Hearing Aids

9.1.1

Early Model Predictions on Speech Understanding

Here some of the pioneering work on the understanding of speech in noise conducted by Reinier Plomp and colleagues in the 1970s and 1980s will be presented. This discussion will apply to the hard of hearing and the elderly, the problems they face with speech understanding in noise, and the limited benefit they may experience from HAs in these conditions. This work may seem outdated, but the reader is reminded that the principles still apply today, and not only for HAs but also for cochlear implants (see chapter: Cochlear Implants ). I start with a quote from arguing the need for his approach:

Our insights into why hearing-impaired people appear to be so seriously handicapped in everyday listening situations seem to be very scanty. This lack of knowledge particularly manifests itself in the uncritical way in which hearing aids are assumed to be of benefit. Since most conductive defects in the transmission chain up to the cochlea can nowadays be successfully rehabilitated by means of surgery, the great majority of the remaining inoperable cases are sensorineural hearing impairments. Although it is generally recognized that electronic amplification cannot compensate satisfactorily for these losses, it is remarkable how much hearing-aid prescribers expect from careful selection and fitting followed by good training. On the other hand, many hearing impaired appear to be rather disappointed about their hearing aids.

This was amplified 30 years later by :

There are two difficulties that accompany hearing loss: loss of ability to hear quiet sounds (usually accompanied by a much smaller loss of ability to hear stronger sounds), and loss of ability to understand speech, especially speech in noise. The two losses often but not always go together: either loss can occur without the other. Someone who cannot even detect quiet speech may be able to understand almost as well as someone with normal hearing at an extremely noisy party, while another person who can still detect quiet speech may be unable to understand speech in the presence of noise at any presentation level.

approach to this problem provided a scheme for modeling the observed differences between normal hearing (NH) and hearing-impaired persons. This approach is as relevant today as it was in 1978 and may be applied to the understanding of speech perception with HAs. The effects of hearing loss are well illustrated by Fig. 9.1 , where the speech-reception threshold (SRT) is plotted against the level of the interfering noise. The SRT is the average A-weighted (see chapter: Hearing Problems ) speech level at which 50% of two-syllable words are repeated correctly by the listener. These days mostly dB SPL, instead of dB(A), is used, but that hardly makes any difference. The figure shows a reference curve for NH persons, and curves for four groups of people with SNHLs ranging from 25 to 40 dB. The hearing loss results in an increasing SRT in quiet (0 dB(A) masking). The linear rising part of the NH reference curve (for levels >30 dB(A)) shows that, over a large range of noise levels, the SRT corresponds to a constant SNR of typically –5 dB. This becomes about 0 dB for people with 40 dB hearing loss.

Plomp’s model uses two parameters to characterize the SRT–noise level curves ( Fig. 9.2 ). Parameter A (for attenuation) is equal to the hearing loss as determined by the pure-tone audiogram and is mainly responsible for the substantially higher speech levels required by the hearing impaired at low noise levels. At higher noise (and speech) levels, well above the elevated hearing threshold (i.e., in the linear rising parts of the curves), there remains a difference between the various curves. This is quantified by the parameter D (for distortion), indicating that hearing-impaired persons typically require a better SNR for achieving the 50% correct score. According to Plomp’s framework the D -term, also called hearing loss for speech in noise, reflects the main problem in speech communication for the hearing impaired. As background noise is common in daily life, HAs are only of limited benefit in compensating for the underlying distortion caused by the hearing loss.

considered two (idealized) classes of hearing impairment: hearing loss of class A with attenuation in quiet and hearing loss of class D , comparable with a speech deficit in noise. Hearing loss of class A (□) shifts the normal SRT curve ( Fig. 9.2 ) by the amount of hearing loss in quiet (in this example by 30 dB) but approaches the normal curve at higher background noise levels. The listener with hearing loss would need a 50 dB(A) SRT for noise levels up to 40 dB(A), slowly increasing to SRT=60 dB(A) for 65 dB(A) noise. We see that this represents a considerable SRT loss in quiet but nearly normal SRTs at a 60-dB(A) background noise level. The pure class D (▲) represents a parallel upward shift (10 dB in Fig. 9.2 ) of the normal SRT–noise level curve. So this represents a minor loss of SRT for normal speech levels (~65 dB(A)) in quiet, but a substantial handicap above a 60-dB background noise level, unless the speech is substantially amplified. The more realistic combination of both class A and D (■) shows a substantial loss of speech understanding both in quiet and in noise.

