Compression

You won't typically have very much control over expansion (whether it be the knee point or the ratio) The most control you're going to get is control that says gain for very soft, which will allow you to

decrease gain for very soft inputs

Modern hearing aids have attack times varying from less than 1 msec to a few seconds, although the majority of manufacturers apply relatively short attack times (1-20 msec), at least for sudden high intensity sounds, to help ensure...

that these brief bursts are not uncomfortable for the patient

Concept of "effective compression ratio"

refers to the fact that the effect the CR has on the change in gain with changes in level depends on SIGNAL LENGTH AND TIME CONSTANTS

What are two ways we can define compression?

1) Compression threshold/threshold of activation 2) Compression ratio- how much we turn down gain with increasing input (change in input/change in output)

How much amplification do you need to make average speech comfortable? (In percent of hearing loss. Or fraction of hearing loss.)

1/3 gain if you have a mild degree of hearing loss As you get to a higher degree of hearing loss, 2/3 gain

What's the difference between ADRO and Speech Guard in terms of the linear sliding window?

ADRO is mathematically based. It's based on probabilities. And Speech Guard is purely level based.

Want fast release after _____ and slow release for ____

Door slams Speech

For speech, if we look at the loudest phonemes and the quietest phonemes, what is the range? Why is this important?

15 dB -Maxima are about 15 dB above the average -Minima are about 15 dB below average Why is this important? 15 dB below 53 (soft) is very soft 15 dB above 83 (loud) is very loud This means TOTAL dynamic range of speech for a wide range of vocal efforts is 60 dB

94% of speech fits into this patient's

2 channel hearing aid (i.e., 2 channels of compression) NOT CHANGING THE COMPRESSION RATIO, ONLY TIME CONSTANTS Using FAST compression (phonemic compression) time constants, what percentage of speech fits into the patient's residual dynamic range?

Size of speech box shrunk 82% of speech fits into this patients residual dynamic range

2 channel hearing aid (i.e., 2 channels of compression) NOT CHANGING THE COMPRESSION RATIO, ONLY TIME CONSTANTS Using SLOW compression (syllabic compression) time constants, what percentage of speech fits into the patient's residual dynamic range?

Don't make the mistake of telling a hunter than his hearing aids will work as earplug (i.e., they have a lot of compression) because it's going to shut down the gain and then give them a _____ You just caused this person hearing loss!

2mm vent

Popular prescriptive methods (NAL-NL2 and DSL v5) limit compression ratios of ____ or less for input compression What about output compression?

3:1 Want output compression a lot higher because you don't want the output to change very much when you change the volume. Often will see 10:1 or higher.

Input = 100 dB SPL CT = 40 dB SPL CR: 3:1 Gain: 30 Output?

40 + (60x1/3 = 20) + 30 = 90 dB SPL How can output be lower than the input? Fast acting compression with the output limited below the input This is the idea behind hunter's earplugs. Only works if the ear is sealed off (no venting obviously)

What is the long-term average level for soft speech?

53 dB SPL (40 dB HL)

What is the long-term average level for conversational speech?

63 dB SPL (50 dB HL)

When you're fitting to 55, 65, and 75 in clinic, which one do you start with?

65 for most manufacturers 55 Then 75 BUT it depends on the manufacturer

In the real world (when accounting for individual talkers), what is the dynamic range for speech more like?

80 dB SPL

What is the long-term average level for loud speech?

83 dB SPL (70 dB HL)

Think of SAAVC as ...

A 110-year old man running the gain change level that is controlling the release time RELEASE TIME IS TOO SLOW He just can't move quickly enough, and by the time he starts moving, the input is already back up SAAVC is effective at controlling the level for sounds over long durations but not for amplitude changes that occur very quickly

What level do we want to be able to hear average speech?

A comfortable level Thus, for a long period of time, the general idea behind hearing aid gain was "let's get the average speech around the MCL"

The idea is: when you've got fast-acting signals, you can compress the dynamic range of speech for...

A single talker In the picture (the two signals), minima are the same. But the maxima are different. The only thing that's different between the two is the time constant.

Compression basically works in a very frequency and level specific way, so you can

Adjust it in a very level and frequency specific way so you can target what the complaint is.

