Starkey Research & Clinical Blog

Prescribing Compression for Severe Hearing Loss

Souza, P.E., Jenstad, L.M. & Folino, R. (2005). Using multichannel wide-dynamic range compression in severely hearing-impaired listeners: effects on speech recognition and quality. Ear and Hearing 26(2), 120-131.

This editorial discusses the clinical implications of an independent research study. The original work was not associated with Starkey Hearing Technologies and does not reflect the opinions of the authors.

Most modern hearing aids feature multiple signal processing channels and wide dynamic range compression (WDRC). For listeners with mild to moderate hearing loss, WDRC can offer improvement in speech intelligibility and sound quality in quiet conditions (Souza, 2002). Small benefits of WDRC over linear amplification have also been observed in the presence of background noise. (Moore, et al., 1999).  Most of the research on WDRC has been on individuals with mild or moderate hearing loss. There is considerably less data available on WDRC performance for severely hearing-impaired participants. In fact, many clinical practitioners believe that patients with greater amounts of hearing loss prefer, and benefit most, from linear amplification.

Severe hearing loss is accompanied by reduced frequency selectivity (Faulkner, et al., 1990; Rosen, et al., 1990) and temporal resolution (Lamore, et al., 1990; Nelson & Freyman, 1987). Beyond audibility constraints of hearing loss alone, these impairments further limit ability to identify and discriminate speech cues. Because spectral cues may be limited or unavailable, severely impaired listeners rely on other cues such as variations in the speech amplitude envelope over time (Rosen et al, 1990). For these reasons, many compressor designs are constrained to minimally degrade the speech signal.

One particular concern about WDRC circuitry is that natural variations in speech amplitude may be altered, reducing the availability of amplitude-related cues. The result can be degradation in consonant perception (Souza & Turner, 1998) or overall sentence recognition (Souza & Kitch, 2001b; Stone & Moore, 2003; VanTasell & Trine, 1996).  The reduction of amplitude cues, in combination with impaired frequency selectivity, could result in poor performance for severely hearing-impaired individuals using WDRC hearing aids.  Indeed, in a study with severely hearing-impaired participants, Souza and Bishop (1999) found smaller improvements in sentence recognition for WDRC amplification than for linear amplification.  DeGennaro, et al., (1986) also found no advantage with a WDRC system for severely hearing-impaired listeners, despite the fact that the compression system provided improved audibility over the linear system.

In contrast, other studies suggest WDRC benefits for individuals with severe hearing loss. Barker, et al., (2001) found that listeners with severe hearing loss preferred single channel WDRC to either output compression limiting or peak clipping.  Although user preference is a critical element of a successful hearing aid fitting, it is inarguably more important to ensure adequate speech recognition.  While many prescriptive formulas attempt to find a balance, there is no consensus on the most appropriate amplification characteristics to ensure speech recognition and acceptable sound quality for hearing aid users with severe hearing loss.

The purpose of Souza et al.’s study was to examine speech recognition performance and speech quality judgments for severely hearing-impaired listeners using four different types of amplification:

1.  Linear with peak clipping

2.  Linear with output compression limiting

3.  Two-channel WDRC

4.  Three-channel WDRC

Thirteen participants with severe, sensorineural hearing loss and most of who had previous hearing aid experience, participated in the study. Seven participants with normal hearing also participated as a control group.

Speech recognition was evaluated using the Nonsense Syllable Test (NST; Resnick, et al., 1976).  Speech quality judgments were obtained with a paired comparison task, using sentence stimuli from the Connected Speech Test (CST) (Cox, et al., 1987).  For each stimulus pair, participants heard the same sentence processed in two different amplification conditions and were asked to select the one they preferred. Listeners were specifically instructed to avoid using loudness as a primary criterion for their preference.

Speech materials were process offline with a master hearing aid simulation of each amplification type described above, signal presentation was at 70dB SPL via an ER-2 insert earphone.  Frequency and gain response was determined based on the mean audiometric thresholds of the participant group.  The targets themselves represented an average of the NAL-RP (Byrne et al., 1990) and NAL-NL1 (Dillon, 1999) targets for conversational speech.

As would be expected, the speech recognition scores for normal-hearing participants were high and equivalent for all test conditions.  Hearing-impaired participants showed poorer performance for peak clipping and multichannel WDRC conditions than for compression limiting. The only statistically significant difference was between compression limiting and 3-channel WDRC.

A more detailed feature analysis was conducted to determine which speech features were poorly transmitted by the 3-channel WDRC system, including place, voicing and manner cues. Place cues were not transmitted effectively by any of the amplification systems, probably because the hearing-impaired participants’ had poor frequency resolution and place is mainly transmitted by spectral cues (Rosen, 1992).  The two WDRC amplification types preserved voicing information slightly more than compression limiting or peak clipping.  Of primary interest was manner of articulation, as it is primarily transmitted via amplitude envelope cues (Rosen, 1992).  There is an expectation that amplitude envelope is more distorted by fast-acting WDRC than either peak clipping or output compression limiting.

