Most clinicians have a standard set of tests they perform to validate hearing aid fittings. These may include real-ear measures or speech-in-noise tests. A set of non-standard tests are often performed as well. Almost every clinic has a coffee can filled with nuts and bolts, a teacup and spoon or a heavy metal paper weight. These items are all used for one purpose: to generate a loud bang or transient impulse to test the wearer’s aided tolerance for such sounds.
Transient sounds consist of an intense pulse of acoustic energy rising well above the long-term average of the surrounding environment. They are typically broadband and always brief (transitory) in duration.1,2 Aside from the intensity and brevity of transients, they are also characterized by a leading edge that rises extremely rapidly.1
The traditional approach for controlling transients in a hearing aid is through the use of a transient limiter or peak clipper. Peak clippers are used because they are instantaneous; they are fast enough to react appropriately to the quickly rising leading edge of the transient. Traditional automatic gain control (AGC) circuits with attack times of 8-10 ms are far too slow. However, a peak clipper, being instant, will work for any extremely intense transient.
Unfortunately, peak clipping also has limitations that render it inadequate for many common transients. Peak clipping generates considerable harmonic and intermodulation distortions that degrade sound quality and speech perception. Furthermore, peak clipping can only occur when the hearing aid is at or near saturation. There are many transients such as dishes clattering, keys jingling, car doors shutting, or many children’s toys that hearing aid wearers find very uncomfortable. 3 However, many are not nearly intense enough to reach the peak clipper. In fact, the rapid rise time of the transient—and not just the overall sound pressure level—often causes the annoyance.4,5 Hence, peak clipping is insufficient to relieve the annoyance caused by many every day transient signals.
Reducing the annoyance of transients or any noises in general is important for three reasons, as cited by Dornic and Laaksonen6:
First, annoying sounds reduce the over-all quality of human life. Second, annoyance should be seen as a warning of threat to health. Third, annoyance can be seen as a mediating factor in the effects of noise performance.6
Problems in Controlling Transients
Controlling transients using traditional approaches has proven ineffective in the past. Most hearing aid circuits adapt too slowly. Those that are fast enough generally create distortion and are reserved only for very high intensity inputs.
To effectively control all transients and provide a comfortable listening experience that maintains good sound quality, the control system must achieve at least three objectives:
1) Instantly detect and control the leading edge of the transient.
2) Never mistake speech for an undesirable transient and therefore never negatively impact speech perception or quality.
3) Transient control must be adaptive providing proportionately larger reduction for more intense transients and little or no reduction for weak transients. Thus, a natural sounding relationship is maintained between transients of any intensity with the surrounding acoustic environment.
Once these three conditions are met, the perceived sound quality of the hearing aid goes up dramatically just as the annoyance from transients is reduced.
A New Approach
AntiShock is a new feature for the hearing aid industry with a detection and correction algorithm that restores normal loudness perception for transient impulses. AntiShock operates within the time frame at which the transients actually occur. Therefore, it neither distorts sound quality nor affects speech perception.
A characteristic transient modulation consisting of a rapidly rising amplitude spike on the order of microseconds triggers antiShock to react (Figure 1).
Figure 1 demonstrates three relevant points about antiShock:
• AntiShock responded instantly to each transient without the overshoot that would occur if an AGC circuit had been used.
• Bearing in mind that a knife on a cutting board is not an intense transient, the signal level of the leading edge has been reduced proportionately. The sound was neither eliminated nor distorted, merely proportionately reduced.
• After the leading edge passes, there is a slight undershoot which momentarily alters the shape of the waveform relative to the original. However, the duration of the undershoot is less than 60 ms, shorter than the duration of the damped trailing edge of the transient.
Never affects speech. Some transients do occur in normal speech. For example, plosives such as /k, t, d/ have a transient burst at the release of air. No speech should be impacted by antiShock, but these phonemes are particularly vulnerable to misdetection due to their transient nature and duration which is less than 60 ms.
Figure 2 shows an example of processed and unprocessed speech to demonstrate the impact (or lack thereof) of antiShock on speech. Ignoring the speech for a moment, observe the amplitude and duration of the leading edge of the initial transient in Track 2. It rises to its peak amplitude in under 1 ms. The duration of the most intense portion of the pulse (colored) is approximately 15 ms. After the first 15 ms the amplitude of the transient is quite significantly damped but still large relative to the speech that follows it. In Track 3 the amplitude of the transient is still larger than the following speech, but it has only been reduced enough to reestablish a correctly proportioned relationship between the two. In other words, the transient is still audible, and it still sounds like a knife on a cutting board.
Otherwise, except where there is a transient present, Tracks 1, 2 and 3 are virtually identical. AntiShock controls the leading edge of the transient just enough to reduce, but not remove it. The algorithm releases before the trailing end of the transient to maintain normal perception of the knife sound relative to the speech that follows. The complete release occurred just under 80 ms after the leading edge of the transient passed.