In line with these findings ( ), it is clear that audibility (e.g., the result of amplification) cannot explain the less than optimal speech recognition of people with severe losses listening at high sensation levels. The data from suggested that audibility as quantified by the speech intelligibility index (SII) over-predicts speech performance at high sensation levels for listeners with severe hearing losses. The SII also underestimated speech scores at low sensation levels in many cases. The SII is computed as SII( f )=(SNR( f )+15)/30, here f stands for the various frequency bands. As the total range of SII (0≤SII≤1) corresponds to 30 dB, every 3 dB increase in the D -effect means a decrease of 10% in SII.

Perhaps the most important consequence of the decline in hearing sensitivity with aging is difficulty in understanding speech. The distortion factor D ( Fig. 9.3 ) increases sharply with age and adds to the problems of loss in hearing sensitivity (cf. Fig. 9.1 ). In 140 male subjects (20 per decade between the ages 20 and 89) and 72 female subjects (20 per decade between 60 and 89, and 12 for the age interval 90 – 96), measured the monaural SRT for sentences in quiet and at four noise levels (22.5, 37.5, 52.5, and 67.5 dB(A) noise with long-term average speech spectra). The data were described in terms of the model shown in Fig. 9.2 . A + D effects increased progressively above age 50 reaching values of 20–40 dB for subjects between 80 and 90 years old ( Fig. 9.3 ). D -effects also increased progressively above age 50 reaching values of 5–10 dB for subjects between 80 and 90. However subjects with the same hearing loss for speech in quiet may differ considerably in their ability to understand speech in noise. Thus, SRT in quiet is a poor predictor of SRT in noise, indicating that the SRT in noise should also be measured when a good picture of a person’s hearing ability is required. The data confirm that the hearing handicap of many elderly subjects manifests itself primarily in a noisy environment. Noise levels in rooms used by the aged must be 5–10 dB lower than those for NH subjects for acceptable speech perception.

9.1.2

Age Effects on Aided Hearing in Noisy Environments

The most important reason for dissatisfaction by those with SNR loss in the 10–15 dB range is their resultant inability to understand speech in the presence of noise at restaurants, parties, etc. Some report that they take off their hearing aids in loud social situations where audibility is not a problem but their inability to separate speech from noise (typically interfering with speech) is.

measured intelligibility for nonsense syllables in modulated noise as a function of modulation frequency for young and elderly (clinically) NH listeners. Speech recognition in interrupted noise was poorer for older than younger subjects (compare ; ; Fig. 9.3 ). Small age-related differences were observed in the decrease in score with interrupted noise relative to the score without interrupted noise. Not only elderly, but also middle-aged listeners often complain about difficulties with conversation in social settings, even when they have normal audiograms ( ). investigated whether early aging influences an individual’s ability to communicate in everyday settings. They found that age affects which auditory evoked potential component predicts communication performance in reverberant conditions. Whereas in younger adults, envelope cues (as reflected in the auditory steady-state response) predicted performance, in middle-aged listeners reliance relied heavily on the temporal fine structure (as measured in the frequency-following response), which is more disrupted by reverberant energy than temporal envelope structure is. assumed that hearing loss results in an unbalanced neural representation of speech: The slowly varying envelope is enhanced, dominating representation in the auditory pathway and perceptual salience at the cost of the rapidly varying fine structure. Under that assumption they envisioned to ameliorate this by auditory–cognitive training to reduce the emphasis on the speech envelope in older adults (ages 55–79) with NH and with hearing loss. They found that the group with hearing loss experienced a reduction in the neural representation of the speech envelope presented in noise, approaching levels observed in NH older adults. No changes were noted in the NH group.

combined pupillometry, which measures processing load as reflected by pupil dilation ( ), with an SRT task and could show that the SNR affected processing load as reflected by changes in the pupil dilation response. They observed that the pupil response was larger in the single-talker masker conditions than in the fluctuating noise conditions. They suggested that this reflects increased processing load evoked by semantic interference during the perception of speech, independent of intelligibility level. assumed that the benefit of HAs is not primarily reflected in better speech performance, but that it is reflected in less effortful listening in the aided than in the unaided condition. They measured pupil dilation in 32 NH participants while listening to sentences masked by fluctuating noise or interfering speech at either 50% and 84% intelligibility. concluded that better cognitive abilities not only relate to better speech perception, but also partly explain the ability to carry the higher processing load in complex listening conditions.

