Notched Music Therapy (DIY, AudioNotch, etc.)

Do you know if this app can notch all sound output on the device? It seems unlikely, since it's not in the description. I think something like that would be the best solution.

Does anyone know of such an app? No pre-processing? Live notch filtering?

 

This app seems to do something close to what I described.

DSP Audio Filter
https://play.google.com/store/apps/details?id=com.wolphi.filter&hl=en

The problem with this app is that it takes the input from the microphone as the audio source. So that will not work for the intended purpose.

The advantage of such an app would be that you can use any audio app to play any sound file, local or streamed, and get notched output instantly. It may be a bit heavy on the CPU, but I would be willing to try it out. It can't be much worse than Facebook app and others that are already working in background and overloading the CPU. It should not be a problem if it's a high spec device.

 

Yes!

My understanding is that you need to widen the notch.

• More secure in the T frequency, narrower notch, more effective treatment.
• Less secure in the T frequency, wider notch, less effective treatment.

 

Q?

Notch Filter?

Just... what? Don't leave us hanging!

 

I am no audiologist, but I can tell you that this depends very much on the individual damage in the persons cochlea. Some may hear it very little, some may hear it pretty good.

Also, in normal hearing, it's not only one single frequency that is stimulated. Surrounding cells help out too. It's exactly this that notch therapy relies on. It assumes that the surrounding cells, around the dead cell, are healthy. How far off or away these healthy cells are from the damaged cell is what should determine the width of the notch. There is of course no easy way for us to know such thing, since we cannot see the cells.

You don't seem to be deaf at 11800 Hz then do you? If you have a damage at that frequency, the surrounding cells are helping you hear it anyway. This is what may be causing the tinnitus, the recruitment of surrounding cells to do double duty to compensate for the loss of the cell or cells around 11800 Hz.

That audiologist needs education. They are fitting people with hearing aids. They should know about "recruitment".

You are absolutely smarter than that audiologist!

 

Dd you make any of these sites?

 

Interesting twist! I have read about this here, before it was released:

http://www.audiologyonline.com/articles/notch-therapy-new-approach-to-18365

This is the first portable, self-contained device that I know of, that relies on the notch therapy theory. In the past, hearing aids have been fitted with noise functions that mask the tinnitus tone. But this may be the first device that subtracts sounds at the tinnitus frequency.

I don't know what to say about it, other than that they are relying on notch therapy, with a twist. The twist is hopefully obvious, so I don't have to explain it. But I have no experience with this.

 

This has been mentioned before, but in combination with DAW software like Garageband, ProTools, Cubase, Ableton Live, etc.

Is this equalizer all you need? Don't you need the virtual cable tool (Virtual Audio Cable)?

 

If you read this thread from the first page, you will find that there are about 10 members that have had some success with this, 2 have had bad experience.

Members with good experience:

• rogi
• dan01
• mcauth
• tintin31
• Khoo
• Bill Thomas
• smel
• don dowel
• kevinjohnston1
• Ambassador

Members with bad experience:

• 65vwbus
• George Albertron
• rrf

So 10 of 12 members had success. In other words 83% success rate? Well I would not put a lot of trust in this. It's not scientific. As @GregCA said the scientific papers are better source for conclusive evidence.

That's interesting! Can you post a link here later on? Since this relates to notch therapy as well as TMS.

 

So your T is around 4 kHz?

How long?

Notched?

What is constant?

So that's regular white noise, not notched?

Are you using AudioNotch service for this?

Sorry about all these questions! I just try to get a better picture of the situation. I am considering NMT or ACRN for myself.

 

One of my files, pre-process:


Post-process:



Question! How deep does the notch need to be?

You can see that at about 4000 Hz it's -68 dB. Is it necessary to kick this into the -90 or -100 dB zone?

It appears that you can improve this notch by repeating the process.



It looks like it goes down all the way. But there is no way to zoom out in this tool or change the scale of the vertical axis.

It says -90 dB on the axis, and -98 dB on cursor. The same reading is -80 dB in Audition.



Third process in Audacity:



Still says -90!

Viewed in Audition:



Says -92 on cursor!

After this quick experimentation I can only conclude that if you are going to use Audacity, make sure to process the file at least twice. Audition and other high quality, and payed software, is better at this and it may not require as much processing to get the desired results. I guess they have better algorithms for this sort of stuff. But Audacity is not bad... not bad at all! If you know what you're doing and what you want to achieve with it.

 

What do you mean by "150 order"? You mean -150 dB?

Is Butterworth an actual filter inside Cool Edit, or an algorithm/function of the Notch filter? Does it have an equivalent in Audacity?

I can confirm that Audacity comes a bit short when using the Notch filter in there, compared to other software. I compared it to Adobe Audition.

Question! So what's the target attenuation then? Is it -100 dB? Maybe even -150 dB?

