Ear & Hearing recently published a nice paper sponsored by Oticon on the additional benefits of speech envelope preservation, but you need to wade into it before you find “The results indicate that envelope fidelity is a primary factor in determining the combined effects of noise and binary mask processing for intelligibility and quality of speech presented in babble noise,” as the article title, Relationship Among Signal Fidelity, Hearing Loss, and Working Memory for Digital Noise Suppression, is slightly misleading. Of course, Nina Kraus’ “Brainvolts” lab at Northwestern has its’ fingerprints all over this: There’s a lot of Really Good Stuff coming out of her lab.
UPDATE: We added reading span test resources in Bootnote D (12/13/2015)
We at The Hearing Blog have been banging the proverbial drum for the benefits of hearing aid speech envelope preservation for some time now; and since this new journal article is quite relevant we’re adding it to our roster on the topic; and also a discussion afterwards.
Relationship Among Signal Fidelity, Hearing Loss, and Working Memory for Digital Noise Suppression (Abstract)
This study considered speech modified by additive babble combined with noise-suppression processing. The purpose was to determine the relative importance of the signal modifications, individual peripheral hearing loss, and individual cognitive capacity on speech intelligibility and speech quality.
The participant group consisted of 31 individuals with moderate high-frequency hearing loss ranging in age from 51 to 89 years (mean = 69.6 years). Speech intelligibility and speech quality were measured using low-context sentences presented in babble at several signal-to-noise ratios. Speech stimuli were processed with a binary mask noise-suppression strategy with systematic manipulations of two parameters (error rate and attenuation values). The cumulative effects of signal modification produced by babble and signal processing were quantified using an envelope-distortion metric. Working memory capacity was assessed with a reading span test.D [Reference added ~Ed. ] Analysis of variance was used to determine the effects of signal processing parameters on perceptual scores. Hierarchical linear modeling was used to determine the role of degree of hearing loss and working memory capacity in individual listener response to the processed noisy speech. The model also considered improvements in envelope fidelity caused by the binary mask and the degradations to envelope caused by error and noise.
The participants showed significant benefits in terms of intelligibility scores and quality ratings for noisy speech processed by the ideal binary mask noise-suppression strategy. This benefit was observed across a range of signal-to-noise ratios and persisted when up to a 30% error rate was introduced into the processing. Average intelligibility scores and average quality ratings were well predicted by an objective metric of envelope fidelity. Degree of hearing loss and working memory capacity were significant factors in explaining individual listener’s intelligibility scores for binary mask processing applied to speech in babble [Emphasis ~Ed.]. Degree of hearing loss and working memory capacity did not predict listeners’ quality ratings.
The results indicate that envelope fidelity is a primary factor in determining the combined effects of noise and binary mask processing for intelligibility and quality of speech presented in babble noise. Degree of hearing loss and working memory capacity are significant factors in explaining variability in listeners’ speech intelligibility scores but not in quality ratings [Emphasis & link added ~Ed.].
Kathryn Arehart & James Kates at CU-Boulder, Pam Souza at Northwestern’s Auditory Neuroscience Lab, and Thomas Lunner & Michael Syskind Pederson at Oticon
The Hearing Blog explains:
From our previous article about hearing aid speech envelope detection titled “Speech Envelope Detection vs AGC Attack & Release in Hearing Aids,” we discuss the fundamentals of where it’s needed, why it’s needed, and what to do to enable it for the patients who need it.
Related speech envelope preservation coverage:
- Individuals With Hearing Loss Process Sound Differently In Noisy Environments;
- Speech Envelope Detection vs AGC Attack & Release in Hearing Aids;
- More On Speech Envelope Preservation Benefits For Limited Working Memory;
- First Person Report: Not Everyone Can Adjust to Digital Hearing Aids
After A Lifetime Of Analog(ue);
- Temporal envelope and fine structure cues for speech intelligibility,
by Rob Drullman (Abstract | PDF)
First off, Edgar Villchur‘s Wide Dynamic Range Compression (WDRC) is a wonderful innovation for the vast number of people who have hearing loss, and is probably responsible for cutting the number of hearing aid credit returns in half as it vastly increases patient satisfaction. In fact, many hundreds of thousands (perhaps well into the millions) of people were helped with the K-AMP hearing aid circuit, which was the grounbreaking analog WDRC hearing aid circuit developed by Mead C Killion in 1988. However, by the very nature of the beast with its fast AGC attack & release times, WDRC has a tendency to “squash” the speech envelope, hence this article series.