Detailed analysis of manner transmission revealed that amplification type affected phoneme categories differently. Fricatives were fairly well preserved by compression limiting and WDRC systems but were less well transmitted by the peak clipping system.  The 2-channel WDRC system transmitted nasality better than peak clipping or compression limiting, but the 3-channel WDRC did not preserve nasality cues as well. Because nasality is primarily transmitted via the low-frequency nasal murmur (Kent & Read, 1992) the channel crossover in the 3-channel system may have disrupted this cue.  Affricates, a combination of a stop and fricative, were best preserved by linear amplification – peak clipping or compression limiting – and were negatively affected by WDRC. The authors surmised that poor affricate identification may have been related to the time parameters of their WDRC algorithm, causing amplitude decreases after the stop to be perceived as bursts, resulting in misidentification of affricates as stop consonants.

Speech quality ratings by the hearing-impaired participants showed a clear preference for compression limiting. There were significant differences between all comparisons with the exception of peak clipping versus two-channel WDRC. No normal-hearing participants preferred peak clipping and their preferences were equally divided among the other options.

Overall, the results of this study indicate poorer speech recognition and sound quality for WDRC compared to compression limiting. Previous work indicated that compression benefit decreased as pure tone thresholds exceeded 70dBHL (Goedegebure, et al., 2001; Souza & Bishop, 1999), and Boothroyd et al., (1988) concluded that amplitude distortions were responsible for poor performance with a two-channel, fast-acting compression system.  These assumptions may be supported by the current work, in which stop-affricate confusions were considered to be related to misinterpretation of amplitude cues.

Though compression limiting yielded better speech recognition and sound quality in the present study, the authors caution that this should not preclude the use of WDRC systems with severely hearing-impaired individuals. For example, they acknowledge that the short release times they used may have affected amplitude envelope cues. Because of the wide variability in parameters among current WDRC circuits, there will be variability in their effect on speech envelope cues as well. Preservation of speech amplitude envelope is likely to aid word recognition and clarity for users with severe loss and may be achieved with longer attack and release times (Kuk & Ludvigsen, 2000).  Alternatively, shorter attack and release times, combined with lower compression ratios may also reduce the detrimental effects of WDRC on amplitude cues.

Souza and her colleagues point out that the frequency responses they used in the study may have deviated from individually prescribed targets, compromising performance. They applied their prescriptive formula to the average hearing loss of their participant population, which may not have provided as close a match to target as would be achieved in a typical hearing aid fitting. In general, their frequency responses over-amplified slightly in the low frequencies and under-amplified in the high frequencies.  They used a fixed compression ratio of 3:1 across channels for all participants, higher than those prescribed by prescriptive procedures for clinical use; additionally, the use of one ratio across channels does not represent the typical prescription.  These deviations from clinically prescribed settings may have affected the speech recognition scores and sound quality preferences in their study.

The hearing-impaired individuals in this study clearly preferred the sound quality of compression limiting over the peak clipping or WDRC options. The improved audibility of high-frequency information with the WDRC systems may have adversely affected sound quality for some listeners, especially those not accustomed to high frequency emphasis. Clinicians are well aware that when individuals wear new hearing instruments with better high-frequency amplification, their initial description of the sound quality is “tinny” or “metallic”. It can take days or even weeks for them to adjust to the new frequency response, before the increased high frequency audibility is received favorably.

Similarly, many experienced hearing aid users react negatively to WDRC because it sounds softer than the linear amplification they have become accustomed to. Participants in this study were specifically instructed to disregard loudness in their sound quality judgments, but the relative decrease in loudness in the WDRC trials may have affected their preferences anyway. For experienced users of powerful hearing instruments, a decrease in loudness can be interpreted initially as a decrease in performance, despite accompanying improvements in speech recognition ability.

While educative, these findings are based on atypical compressor parameters; thus, applying them directly to clinical decision making should be done with caution. Beyond the considerations of gain and time constants, sound quality judgments based on everyday performance might be very different, especially after use over a longer period of time.  Hearing aid users are known to experience a period of acclimatization to new hearing aids, the duration of which varies depending on their degree of loss and prior user of amplification (Keidser et al., 2009).  It is reasonable to assume that a WDRC circuit providing superior high-frequency audibility, though initially perceived as “too soft” or “too tinny”, would sound more natural and clear after an extended period of consistent use.