Can someone back this up? It makes sense to me, but that's not very scientific. Music is a very powerful thing though, and I plan on using music files instead of noise. Music activates many areas of the brain, such as memory and emotion.

These are very good two points!

• Noise vs. Music
• MEG vs. interview? (TFI questionnaire?)

 

Yes, but it's not enabled by default!
http://manual.audacityteam.org/man/classic_filters.html

The highest order that can be specified for Butterworth filter in Audacity is 10. That's far from 150 specified in the study.

 

It is apparent how? I looked at the PLOS article and the PNAS article.

You can't easily achieve this type of notch with regular audio software:



What you are looking at is likely the output from Matlab! Highly advanced and complex software used by scientists and engineers all over the world. My guess is... they used IIR filter functions butter and buttord to specify Butterworth filter algorithm and lowpass and highpass cutoff. In other words they did this programmatically, not by using your everyday audio editor software.

This realization makes me doubt the efficacy of the notch therapy being sold by the many apps and web services. They need to be able to replicate the filtering protocol exactly! I am not so sure they do. They could do it of course, but they would need to do a lot more investigation into this and implement it correctly in code. This is more about code than audio.

By the way, Adobe Audition software that I am using is the successor to Cool Edit. I am not sure it even has a Butterworth filter of any sort. Will have to do some checking.

The closest you can get is to use the Butterworth filter in Audacity. You will find it after you add it. It looks like this:



You can see I have put in 5020 Hz which is approximately the result of 7100 Hz divided by square root of 2. This will give you the lower cutoff. But you can't do a lowpass and highpass at the same time with this. So it's immediately useless! It may be possible to approach it pro grammatically without the GUI part, if there is support for it. I will have to do some checking.

Another pitfall of this filter is that it only allows for 10th order calculations. I finally get the order part! It's math! It's that boring school stuff again, such as equations, polynomials, etc.

https://en.wikipedia.org/wiki/Total_order
https://en.wikipedia.org/wiki/Higher-order_function

Yada, yada, yada... there are some inherent problems with higher order functions that can cause round off errors. I suspect this is why the filter designers left it at 10 as the highest option.

Notch therapy made at home... easier said than done. Properly...

Update:
OK, so I checked and turns out Audition has a whole new set of filters called scientific filters since Adobe CC release (not sure which version).

As for Audacity, there is a plugin available for it that provides proper band stop filtering.
http://wiki.audacityteam.org/wiki/Nyquist_Effect_Plug-ins#Band_Stop_Filter
I don't think it uses Butterworth algorithm, and probably not a function of 150th order. But it's as close as you can get, using free software.

 

Source? You mean like what files I'm using? Right now I'm just experimenting a bit with music files. I will not be using noise.

 

No, I have no document. I wish I did. I am trying to put the puzzle together by reading different sources on this. The authors were not too specific about the filtering protocol.

I was taught in early school years that one of the criteria for a scientific experiment is that it needs to be reproducible by others. In other words it needs to be very well documented. It seems to me the authors did not give much weight in describing the filtering protocol. It's as if they just assume that it would be understood by reading the few lines above.

Yes, they are professionals, and the article is aimed at other researchers who probably understand what the authors mean. They can read between the lines. But I'm not dumb either... but it makes more sense to put everything in writing, instead of assuming that the reader will understand.

For example, they mention that they used Orbiter 922DH clinical audiometer. This is very specific, and important piece of information. But why is the software less important? If they did use Matlab, as a researcher I would like to know that. I would also like to know what commands and options they used. So that, as a professional researcher (which I'm not), I can replicate the experiment properly without deviation and guesswork. It just makes sense... being a scientist and all that...

I know they (researchers) have been playing around with this ever since that Italian paper on Windowed Sound Therapy came out.

I wonder if this has anything to do with it...

 

So... some of the tools or options that have been suggested earlier:

• Notch filter
• Equalizer

What you really want though is:

• Band stop filter
  • Butterworth

Notch filters are just that, they make very small notches in the "energy spectrum" (dB/Hz) as they call it. When the notch is an octave wide, it's no longer a notch filter but a band stop filter (or crazy wide notch filter, notch is really a subgroup of band stop). Why it's important to pick the right kind of filter is because they all apply different algorithms and work in slightly different ways internally (the fine details in code, math and physics).

It's not enough to just achieve a certain attenuation, you need the right width and curvature in the spectrum. That is if you want to properly replicate the research they did.

I don't want to claim that doing it exactly right will give you better results. But if you want to be scientific about it, that's what you need to do.

The order number is what gives the filter that steep slope. This is what the first 5 order slopes look like:


The steep slope is not the only quality of this filter. More on the science behind Butterworth:
https://en.wikipedia.org/wiki/Butterworth_filter

It's not without reason that Adobe grouped this filter in the category "scientific filter".