What happens to the compression ratio for slow acting, syllabic, and phonemic compression as we decrease the length of segment?

All of these are set to 4:1 compression No gain change when we get to a short enough segment. Whereas if we do faster time constants, especially phonemic with look ahead, we can effectively compress all the way down to very short segments, even the length of phonemes

In most low threshold compression hearing aids, compression is

Always on

Input = 70 dB SPL CT = 50 dB SPL CR = 2:1 Linear gain= 35 What's the output?

Amount below CT is for free = 50 dB SPL Amount above CT is compressed = 20 x 1/2 = 10 dB SPL Add gain (+35 dB SPL) Output = 95

What is the only way to get more compression in one frequency range than another?

To have multiple channels

Input = 70 dB SPL CT =20 dB SPL CR = 2:1 Linear gain = 50 Output?

Can't be higher than 120 dB SPL (linear gain rule) 20 + (50 x 1/2 = 25) + 50 = 95 dB SPL

In order to address clinical problems with gain, it's important that you actually understand the problem. How can you quantify/understand the problem?

Ask: What specific sounds? What specific input levels? What specific frequency ranges (or is it broad band)? Is it discomfort, annoyance, distortion, or other factors?

We've been talking about compression as if it's static. Input comes in, we adjust gain. The reality is in almost all hearing aids, (not all), this does not happen relatively instantaneously. It has a time cost to it. All hearing aids that use compression are going to have ____ and ____ times

Attack and release E.g., An input comes in, it says oh, that input is that level, I need to start turning down gain and get gain to the target level

When do you have over vs. undershoot distortion?

Attack times- overshoot (in the process of turning it down) Release times- undershoot

The vast majority of current high-end hearing aids use... Especially if there is a...

BOTH input compression and output compression Especially if there is a volume control

AGCi is used to ensure the desired loudness relationship for soft, average, and loud inputs Conversely, we are also interested in ensuring that loud sounds are not too loud at the hearing aid wearer's ear Conceptually, AGCo with a high CK makes sense for limiting the MPO Consequently, it is quite common for hearing aids to implement

Both AGCi and AGCo

What's the idea behind the linear sliding window?

Both compression and linear are good and bad Let's combine the two Keep it linear whenever we can, but realize that changing gain for higher or lower levels is a good thing

What do we call breaking the frequency ranges up?

Breaking up into CHANNELS

Hearing aids need to look at the incoming input levels and adjust the gain. What is the only way that is going to be able to do that?

By looking at different frequency ranges differently It's going to have to break the frequency ranges up

Unaided audibility: only about 23% of dynamic range of speech fits/is audible

Calculate audibility- the percentage of the area of dynamic range of speech that fits

Some modern hearing aids have multiple output compression knee points within each _____. There are others than have one compression knee point in each ____ or even one total output compression knee point across all ____. Why are there so many different ways of doing this?

Channel Channel Channel You've only go so much signal processing power. You want to do things where you're as efficient as possible in your processing power so that you save your processing for the things you think are really important

What is this called? What concept does it use?

Channel free processing Instead of "channel free," Dr. Ricketts would call it "lots of channels" because it's applying gain in FFT points Uses concept of parallel processing

When you select a brand, you're selecting ______

Compression Sometimes it's when you're selecting a model

From Dr. Rickett's dissertation: Speech recognition is better for soft with _____ Additionally, 75% of patients said linear was loud or too loud. It was only true for about ____ of patients fit with the compression

Compression (this is true regardless of time constant) 12% of patients fit with compression I.e., Soft speech is more audible and loud sounds are not too loud comparatively to not using compression

How does compression work?

Compression has an activation threshold. Above that level, gain goes down with increasing input

Bottom line, for most cases when you're trying to squish these things together, what you're doing is adjusting ______. Though you might affect compression threshold.

Compression ratio

If you made average and louder closer together, what did you just adjust (time constants, compression ratio, or compression threshold?)

Compression ratio

Previous work suggests that working memory may influence speech rec performance as a function of _______ in very controlled conditions

Compression speed Does this translate to real patients with modern hearing aids?

If gain for average to gain for loud don't move, you probably did something with

Compression threshold

Where is the "level" that triggers the compression located in the logic stream?