We have identified at least four classes of people where this fast attack/release is a detriment to speech discrimination, though:
1) Auditory Neuropathy Spectrum Disorder (ANSD), or more precisely, those people who fall into the auditory dys-synchrony part of the spectrum, usually due to missing inner hair cells: Because of the distortion brought on by the neural firing loss of synchrony, these patients lack the temporal resolution to detect the fine structure of speech, and instead fall back on envelope detection as an adjunct to lipreading. This lack of temporal resolution can be seen in gap detection times: Instead of them being in the 4-9 msec range across the speech range, they can be in the hundreds of milliseconds. For a more complete discussion of this mechanism, start with Auditory Neuropathy Spectrum Disorder and Hearing Aids: Rethinking Fitting Strategies1 by Susan Spirakis PhD, who also presented this at the Auditory Neuropathy Spectrum Disorder Conference 2012 in March;
2) Along the lines of ANSD, people who have significant cochlear dead zones in the mid-to-high frequencies, as detected in the Threshold Equalizing Noise (TEN) test,2 would intuitively appear to be unable to resolve the fine structure of speech: More research into this needs to be performed;
3) Profoundly deaf people, especially prelingually deafened adults, absolutely detest WDRC, 8, 9 as they too lack temporal processing ability, 10 and instead use envelope detection as an adjunct to speechreading. ReSound has it right in the Sparx (DanaLogic 6090 for NHS), where WDRC can be disabled with linear amplification used in its place; and full linear programming has been extended to their Enzo, LiNX², and Enzo² hearing aids (but not to the 1st generation LiNX sold at Costco). That being said, although we’re now fitting the Sparx on profoundly deaf people with excellent results, we have yet to enable WDRC, as the patients actually understand speech better with these instruments running in linear mode. For much more on this, please see “First Person Report: Not Everyone Can Adjust to Digital Hearing Aids After A Lifetime Of Analog(ue);”
4) There’s some tantalizing new research coming out of Nina Kraus’ “Brainvolts” Lab at Northwestern, noted for their neurobiology angle to audiology, by Pamela Souza PhD that correlates working memory ability and long term memory to AGC attack/release times. From 20Q: Cognition Measures -They Might Change the Way You Fit Hearing Aids 3
10. You said something about working memory being important in cases where a particular type of hearing aid processing might not be appropriate. What did you mean by that?
I’m glad you asked, because I think this is one of the most interesting and potentially valuable reasons to think about working memory. There is evidence in the research literature that working memory affects the way an individual may respond to hearing aid processing. Let’s focus on two different aspects of hearing aid processing: fast-acting wide dynamic range compression (WDRC) and frequency lowering. Both of these are sometimes used in digital hearing aids, and both offer potential for improved speech-sound audibility and recognition. But both also alter the acoustic properties of the signal in a way that may not be suitable for every patient.
11. OK, I’m interested. I sometimes fit products that have a fast compression release time. Should I be thinking about working memory when I do?
First, let’s think about what a short release time (with a low compression kneepoint) will do to speech. The hearing aid gain is adjusting very quickly, so the hearing aid improves consonant audibility but at the expense of more alteration of the signal. The longer the release time, the more similar the amplitude variations (or “envelope”) of the output signal will be to unprocessed speech. To understand where working memory comes in, consider one study in which the experimenters measured working memory, attention and reaction time.11 They fit their patients with two-channel WDRC with two different compression settings: a fast WDRC with a 40 ms release time and a slow WDRC with a 640 ms release time. Patients wore each compression setting for 10 weeks and speech recognition was measured in unmodulated and modulated noise. For the modulated noise, patients with better working memory did better with fast-acting WDRC. In fact, the cognitive score explained 39% of the variance in performance, while the amount of hearing loss explained only 3%! So, this really supports the idea that in a complex listening task (in this case, the noise is modulated so that you need to “glimpse” the signal of interest through the noise) and when fast WDRC is altering the signal, cognition matters. That reinforced earlier work which showed a similar relationship between cognitive ability and release time.12, 13, 14
A second point is the definition of the temporal envelope. We adopted the Hilbert Envelope, although many studies use a method of rectification and low–pass filtering. The Hilbert envelope has the clear advantage that it accurately follows all amplitude modulations of a frequency band, running smoothly over the actual waveform. As a consequence, any modification of the envelope can be performed without keeping unwanted (original) temporal modulations intact. In this way one can precisely control the envelope cues transmitted to the listeners.
A note on the protection and extraction of the speech envelope in a processed signal: There are several ways to preserve it, including slow AGC attack & release times, as Starkey did up to their S series; and using the Hilbert transform,4, A as Med-El uses in their Fine Hearing CI stim.5 Oticon aids also have contained several generations of speech envelope preservation, with their older “Speech Guard” and newer “Speech Guard S” processing. B One competitor’s lab testing revealed the older Speech Guard processing to be slow AGC attack & release; however the newer Speech Guard S may indeed be a switch to the Hilbert transform. In addition to Oticon, William Demant’s Bernafon division calls it “Speech Cue Priority.”
