearing instrument to affect the real-world benefits of the feedback cancellation system. These examples show how the extra 10 dB available gain may provide static and dynamic benefits, and how this 10 dB may be sufficient, insufficient, and unnecessary. For simplicity, the following examples assume that the feedback path remains the same during the FBC process and that we consider the condition at one frequency only. One should be aware, however, that in practice feedback occurs at multiple frequencies depending on the listening situation.

Example 1—Typical Fitting. In this example, the wearer requires a target insertion gain (TG) of 40 dB at a particular frequency, and the hearing instrument can provide a maximum insertion gain (HAG) of 50 dB at that frequency. This suggests that the hearing instrument can meet the wearer’s gain requirement. Let us assume that, because of the use of a moderate size vent, the available gain with FBC-off (AGFBC-off) is only 35 dB. This may have been done to minimize the occlusion effect or to increase the physical comfort of wearing the chosen aid, ie, open fit. Insufficient AGFBC-off may also be a result of a loose fitting or an imprecise impression of an ear canal that has undergone surgery, eg, total mastoidectomy.

Because the FBC system provides 10 dB more available gain, AGFBC-on becomes 45 dB and feedback does not occur. This is an example of a static benefit offered by the FBC. Because the gain target is reached and no feedback occurs, this is also an example of sufficient benefit provided by the FBC system.

Example 2—Looser Fit or Larger-Than-Required Vent. This example involves a fit that is much looser, or one in which a larger vent is used, such that AGFBC-off becomes 25 dB instead of 35 dB. This results in an AGFBC-on of only 35 dB. In this case, the static benefit moves the gain closer to the target; however, it is insufficient to reach the gain target.

Example 3—Gradually Loosened Fit and/or Growing Ear Canal. An elderly individual who has a soft skin texture (and also a straight ear canal) often finds that the hearing aid works way out of the ear after it has been worn for some time. This could also apply to the case of a child (eg, <10 years old) whose ear canal increases in size as he/she grows. Assume that the individual still requires a TG of 40 dB from a hearing instrument that has a HAG of 50 dB. At the initial fitting, the AGFBC-off may be 35 dB. With the FBC system on, the available gain becomes 45 dB. The immediate static benefit of the FBC system is sufficient to the wearer.

On the other hand, as the fit of the hearing instrument changes over time, the AGFBC-off may decrease to 30 dB and then to 25 dB. This reflects changes in the seal (or feedback path) of the hearing instrument. Because the AGFBC-off is now lower than the TG of the wearer, feedback occurs after the hearing instrument is worn for some time. However, the FBC system adds 10 dB more available gain so the AGFBC-on becomes 40 dB and 35 dB instead. This suggests that the wearer may be able to use the hearing instrument even though the feedback characteristics have changed gradually. As the feedback path changes even more, ie, to 25 dB AGFBC-off, the action of the FBC system may eventually reach its maximum. This would result in insufficient gain and insufficient benefit for the wearer.

The benefit of the FBC system in this case is not only its ability to provide 10 dB more gain to meet the initial gain target, but rather its ability to maintain a feedback-free condition over a longer period of time as the feedback path changes. This is beneficial in that the earmold/hearing instrument shell may not have to be immediately replaced, eg, growing child, in order to correct for the feedback problem. But, since the change in the feedback path is continuous, one would eventually need to replace/modify the earmold/shell and redo the feedback test to control for the feedback problem. However, time is gained before a new earmold/shell is required.

Table 1. Examples of  Real-Life Situations
Example	TG	HAG	AGFBC-off	AGFBC-on
1	40	50	35	45
2	40	50	25	35
3	40	50	35- 30- 25	45- 40- 35
4	40	50	40	50
5	40	40	35	40
6	40	40	40	40
7	40	30	30	30
8	40	30	25	30
9	20	50	30	40
Table 1. Summary of examples in this article illustrating when the feedback cancellation system (assuming one fixed frequency) may be beneficial to the wearer. Target gain (TG) refers to the desirable maximum insertion gain appropriate for the wearer’s hearing loss. Hearing aid gain (HAG) refers to the potential insertion gain available from the hearing instrument. Available Gain w/ FBC-off (AGFBC-off) is the maximum available insertion gain from the hearing instrument under the current venting/earmold condition in the FBC-off position. AGFBC-on refers to the maximum available insertion gain in the FBC on-position. In this illustration, AGFBC-on is 10 dB higher than ABGFBC-off.

Example 4—Intermittent Feedback. The FBC mechanism provides dynamic benefits as well. Example 4 illustrates the case where the TG of the wearer is 40 dB and the HAG is 50 dB. If the AGFBC-off were 40 dB during the feedback test, the AGFBC-on of 50 dB would not improve the static target fit since the wearer only requires 40 dB of gain. On the other hand, if one remembers that the feedback path can be temporarily modified by the conditions of the wearer’s ear canal (eg, chewing, laughing, near a reflective wall, placing a telephone handset close to the ear), a lower AGFBC-off may occur in real-life, even though it is 40 dB during the feedback test. The advantage of 10 dB more available gain in the AGFBC-on would provide a greater safety margin against intermittent feedback even though the TG of the individual does not require the FBC to be active.

It is well known that despite the absence of feedback, sound quality may be less than optimal if gain on the hearing instrument is close to its feedback limit.3 In this example, sufficient dynamic benefit is provided to the wearer and intermittent feedback or degraded sound quality is minimized in real-life.