Conceptually' the level at the microphone vs. the level at the receiver

What is the main use of input compression?

Control incoming levels over broad range of inputs (low-threshold WDRC compression) Can also be used as a compression limiter (usually done in modern digital hearing aids to reduce input dynamic range into the digital system)

If a patient complains that all sounds are too loud including speech, warning sounds, and soft speech What should we do?

Decrease overall gain! Also might want to mess with output control Decreasing overall gain doesn't necessarily decrease output. It depends on how much gain is being applied relative to where MPO is You can decrease overall gain, and then you can decrease output, and then you could run an REAR90 on the person and get some feedback from them of whether those very loud sounds are too loud as well

Amount of compression decreases with _____ unit length

Decreasing

Study on kids comparing: sliding windows, multi syllabic and linear window

Did about 4-5% better sliding linear window on average than either syllabic or linear...4% is not much of anything. But when you look at hearing loss, the more hearing loss they had, the more speech recognition benefit they got from the sliding linear window compared to the syllabic. This is consistent with past data looking at phonemic vs. syllabic, which is you've got a lot of hearing loss, you want less compression.

Sliding linear window and ADRO and Channel Free, all three will say no, it's absolutely not compression. In my head what compression is you adjust gain based on input, and they all...

Do that

When output compression is used for output limiting, we can change the volume a lot and the output above the compression threshold ....

Doesn't change very much This let's us avoid the risk of high inputs exceeding threshold of discomfort

The data in general comparing fast vs slow in big groups of people ....

Don't show big differences

Example of overshoot compression artifact: "shoes": when the O comes in, rather than having a really slow transition, as amplitude is decreased on the O...

End up with a much faster transition If you do it wrong, the O can start to have a burst type of quality to it and can almost sound like a stop Hear a "pop" at the start of the O from the overshoot compression Can mishear what's actually there

Most common distortion

Fricatives will become more affricate sounding

When we talk about things like phonemic compression and syllabic compression, we're talking about time constants that are ___ enough that they can act on and change the amplitude of things as short as phonemes or as short as syllables

Fast

Within the last two, three, four years, a lot of companies have moved to dual time constants, or even slower. Even companies that did fast for at least some of their instruments because this data came out. If you have patient, however, who is younger and has hearing loss and needs to understand speech in noise really well, that might be somebody you try out ____ on

Fast

When we start looking at individuals, Some people with good cognitive ability do better with ____ compression in noise On the opposite end, some people like the sound quality better of ______ compression People with more limited cognitive abilities seem to do a little better with the ____

Fast Slow-acting Slow-acting This suggests cognitive ability or working memory is something we need to think about

Why does it matter if we assign compression based on the level before or after amplification? The most important factor relates to how AGCi and AGCo interact with a volume control Explain this

For AGCo , we can see that because compression behavior is based on the sound level after linear amplification, compression is activated above the same 85 dB SPL output cross all three volume control settings As a consequence, as the user increases the volume, gain increases for input levels below the MPO, but there are no changes to maximum output The effects of changes in volume control with AGCi are considerably different. Because compression behavior is based on the sound level before linear amplification, compression is activated above the same 40 dB SPL input across all three volume control settings Consequently, when the VC is changed using AGCi processing, both the gain and the MPO are affected

How fast is fast? How slow is slow?

For most companies, fast is as fast as half a second Slow might be 30 seconds (so there's a big range of what slow means)

Are more channels better?

From an analysis standpoint, more channels can give you a cleaner look at the signal But if you think about the fact that you're applying gain differently in all of these different channels and speech is fluctuating all over the place in level and frequency ..... the spectrum is going to get smoother and smoother and smoother You're going to lose spectra resolution of the signal

BUT when they looked at individual data... they saw that there's not really a pattern relative to working memory and slow-acting Working memory is NOT a predictor for slow What IS a good predictor for slow is...?

Hearing loss and age Those with more hearing loss are going to do better with slower People who are older are going to do better with slower

You're more likely to reach the output limit if the gain is ___

High

What manufacturers actually do with channels is you've got most of the major manufacturers for most of their products using a relatively high number of channels, 15, 16, 20, that are

Highly overlapping (broad filter skirts)

Attack and release times are NOT the time it takes to turn compression on and turn compression off This is a common MISTAKE What really are attack and release times?