Souza and her colleagues address an important issue for clinicians fitting individuals with severe to profound hearing loss.  As they acknowledge, more study is needed regarding specific WDRC parameters and how they affect speech recognition and preferences in the severely hearing-impaired population.  Of particular interest is how WDRC parameters perform over time in everyday listening situations, outside of the laboratory.  Findings, such as these, illustrate how objective and subjective measures reveal discrete aspects of a patient’s experience, proving information that helps clinicians determine the most appropriate starting point for these individuals.

References

Barker, C., Dillon, H. & Newall, P. (2001). Fitting low ratio compression to people with severe and profound hearing losses. Ear and Hearing 22, 130-141.

Byrne, D., Parkinson, A. & Newall, P. (1990). Hearing aid gain and frequency response requirements for the severely-profoundly hearing impaired. Ear and Hearing 11, 40-49.

Cox, R.M., Alexander, G.C. & Gilmore, C. (1987). Development of the Connected Speech Test (CST). Ear and Hearing 8 (Suppl.), 119S-126S.

DeGennaro, S., Braida, L.D. & Durlach, N.I. (1986). Multichannel syllabic compression for severely impaired listeners. Journal of Rehabilitation Research and Development 23, 17-24.

Dillon, H. (1999). NAL-NL1: A new prescriptive fitting procedure for non-linear hearing aids. Hearing Journal 52, 10-16.

Faulkner, A., Rosen, S. & Moore, B.C.J. (1990). Residual frequency selectivity in the profoundly hearing-impaired listener. British Journal of Audiology 24, 381-392.

Humes, L.E., Christensen, L., Thomas, T., Bess, F., Hedley-Williams, A. & Bentler, R. (1999). A comparison of the aided performance and benefit provided by a linear and a two-channel wide dynamic range compression hearing aid. Journal of Speech, Language and Hearing Research 42, 65-79.

Keidser G., O’Brien, A., Carter, L., McLelland, M. & Yeend, I. (2008). Variation in preferred gain with experience for hearing-aid users. International Journal of Audiology, 47(10), 621-35.

Kent, R.D. & Read, C. (1992). The acoustic analysis of speech. San Diego: Singular Publishing Group.

Kuk, F. & Ludvigsen, C. (2000). Hearing aid design and fitting solutions for persons with severe-to-profound losses. Hearing Journal 53, 29-37.

Lamore, P.J., Verweij, C. & Brocaar, M.P. (1990). Residual hearing capacity of severely hearing-impaired participants. Acta Otolaryngologica Supplement 469, 7-15.

Moore, B.C.J., Peters, R.W. & Stone, M.A. (1999). Benefits of linear amplification and multi-channel compression for speech comprehension in backgrounds with spectral and temporal dips. Journal of the Acoustical Society of America 105, 400-411.

Nelson, D.A. & Freyman, R.L. (1987). Temporal resolution in sensorineural hearing-impaired listeners. Journal of the Acoustical Society of America 81, 709-720.

Resnick, S.B., Dubno, J.R., Hoffnung, S. & Levitt, H. (1976). Phoneme errors on a nonsense syllable test. Journal of the Acoustical Society of America 58 (Suppl. 1), 114.

Rosen, S. (1992). Temporal information in speech: Acoustic, auditory and linguistic aspects. Philosophical transactions of the Royal Society of London Series B: Biological Sciences 336, 367-373.

Rosen, S., Faulkner, A. & Smith, D.A. (1990). The psychoacoustics of profound hearing impairment. Acta Otolaryngologica Supplement 469, 16-22.

Souza, P.E. (2002). Effects of compression on speech acoustics, intelligibility and sound quality. Trends in Amplification 6, 131-165.

Souza, P.E. & Bishop, R. (1999). Improving speech audibility with wide dynamic range compression in listeners with severe sensorineural hearing loss. Ear and Hearing 20, 461-470.

Souza, P.E., Jenstad, L.M. & Folino, R. (2005). Using multichannel wide-dynamic range compression in severely hearing-impaired listeners: effects on speech recognition and quality. Ear and Hearing 26(2), 120-131.

Souza, P.E. & Kitch, V.J. (2001b). The contribution of amplitude envelope cues to sentence identification in young and aged listeners. Ear and Hearing 22, 112-119.

Souza, P.E. & Turner, C. W. (1998). Multichannel compression, temporal cues and audibility. Journal of Speech and Hearing Research 41, 315-326.

Stone, M.A. & Moore, B.C.J. (2003). Effect of the speed of a single-channel dynamic range compressor on intelligibility in a competing speech task. Journal of the Acoustical Society of America 114, 1023-1034.

VanTasell, D. J. & Trine, T.D. (1996). Effect of single-band syllabic amplitude compression on temporal speech information in nonsense syllables and in sentences. Journal of Speech and Hearing Research 39, 912-922.