Example 5—System Able to Compensate Gain Up to HAG. The previous examples show cases where the hearing aid has sufficient potential gain to add the 10 dB available gain to the FBC-on condition. The benefit may become marginal or nonexistent when the hearing instrument cannot provide the available gain. Example 5 shows the case where the HAG is 40 dB and the AGFBC-off is 35 dB. Although the FBC can provide 10 dB gain, the end result is only a 5 dB increase to the available gain (AGFBC-on) as the maximum gain of the hearing instrument is only 40 dB. In this case, the static benefit is only 5 dB, still allowing the wearer to meet the gain target.

Example 6—FBC Compensates to Maximum HAG. In this case, the AGFBC-off is 40 dB, enough to match the wearer’s TG. Although the FBC can add 10 dB gain over the AGFBC-off condition, gain at AGFBC-on is still 40 dB because the HAG is limited to 40 dB. In this case, the FBC does not increase the available gain for the wearer. No static benefit is offered to the wearer.

However, an active FBC would correct for real-life situations where the available gain decreases momentarily during activities such as chewing and placing telephone close to ear. In this case, there is no static benefit, but dynamic benefit offered by the FBC. The frequency of intermittent feedback should decrease.

Example 7—FBC Unable to Compensate Beyond HAG Limits. This example illustrates a frequently misunderstood benefit of an active feedback cancellation mechanism. In this case, the wearer still requires 40 dB of gain. However, the hearing instrument has only 30 dB available gain. This suggests that regardless of the action of the FBC system, the maximum gain available to the wearer cannot exceed 30 dB. In other words, an active feedback cancellation system cannot provide the wearer any gain that is not available on the hearing aid.

In this case, one would not be able to match the gain target or enjoy the dynamic benefits of a FBC system. There is neither static nor dynamic benefit offered by the FBC mechanism to the wearer. A practical analogy against such thinking is fitting a CIC-style hearing instrument to a person who has a hearing loss greater than 100 dB, and hoping that the FBC would allow use of such a hearing device to match the gain requirement of the wearer.

Example 8—Result Is Still Better Than Not Having FBC. One may not completely discount the benefit of the FBC in the above case, however. For the same case (Example 8), if the AGFBC-off is only 25 dB instead of 30 dB, the action of the FBC-on would result in an AGFBC-on of 30 dB. Although the target gain is still not met, the wearer has at least 5 dB more available gain than wearing a hearing instrument without such a mechanism.

Example 9—When FBC is Unnecessary. The last example shows a case where the benefit of the FBC is unnecessary. Example 9 shows a subject who may require only 20 dB TG from the same hearing instrument that has a HAG of 50 dB. If the result of the feedback test yields an AGFBC-off of 30 dB, this would have matched the target gain, as well as provided a gain cushion against any minor changes in the feedback path during daily activities like chewing and placing telephone over ears. In this case, the wearer does not need or use the AGFBC-on of 40 dB, although it is available to the wearer nonetheless.

Importance of an Accurate Feedback Test
In the above discussion only one single frequency is considered. As mentioned earlier, feedback may occur at several frequencies and these frequencies may change according to the real-life listening situation.⁴ It is therefore insufficient to cancel feedback at one frequency alone, and the Senso Diva feedback cancellation system operates simultaneously over a wide frequency range where feedback may potentially occur.

The accuracy of the feedback test plays a critical role in the performance of a feedback cancellation system. Feedback may still occur in the system if the initial feedback path is inaccurately estimated. If this initial estimation is accurate, feedback that may occur will be canceled immediately. However, if this initial estimation is not accurate, eg, from reflective surfaces that are present during the feedback test, which are not present during typical use of the hearing instrument, the FBC mechanism will need longer time to converge to the new feedback path, and thus intermittent feedback may occur.

If the initial estimate and the actual feedback path are too different, the FPS may not be able to completely cancel the feedback frequency. In such a case it may be left to the DCO to limit gain, which may lead to insufficient gain for the wearer. Thus, it is important that the initial estimation of the feedback path during the feedback test is as accurate as possible. If there is any doubt on the accuracy of the feedback test, a new test should be performed under a more realistic condition.

In summary, the examples presented here show that a FBC system allows more usable gain—the extent determined by how much gain is available on the hearing instrument. The provision of additional available gain may enhance the static and dynamic uses of the instrument.

An example of a static advantage includes the use of a larger vent and a looser fit hearing aid while matching a gain target at more frequencies. Such action may improve overall sound quality while minimizing the occlusion effect and enhancing the physical comfort of the fit. An example of the dynamic advantage is the allowance for sudden and gradual changes in the seal of the hearing instrument during daily activities.

In order to maximize the advantages offered by the FBC mechanism, it is important to understand the wearer characteristics and the hearing instrument characteristics. This is important to set proper expectations for the FBC system. In addition, it may be worthwhile, especially in children, to perform the feedback test regularly in order to keep the estimation of the feedback path as close as possible to the actual feedback path. This is because the feedback path may change gradually over time, leading to the likelihood of more intermittent feedback.

References
1. Kuk F, Ludvigsen C, Kaulberg T. Understanding feedback and digital feedback cancellation strategies. The Hearing Review. 2002;9(2):36-41, 48.
2. Kuk F. Perceptual consequence of vents in hearing aids. Brit J Audiol. 1991;25(3):163-169.
3. Cox R. Combined effects of earmold vents and suboscillatory feedback on hearing aid frequency response. Ear Hear. 1982;39(1):12-17.
4. Agnew J. Acoustic feedback and other audible artifacts in hearing aids. Trends in Amplif. 1996;1(2):45-82.

This article was submitted by Francis Kuk, PhD, director of audiology at Widex Hearing Aid Co, Long Island City, NY, and Carl Ludvigsen, MS, manager of audiology research at Widex ApS, Vaerloese, Denmark. Correspondence can be addressed to Hearing Review or Francis Kuk, Widex Hearing Aid Co, 35-53 24th St, Long Island City, NY 11106-4116; email: [email protected].