How long it takes to readjust gain in response to a specific change in input level Attack time: the time it takes to turn gain down in response to an increase in signal level Release time: the time it takes to turn gain up in response to a decrease in signal level

What's the problem with fast attack and slow release?

I would be talking and I'd talk at a particular level, then you'd hear a door slam and then my level would go down and then it would release slowly and you'd miss the first couple of words. The problem with fast attack slow release in isolation is the hearing aid goes dead or appears to go dead when a loud burst happens.

What is look ahead compression?

If time constants are fast enough, you can have a sensor looking at the input - If a high input level is going to come in during the processing time of the hearing aid... you can turn down gain before the signal goes through. Square these off to where there's no undershoot or overshoot distortion

What was early hearing aid provision like?

If we said we want to hear the softest things, then almost all of shouted speech is going to be too loud. If we want to make sure the loudest things aren't too loud, all of soft speech is going to be inaudible. And so what early hearing aid provision did was to say, "we don't care about soft and loud. Let's try to make average comfortable." And then we would look at this loud line of hey, anything louder than that's uncomfortable (threshold of discomfort). And we would clip off any signal that came in that was louder than that and basically limit it. And distort it.

The amount of gain we want for a linear vs. compression/non-linear hearing aid for an average input is...

In the same ballpark. Our goal for average speech hasn't changed.

Can think about the threshold of activation as the signal level at the microphone (_____) or at the level of the receiver (_____)

Input Output

Linear sliding window

Input window that is based on level If the input falls within that window, gain is linear (no change) If inputs fall above or below this window, this triggers a FAST shift in the amount of gain Then it will be linear again

What's the effective compression ratio at 2kHz ?

Inputs are 50, 65, and 80 Difference in input = 15 dB Difference in output? = 7.5 dB Compression ratio: 2:1 (also depends on the time constants) Compression threshold looks like it's below 50 because it's compressing the same

What are dual or multiple time constants? (Currently one of the more popular ways to do things)

Inputs identified as being in the range and dynamics of speech are going to have relatively slow time constants But if a high level signal comes in, especially a high level short signal, release times are going to get much faster.

We don't want adjacent channels behaving very differently What can we do to get around that?

Instead of having 20 independent channels, we can have 20 highly overlapping channels (broad filter skirts)

The patient complains that low level sounds like fans and paper driving him crazy What should we do?

Issue might be that it's too loud Or it might just be that the patient has simply not heard these sounds before and now they find them annoying Possible fix for this might be counseling OR it might be a mix of counseling and some changes

Channel Free

It's do an FFT of the signal. Basically does a look ahead and then adjusts the filter shape based on the input level to give it this very smooth output response. You could either look at it as they do oh, it doesn't have channels or it has lots and lots of channels because it's looking at every point and adjusting gain.

Linear processing can maintain natural amplitude dynamics, but at the cost of ....

Limiting audibility, particularly for soft speech

The idea of providing the same amount of amplification no matter what the input is, up to a maximum output, is called ...

Linear gain

How does the normal ear work?

Linear, then a lot of compression for fairly mid-level inputs and then it goes back to linear

You can say the compression ratio is 2:1, 3:1, 4:1, but that might only be true if we're talking about really really____ segments

Long

SAAVC 7:1 (2 second release)

Long running soft speech, long running average speech, long running shouted speech got closer together in level They fit better inside the person's dynamic range. But what about maxima to minima of the phonemes? They don't change at all!

If we provide the same amount of amplification regardless of input level, what happens?

Loud is going to be too loud Soft is going to be too soft

What might be a fix for a patient in which low level sounds were bothering him?

Lower the compression ratio

The control in the hearing aid that allows you to adjust the output compression threshold is most often called....

MPO (maximum peak output)

What is the main use of output compression?

Mainly used to limit output for sounds (OSPL 90)

If we have a properly fit hearing aid, and we compare linear fitting to compression, what's the difference relative to soft speech versus loud speech?

Makes soft speech louder. Loud speech softer. Makes them closer together.

What happens when you increase compression ratio?

Makes them even closer

What might expansion be called in the manufacturer's fitting software? Why?

