When fitting any modern digital hearing aid we are able to specify differing gains (and outputs) for soft speech, medium speech, and loud speech. These “level dependent” filters are a hallmark of modern hearing aid technology. There is some clinical controversy here in that one manufacturer may consider loud speech to be 75- 80 dB SPL and another 85 dB SPL. However, the result is similar- more gain is specified for soft level inputs than for high level inputs. This is in accordance with the loudness growth functions (and corresponding equal loudness contours) and were first delineated by Fletcher and Munson (1933) and also by Sivian and White (1933).
The equal loudness contours (that specify more required gain for softer level inputs) have been replicated with all sorts of stimuli over the years and for the most part appear to be stimulus-independent. One can recognize the use of pure tones as stimuli by the 0 phon curve. Unlike other curves on the equal loudness contours, the 0 phon curve is really one of detectability (yes or no) and not really a judgment based on equal loudnesses in the same sense that the other contours are derived. If a pure tone is used in a minimal audible field (MAF) paradigm, take a look at 1000 Hz. It’s not 0 dB SPL but 3 dB SPL . Puretones have a crest factor of 3 dB – the difference between their instantaneous peak intensity and their RMS value- so the peaks are detected at +3 dB SPL relative to their RMS averages. The RMS average may be 3 dB SPL for a 0 phone MAF curve, but that is only because the peak of the pure tone stimulus is a 0 dB SPL)…. But back to the issue at hand.
So, is soft-medium-loud music the same as soft-medium-loud speech? Initially one would say, of course- what a silly question. The absolute levels of the speech would be lower than those for music, but the relative differences should be the same. After all, the equal loudness contours were established with different stimuli, and except for some “artifactual” differences found in the 0 dB phon curve, the differences are negligible. And for those with a sensorineural hearing loss and perhaps a pathological loudness growth function, it is reasonable to assume that the way that the compression is set for the speech (in quiet) program, should also be set in any music program. After all, the setting of the compression parameters is based more on the nature of the cochlear damage than on the nature of what’s coming into a hearing aid per se. This was actually felt to be the case, and Chasin and Russo (2004) performed a calculation to demonstrate this. Davies-Venn, Sousa, and Fabry (2008) showed that this was indeed the case clinically.
But let’s take a step back (and this would be a really neat Capstone project waiting to be done), and look at how speech changes in intensity versus that of music. Is the setting of compression in a hearing aid just a cochlear phenomenon or is there an additional attribute of the input signal that needs to be assessed. And I should point out that at this point, after having seen hundreds (if not thousands) of hard of hearing musicians over the past 30+ years, I am still not sure whether we should alter the level dependent nature of the compression circuitry for music from the one we selected for speech.
What is it in speech that changes as the speech changes from soft to loud? The obstruents certainly do not change- these are the siblants, affrictates, fricatives, and stop plosives. In speech, the difference between a quiet /s/ and a very loud /s/ is negligible. The difference lies in the lower frequency sonorants only- the vowels, the nasals, and the liquids (/l/ and /r/). These sonorant sounds have most of their energy at their formant locations and for the first two formants (F1 and F2) are generally below 2200 Hz. So loud speech is characterized by the growth in the lower frequencies relative to the higher frequency sounds. In a multi-channel hearing aid (where the higher frequency sound energy is treated differently than the lower frequency sounds) the difference between soft and intense speech is merely a low frequency growth in the input to the hearing aid- or at least it should be. I suspect that different manufacturers implement this change in different ways and use differing fitting philosophies.
In contrast, instrumental music as an input to a hearing aid would have a substantially different behavior. When a stringed instrument is played louder, there is an almost uniform increase across the spectrum. If the fundamental is played 10 dB more intense (e.g., at A[440 Hz]) then the upper frequency harmonics are also played at approximately a 10 dB more intense level. The spectrum of quiet string music has the same shape as the spectrum of intense string music. This is certainly in contrast to the low frequency increase of louder speech.
What about woodwinds? If one were to measure the spectral energy of my clarinet (or any other reeded woodwind) one would find that as the music transitioned from quiet to intense, the lower frequency fundamental energy wouldn’t change all that much- perhaps several decibels at most- but the higher frequency harmonic structure would increase substantially. This is directly related to the non-linear behavior of the reed. This is also something that any clinician can assess with their real ear measurement system.
When we say that stringed music is louder, we are therefore referring to an increase almost uniformly across the frequency range. When we say that woodwind music is louder, we are referring to an increase almost exclusively in the higher frequency region. And when we say that speech is louder, we are referring to an increase almost exclusively in the lower frequency region.
|Low frequency channel||High frequency channel|
|Reeded woodwinds||No change||Increased|
Although the low frequency channel of a speech program may be similar to that of strings, this is certainly not true of woodwinds. The Table summarizes the similarities and differences between speech, stringed instruments, and reeded woodwinds, as the intensity increases from soft to loud.
When it comes to a higher frequency channel, there is no similarity at all for speech, stringed instruments and reeded woodwinds. There would be no increase in the intensity of the obstruents with loud speech. There would be the same increase in the intensity of the harmonics (as for the lower frequencies) with strings. And there would be a much greater relative intensity in the higher frequency region with a reeded woodwind when played at an intense level.
However (and it is always the “however” statement that makes for an interesting capstone project), even though there are differences between soft and loud inputs for musical instruments which is not the same as the differences between soft and loud inputs for speech, should we be setting our music programs in our hearing aids differently?
Perhaps, it is this high frequency very intense signal, that is “almost” too loud, that makes a saxophone or clarinet sound like it does, and no differences should be made. Also, perhaps this is really a non-issue after all. Loud speech is roughly 80 dB SPL, and quiet music is roughly 80 dB SPL. If we are examining how we should be setting a hearing aid to handle loud music (at input levels of 100-110 dB SPL) perhaps this has nothing to do with speech because speech is never that intense? It may not be an issue if we were only trying to compare the loud inputs of speech and music but there is no such thing as loud speech at 100 dB SPL. It would be neat to do a study of loud instrumental music but reduced in volume to be normalized with that of average speech (also at 65 dB SPL) just to look at this “spectral difference” issue (or non-issue?)