Research is piling up showing speech envelope preservation is a vital part of the signal processing chain for hearing aids as it can provide surprising benefit when properly applied. However, just like with different frequency lowering techniques, the preservation algorithms used by Oticon (WDRC with Hilbert transform), Starkey (slow AGC), and ReSound (linear response) differ, and hence would somewhat be aimed at different groups of patients; with the caveat of wireless connectivity for improved S/N thrown into the mix for patients in Red Flag Matrix quadrants II, III & IV.
- Auditory Neuropathy Spectrum Disorder and Hearing Aids: Rethinking Fitting Strategies by Susan Spirakis PhD: Hearing Review, Fall 2012 edition;
- Threshold Equalizing Noise (“TEN”) test, by Prof Brian CJ Moore (ret.), Cambridge University Auditory Perception Lab;
- 20Q: Cognition Measures -They Might Change the Way You Fit Hearing Aids by Pamela Souza PhD with Gus Mueller;
- Hilbert Transform, Wolfram Mathworld
- How cochlear implants encode speech. Jay T Rubinstein, Current Opinion in Otolaryngology & Head and Neck Surgery 2004,
- Temporal envelope and fine structure cues for speech intelligibility. Journal of the Acoustical Society of America; 1995 Jan;97(1):585-92. Rob Drullman (Abstract | PDF)
- AudiologyOnline Course #23839 (Live), #24805 (Recorded): Understanding and Managing Severe Hearing Loss Tuesday, August 12, 2014 at 1600 UCT (12:00 pm EDT). Course presenter: Pamela Souza PhD.
- Using multichannel wide-dynamic range compression in severely hearing-impaired listeners: effects on speech recognition and quality: Pamela Souza PhD, LM Jenstad, and R Folino. Ear and Hearing 26(2), 120-131 (2005)
- Severe Hearing Loss – Recommendations for Fitting Amplification, by Pamela Souza PhD. AudiologyOnline, January 19, 2009
- Prescribing Compression for Severe Hearing Loss, by Jason Galster in the Starkey Evidence Blog, April 1st, 2012;
- Interactions between cognition, compression, and listening conditions: effects on speech-in-noise performance in a two-channel hearing aid. Lunner T1, Sundewall-Thorén E. Journal of the American Academy of Audiology; 2007 Jul-Aug;18(7):604-17;
- Linear and nonlinear hearing aid fittings — 1. Patterns of benefit. Gatehouse, S., Naylor, G., & Elberling, C. (2006): International Journal of Audiology, 45, 130-152;
- Linear and nonlinear hearing aid fittings — 2. Patterns of candidature. Gatehouse, S., Naylor, G., & Elberling, C. (2006); International Journal of Audiology, 45, 153-171;
- Stuart Gatehouse: The International Perspective. Dianne J. Van Tasell, PhD and Harry Levitt, PhD. Trends in Amplification; 2008 Spring; 12(2): 80–84. [Full article courtesy of SAGE Publications.]
B) While training on “Brain Hearing” at Oticon’s palatial New Jersey headquarters, your humble editor auditioned a pair of Alta2 Pro aids, and we could clearly detect the “pop” from the speech envelope when Speech Guard S was enabled. When we asked a certain high-ranking official if they upgraded from slow AGC times to Hilbert transform, we got sort of a dirty look, like we uncovered some deep, dark secret. Although impressed with the sound quality, and can see using it for a few patients, we were disappointed this important audiological feature was only in the top-of-the-line Alta2 Pro and Nera2 Pro versions. What’s more, since we place a premium on wireless connectivity to improve the signal-to-noise ratio, especially for patients who have a >12dB Acceptable Noise Level (ANL; more) score and/or >7dB QuickSIN score (the Red Flag Matrix), we really like the ReSound wireless aids instead, as the Mini Mic solves many problems before they start, without having to resort to an awkward streamer;
C) Anyone who operates a single-sideband suppressed carrier ham radio transceiver since the 1950’s will instantly recognize the Hilbert transform, as that is how the carrier is modulated. For more on this, the reader is directed to Single Sideband, SSB Modulation; and click the Next>> link for more. Also, see the nice graphic here (be sure to peek at the “Anecdote” section, too).
D) A good explanation of the reading span test can be found on the Cognitive Fun! website here, with their version of the test here; and another version from Pitt here. Scholarly explanations & uses can be found here~