Mic noise reduction Especially for hearing aids with a low knee point, they will have expansion (turning down gain below a knee point) ... One of the low level inputs that we might want to turn down is the noise from the microphone If we turn gain down for very soft, we can help this out Other names include: gain for very soft or soft squelch

With digital hearing aids, we can break frequency ranges into ____ channels where we can look at fairly ____ frequency ranges

More channels! So we can look at level of the input signal in SMALLER frequency ranges and then apply compressor gain within those levels

Does instantaneous SNR change with compression?

No

Are handles related to channels?

No Typically have more channels than handles

If we use fast enough time constants, high enough compression ratios, and a few other tricks... we can squish everything into that dynamic range ... BUT Is maximum audibility for all speech input levels optimal?

No We are distorting speech by changing the amplitude relationship. (we're making the minima and maxima closer together). E.g., sporting event on TV- everything is the same level ... they're compressing the heck out of the signal They're using is very high levels of fast acting compression to make that work If you have normal hearing- this does nothing to intelligibility If you have hearing loss, it decreases intelligibility

Does normalization work? I.e., adjusting gain so that the output dynamic range from the hearing aid changes to be somewhat similar to the residual dynamic range

No, it's too much compression if we do things too precisely

Phonemic 7:1 (20 msec release)

Now...not only do we have soft, average, and loud speech closer together, BUT we've also got maxima and minima closer together And so we've not only compressed speech in the long term, we've compressed speech for a singer talker GERBIL ON CRACK

What is the mechanism behind losing dynamic range when we have hearing loss?

OHCs are compressors THat's their job. They provide more gain for soft sounds and less gain for loud sounds. The first thing that goes with SNHL usually is OHCs

How do you measure where output compression threshold is set (in the test box)?

OSPL 90

They are the same! How could that be?

Our goal is to make average comfortable If we lower our compression threshold, we need a lot more linear gain to get the same outputs for higher inputs Because of our low compression threshold, we're making the amplifier work a lot harder to give us the same output

What's the advantage of conceptually thinking about this an input compression?

Our goal is to take low level speech and make it soft, average speech and make it average, loud speech and make it loud That's based on input level

Gain + input =

Output

If you are thinking of limiting the output of the hearing aid, conceptually people think about ______ compression

Output compression With output compression the angle is a lot steeper

When you're adjusting the MPO, you're adjusting the...

Output compression threshold

In modern hearing aids, because we're using low threshold compression, peaks don't get to ____ as often as they used to

Output limit

What is the process we use to ensure that high level inputs are limited? In most hearing aids how is this done?

Output limiting In most hearing aids, this is done by output compression limiting

Look ahead is used to suppress...

Over shoots and under shoots

You've got a few products out there, not many left, but a few, that use a smaller number of channels, maybe, 6, 7, 8, 9, that aren't very

Overlapping

Back in the days of linear hearing aids, the way that we did limiting was...

Peak clipping Cut off the input level (created quite a bit of distortion)

What happens with spectral smoothing?

People do worse Especially in noise

How did the idea of loudness normalization come about?

People wanted to make up for the active compressive mechanism that people with hearing loss lost. Except it's gone and you can't make up for it.

E.g., of problem with perception +10 SNR Apply 7:1 syllabic compression What happens to perception?

Perception is that's a lot noisier. Noise in the valleys within the speech segments are going to be brought up because it's +10. The noise is softer, bring it up and make it closer. So the effect on speech recognition is nothing for people with NH because the instantaneous signal to noise ratio hasn't changed but the I hear the noise between those high level speech segments much more clearly. BUT for people with hearing loss, they use amplitude difference cues as speech recognition. Melting them away hurts them

Originally people said we need fast acting compression with lots of channels, we're going to restore normal loudness! What happens if we do that with patients with hearing loss?

Problem with perception

What should we do instead? How is this accomplished?

Provide more gain for soft sounds Less gain for loud sounds This is the job of compression

When will manufacturers only have output compression? (i.e., when will they not have low level compression) Why?

Severe to profound losses If you use input compression with a low compression threshold, you need more gain for soft. If you've got somebody with a profound hearing loss, you need a lot of gain for soft More than what a lot of hearing aids can put out.