A similar approach was followed by who stressed that: “Age-related problems in understanding spoken language are exacerbated by perceptual stressors such as noise and by cognitive stressors such as memory load.” Signal-processing technologies in HAs, designed for older adults, have to include not only improved audibility but also reducing stress on the listener during this information processing. Consequently, argued that: “Long-standing approaches to rehabilitative audiology should be revitalized to emphasize the important role that training and therapy play in promoting compensatory brain reorganization as older adults acclimatize to new technologies.” also noted that: “When the listening conditions are matched so that it is as difficult for younger adults to identify individual words as it is for older adults, apparent age-related declines in cognitive performance on measures of memory, attention, and discourse comprehension largely disappear.” found: “Evidence to suggest that perceived social support is a significant predictor of satisfaction with hearing aids.”

9.1.3

Effects of Hearing Aids on Sound Localization

In their study on the evolution of human hearing, concluded that high-frequency sensitivity in mammals evolved principally as a consequence of the marked improvement it provided in sound localization (see chapter: Hearing Basics ). Besides high-frequency hearing, localizing spatially discrete acoustical sources in a noisy world is the most effective means of suppressing noise by the spatial filtering provided by binaural hearing. Such a filter passes acoustical signals and noise from the general direction of the sound source, while attenuating most of the noise power from other directions ( ). High-frequency hearing loss is very detrimental for sound localization based on interaural loudness differences. In order to preserve the ability to localize sound based on interaural time (phase) differences, the sound levels at both ears should be comparable also in the case of hearing loss, hence the need for bilateral HAs. investigated the effect of various commercial HAs on the ability to resolve front-back confusions and on sound localization in the frontal horizontal and vertical plane. Corroborating the role of audibility across a wide frequency range, they found that “hearing-impaired subjects reached the same performance with and without the different hearing aids, if in the unaided condition, a frequency-specific audibility correction was applied.”

9.1.4

Hearing Aids at the Cocktail Party

In Chapter 4 , Hearing Problems, we showed that hearing-impaired listeners perform more poorly in cocktail party conditions compared to NH people, and that this problem increases with age. Do HAs help under these conditions? listed a variety of signal processing strategies implemented by HAs may assist the listener with SNHL. Among these are algorithms that implement environmental noise reduction, and thus attenuate unwanted sound sources. In addition there is directional amplification, which emphasizes a source originating from a specific azimuth relative to the head, directly related to improving SNR and enhancing source selection. However, warned that these two forms of noise reduction are effective for certain types of unwanted sounds, however, they do not help the listener in choosing among competing talkers. For that purpose, cognitive factors such as attention become a dominant factor.

had found that spatial release from masking (SRM) with bilateral HAs (mean ~4 dB) was negatively correlated with the amount of hearing loss. With a single HA, SRM was lower (mean ~2.5 dB) and related to the level of the stimulus in the unaided ear. In NH subjects the SRM was on average approximately 10 dB. determined the benefit provided to listeners with SNHL by an acoustic beamforming microphone array, which provides directional amplification, in a speech-on-speech masking experiment. They compared this with bilateral amplification. They found that acoustic beamforming provided a large (mean ~9 dB) spatial release from speech-on-speech masking for SNHL listeners. This is about the same as for natural NH. reported that for most SNHL listeners in the wider masker-signal separation condition, lower thresholds were obtained through the microphone array than through bilateral amplification. suggested that consequently candidacy for highly spatially tuned amplification might depend on performance with conventional bilateral amplification. Especially, individuals with poor performance using natural cues are more likely to benefit from using a beamforming microphone array.

9.4.1