Adaptive control. When the controlling algorithm is adaptive, a balanced perception of transient to speech will be maintained for all shock sounds. If it is static, it may not be strong enough to control very intense transients (eg, a door slamming) or it may be overly aggressive for soft transients (eg, a pen tapping on a table). The goal is to allow the hearing aid wearer to perceive a car door slamming or a pen tapping just as a normal-hearing person would. Neither sound should be distorted, completely suppressed, or missed. Earlier examples showed how well antiShock controls a moderately intense chopping sound.
The flexibility of antiShock to more selectively control soft transients is demonstrated in Figure 3. The morphology of the waveforms passed through the hearing aid look slightly different than the original, for all the same reasons mentioned previously. When Track 2 is compared to Track 3, the speech signal is entirely unaffected by antiShock. Furthermore, even the pen tap is passed with very little effect. AntiShock causes a very minor reduction of the leading edge for about 2 ms. Otherwise, the pen tap and speech are passed with no effect at all.
Field Trials on User Preference
Figures 1-3 demonstrate that antiShock performs as desired for speech, transients, and speech plus transients. However, it does not show if hearing aid wearers will prefer this new feature.
To answer this question, a blind comparison was undertaken. A total of 30 participants with varying degrees of hearing loss were fitted with multiple shell styles of Element hearing aids. They listened to recorded speech and recorded transients separately, then together. They heard each recording with antiShock turned on and turned off in randomized order. Then they informed the tester as to which they preferred. They could give one of three responses that amounted to a preference for antiShock on, antiShock off, or no difference. The results are shown in Figure 4.
The paired comparison results agree with the acoustic measurements shown in the earlier figures. When listening to speech alone, 23 of the 30 participants had no preference (77.6%); they could not tell whether antiShock was on or off. Five people preferred it on (16.7%), and only two liked it off (6.7%).
However, in the other four conditions where transients were presented either alone or with speech, there was a strong systematic preference for antiShock to be on. Between 20 and 22 people preferred antiShock to be on in the presence of a car door slamming or a knife chopping (66.7%-73.3%). Only 2 people preferred it off (6.7%), with 4 people in the case of speech plus the car door (13.3%). A small minority had no preference. This example suggests that antiShock does meet the three criteria for an adaptive transient limiter, and provides superior performance in the presence of transients.
A second, multi-site study (Rochester, NY, and Oldenburg, Germany) was undertaken to further test the efficacy of, and obtain more detailed information about, antiShock. To date a total of 23 participants across both sites have been tested and some of the preliminary results are reported here.
Each participant provided preference comparisons for antiShock on or off (ie, activated or deactivated). In general, the number of comparisons was exhaustive and the data is still somewhat preliminary. Therefore, only the most important dimension of antiShock, that of annoyance, will be reported here. Figure 5 shows the outcome when transients alone were presented without the context of speech or noise. The results were the same for all four signals from the softest (pen taps) to the loudest (car door).
The participants from both sites overwhelmingly preferred antiShock on. The trends were consistently the same for the two locations. Furthermore, the preferences for antiShock activated were consistent across listening environments, including transients, transients plus speech, and transients plus noise.
In other words, there is a substantial reduction in the perceived annoyance caused by the hearing aid when antiShock is engaged. This perception persists even in the presence of the most annoying sounds to which most people are exposed. The preference for antiShock in the presence of transients is consistent and substantial demonstrating a measurable reduction in annoyance.
Summary
Properly controlling transients in a hearing aid requires that three conditions be met. The algorithm must instantly detect and suppress the leading edge of the annoying transient without interfering with the clarity and quality of speech. The algorithm must also be adaptive enough to bring any transient into the range of proportionately normal loudness relative to the surrounding speech environment. Acoustic analysis and subjective annoyance evaluations on hearing aid wearers demonstrates that antiShock meets all three requirements. AntiShock provides a very natural sound quality that is both clear and comfortable without distorting the perception of speech or the target transient.
References
1. Henderson D, Hamernik RP. Impulse noise: Critical review. J Acoust Soc Am. 1986; 80(2):569-84.
2. Voigt, P., B. Godenhielm, and E. Ostlund, Impulse noise—measurement and assessment of the risk of noise induced hearing loss. Scandinavian Audiology. 1980. Suppl 12:319-25.
3. Hellstrom PA, Dengerink HA, Axelsson A. Noise levels from toys and recreational articles for children and teenagers. Brit Jour Audiol.1992; 26(5):267-70.
4. Fidell S, Silvati L, Pearsons K. Relative rates of growth of annoyance of impulsive and non-impulsive noises. J Acoust Soc Am. 2002; 111(1 Pt 2):576-85.
5. Hiramatsu, K., K. Takagi, and T. Yamamoto, Experimental investigation on the effect of some temporal factors of nonsteady noise on annoyance. Jour Acoust Soc Am. 1983. 74(6):1782-93.
6. Dornic S, Laaksonen T. Continuous noise, intermittent noise, and annoyance. Perceptual and Motor Skills.1989. 68(1): p. 11-18.
This article was submitted to HR by Donald Hayes, PhD, Director of Audiology, Unitron Hearing, Kitchener, Ontario. Correspondence can be addressed to Donald Hayes, PhD, Unitron Hearing, 20 Beasley Drive, Kitchener, ON N2G 4X1, Canada; e-mail: [email protected] |