Sliding linear window you can think of as less compression, at least for a

Single talker

A sliding linear window, by function of it being linear for range of inputs, is going to preserve more of the dynamics for a _________

Single talker Picture: Purple vs. gray is peaks vs. valleys for sliding linear window vs syllabic compression

People with low and high working memory are similar for ____ Those with low working memory perform significantly worse with ____ Those with high working memory performed significantly better with ___

Slow Fast Fast

If you're only concerned about sound quality for music, ___ is probably a safer option that ____

Slow is a safer option that fast

Study by Brian Moore Had attack and release times that were fairly -Fast: 10 and 100 would be considered syllabic. -Slow: 50 and 3,000 would be considered slow. They limited compression ratio to 3:1 for fast but they said for slow we can go to higher compression ratios because we're not squishing dynamic range of speech. What did they find in terms of preference?

Slow is preferred for speech sound quality For music, everybody preferred slow or didn't have a preference It isn't the case that you have a person that goes oh, I love fast for speech and I love slow for music People were consistent in their preference, it was just stronger relative to shifting for slow a little bit for music

For a lot of our patients who are older and have a lot of hearing loss and may or may not enjoy listening to music the safe bet is to go to ______ Also for children who are younger perform better with more ____ processing/slower time constants/sliding linear window

Slower Linear processing

Bottom line, differences in average speech recognition and sound quality across different gain schemes are generally ____. Individual differences are there, though, and you get people doing 10 15% better for one type of compression versus another. Best candidacy predictor is hearing loss and age. ______ is a pretty limited predictor.

Small Cognitive ability is a pretty limited predictor

For someone with SNHL, their residual dynamic range from lowest level to loudest level sound is ...

Smaller It shrinks with increasing hearing loss

The higher the compression ratio, the more gain for _____

Soft

Dillon et al., 2003 did not find any preference differences between normal hearing and hearing impaired people in terms of _____

Sound quality Authors attribute trends not to compression differences but rather to correlations with objective measurements that imply that devices with preferred sound quality have: -Low internal noise -High peak OSPL 90 -Low internal delay

And then you've got another way that you could analyze the incoming signal. Rather than breaking everything up into channels and looking at what's in the input, you can take

Spectral snapshots of the incoming signal (FFT) Gain is being applied through FFT rather than in channels

If we are talking about short segments, we might be compressing less because we've got...

The old guy that can't move fast enough

Why does residual dynamic range shrink with increasing hearing loss? In general, threshold of discomfort for normal hearing and hearing impaired listeners listeners is about ... So for listeners with hearing loss... their threshold is going ___, but their threshold of discomfort is ____.

The same ~120 dB For listeners with hearing loss their threshold is GOING UP, but there threshold of discomfort is NOT CHANGING very much (until you get quite a bit of hearing loss and then it changes a small amount)

If you've mistakenly set an output limit low enough you will get ...

The same output for all the different inputs

Plyler et al., 2013 found that there was no difference in: Speech recognition in quiet and noise, outcome measures, satisfaction Between multi-channel WDRC syllabic and channel free processing What does this tell us?

There's not a lot of differences in terms of compression types

In terms of sound quality, why do people tend to prefer what they have?

They're used to a particular compression architecture (e.g., time constants... band filtering or FFT of a particular time) which all change the sound of the hearing aid *Why we stick with a particular manufacturer

Real world outcomes here? Not any different at all Where they're providing more gain is in the noise floor

They've got the same gain for average, they've got the same gain for soft speech, they've got the same gain for loud speech. From a speech input level, these things are no different. What this will create if you use this really low CT is annoyance for a fan

Triangles: Compression threshold: 30 Compression ratio: 3:1 Linear gain: 40 Circles: Compression threshold: 50 Compression ratio: 4:1 Linear gain: 25 Diamonds: Compression threshold 40 Compression ratio: 2: 1 Linear gain: 45

This is ONE CHANNEL. Input/output function, linear agin, compression ratio is going to vary from channel to channel

We want to know...what does compression do for soft, average, and shouted vocal efforts?

This shows soft, average and shouted vocal efforts (53, 63, 83 dB) Speech from maxima to minima is 30 dB (peak to valley is always 30 dB) Let's add compression

If you notice that you're turning up the compression and can't get loud and soft close enough together, sometimes that's the ____ and not the compression ratio

Time constants

There are manufacturers where you adjust 65 first and 55 and 75 will never be right. They are too low and too high and you push them and nothing happens. That's because the....

Time constants (release times) are slow enough that you're never going to be able to force those together

Modern hearing aids vary tremendously with regard to release times Ranging from as short as 20 to 30 msec or less to as long as five seconds or more What happens is the release time is too short vs. too slow?

Too short: the hearing aid may have slight distortions (e.g., audibly pumping of ambient noise in a quiet room- the increased gain in gaps in speech makes ambient noise audible to the user) Too slow: the hearing aid gain may not be restored quickly enough, resulting in dead spots

Define expansion

Turning down gain below a knee point Thinking about this as the opposite of compression

What does compression look like for someone with normal hearing?

Very different than what we do with hearing aids. Pretty linear gain for low level inputs, compressive for mid level inputs, and it goes back to linear for higher level inputs

You can design the hearing aid as input triggered compression or output triggered compression No way to know unless you have a ____

Volume

For input triggered compression, no matter where we adjust the ____, it's always triggering compression at the same input level

Volume In this case, 30 dB input compression

Why are hearing aids not like eyeglasses?

We aren't putting a lens over something. We're taking a system that's damaged and we're trying to feed it a signal that works. And so we can't do the processing for the ear and feed it to the ear and assume that it's going to sound normal. All we can do is give it the best signal we can.

We want to compress more in the ___ frequencies than the ____ frequencies Why?

We want to compress more in the high frequencies (higher CRs) than the low frequencies because we've got a smaller residual dynamic range (Applying more compression in the highs vs. the lows) Even if the hearing loss is flat, you want to provide more compression in the highs than the lows (Dr. Ricketts will explain more as to why in Hearing Aids III)

Patient complains that loud speech is too loud (distorts), warning sounds (e.g., fire trucks) are OK and average speech is OK What are some solutions

We want to get this output line below the circle We could: -Lower overall gain -Increase compression ratio -Increase compression ratio and overall gain -Increase compression ratio and CT All of these things are going to have different outcomes for all the other sounds All of them will potentially fix the problem, but how much of another problem they create depends on what's actually happening ...

Selecting NAL doesn't get you to NAL. It gets you to

Whatever their NAL prescription is So select the prescriptive method in their software that gets you as close as possible to what you want your end point to be on average

Don't be a slave to your prescriptive method!

When you first adjust the hearing aid it will ask you what prescriptive method you want to use By selecting prescriptive method you might be affecting your time constants Your prescriptive method is there as a good starting point but it shouldn't override things like hey, this is an older patient with a lot of hearing loss. I want to use slow acting. I know if I select NAL NL2 it's going to give me faster acting (just an example). I don't actually want that.

The most common type of low threshold compression is....

Wide Dynamic Range Compression

Most hearing aids include a form of input compression where they trigger at a fairly low level and they do so for controlling level across a variety, wide range of inputs What is this called?

Wide Dynamic Range Compression (WDRC)

If you're going to switch manufacturers, switch to a company that is doing something somewhat similarly to the company the patient used to have and sound quality might be fine E.g., You wouldn't want to switch a patient from Widex to Resound. Why?

Widex does things very slow Resound does things fast

Remember the "linear rule"

You can't have more output than with linear! Input = 70 dB SPL CT = 50 dB SPL CR = 2:1 Linear gain= 35 If linear, output = 105 dB SPL If you calculate something higher than 105, you know you're wrong!

What happens when you change volume?

You change your output

Working memory alone accounted for about 30% of the variance. Still the big player is hearing loss. But working memory plus hearing loss accounted for about 70% of the variance for fast. What does this tell us?

Younger people with less hearing loss and high working memory might do a little better with fast Older patients with more hearing loss might do worse with fast

Bottom line: trying to normalize loudness completely for people with significant hearing loss DOESN'T WORK If you use a high enough compression ratio and fast enough time constants with more severe hearing loss to normalize loudness, people with a lot of hearing loss will do worse than they do with _____

linear

Consistently more hearing loss for adults or kids more ____ is better performance, whether that's slower time constants or sliding linear window.

more linear (i.e., less compression)