Exploring voice shifting and amplification for the hard of hearing

[orac]

I know a rather old gentleman who has lost the upper part of his hearing. In the past I have built a simple stereo amplifier based around a NE5532 (IIRC) which worked quite well.
The problem is, he still doesn't hear much due to loosing the upper frequencies of his hearing range.
So I have been searching around the internet a little and most voice changers seem to be based on HT8950, which are either hard to get hold of or suprisingly expensive (one IC about the same cost as a teensy 3.x).
So i thought about using an arduino, but its a bit slow and seem to need alot of external hardware to carry out voice shift.
Having played a little with a teensy 4.0, thought it might fit the bill when coupled with an audio board.
I have seen a few threads on teensy based voice shift but there didn't seem to be an outcome from any of them.

Over all the project will need to be battery powered and housed in a modfied set of headphones (I have acces to 3D printers), will need to be real time (or pretty close) and easy to use.
From my research i will need to use granular pitch shift, with a potentiometer for volume and another for pitch shift.
I was thinking about using a 3.2 or 4.0 due to the smaller form factor of the boards, both are very close in cost so that is not an issue

What is the ouput like of the audio board? I am thinking it maight need some additional amplification for this application.
With the granular ptch shift, is it applyed to both channels, or just a single channel?
Am i wondering down the wrong path and an analogue circuit will be better used?
If this a good way to go, I will be start with the granular example in teensyduino and work from there.

Thanks for any suggestion and insights you can offer.

 

[MarkT]

I know a rather old gentleman who has lost the upper part of his hearing. In the past I have built a simple stereo amplifier based around a NE5532 (IIRC) which worked quite well.
The problem is, he still doesn't hear much due to loosing the upper frequencies of his hearing range.
So I have been searching around the internet a little and most voice changers seem to be based on HT8950, which are either hard to get hold of or suprisingly expensive (one IC about the same cost as a teensy 3.x).
So i thought about using an arduino, but its a bit slow and seem to need alot of external hardware to carry out voice shift.
Having played a little with a teensy 4.0, thought it might fit the bill when coupled with an audio board.
I have seen a few threads on teensy based voice shift but there didn't seem to be an outcome from any of them.

Over all the project will need to be battery powered and housed in a modfied set of headphones (I have acces to 3D printers), will need to be real time (or pretty close) and easy to use.
From my research i will need to use granular pitch shift, with a potentiometer for volume and another for pitch shift.
I was thinking about using a 3.2 or 4.0 due to the smaller form factor of the boards, both are very close in cost so that is not an issue

The 4 has a higher current draw which is important for battery operation - its way overpowered too I think

What is the ouput like of the audio board? I am thinking it maight need some additional amplification for this application.

The headphone output is plenty for me at gain 0.55 or so, but it will depend on the load - the SGTL5000 datasheet gives the low down.

With the granular ptch shift, is it applyed to both channels, or just a single channel?
Am i wondering down the wrong path and an analogue circuit will be better used?
If this a good way to go, I will be start with the granular example in teensyduino and work from there.

Thanks for any suggestion and insights you can offer.

I've not used the granular effect(s). You can hook up audio components however you want BTW, having two
copies of a component is not a problem.

 

[BJB]

https://forum.pjrc.com/threads/40115-Teensy-Hearing-Aid

https://tympan.org/

My personal hearing aid is based on a T3.2 and does 'some' frequency shifting. Have been using it for many years.

My Neighbor, having tried several $$$$ solutions and also having lost most stuff over 4kHz, bought an earlier revision of the Tympan for me to experiment with. Took about three weeks of part-time programming and experimentation. Neighbor has been using it for almost three years, but have to change the earpieces about every 6 to 9 months. The Latest Tympan uses the T4.1

 

[defragster]

The 1062 in the T_4.0 and 4.1 can be clocked lower to use ~10-15% more power than a T_3.2, but offers 2X better performance - thanks to cache and other MCU features. Also the added RAM and onboard flash size.

Looking here gives graphs of relative coremark perf and current used across the PJRC ARM MCU's: github.com/manitou48/teensy4
near 200 Mhz seems like a sweet spot could be arranged in setup(), or the 396 MHz build option is ~5X faster at ~2X the current of a T_3.2 - and 528 MHz is much faster without the extreme current used at 600 MHz where the MCU voltage is pushed up:

 

[orac]

https://forum.pjrc.com/threads/40115-Teensy-Hearing-Aid

https://tympan.org/

My personal hearing aid is based on a T3.2 and does 'some' frequency shifting. Have been using it for many years.

My Neighbor, having tried several $$$$ solutions and also having lost most stuff over 4kHz, bought an earlier revision of the Tympan for me to experiment with. Took about three weeks of part-time programming and experimentation. Neighbor has been using it for almost three years, but have to change the earpieces about every 6 to 9 months. The Latest Tympan uses the T4.1

I am glad to have asked now, this shows that what I am trying to do is possible. I wasn't able to find much info for getting the tympan here in the UK.
I will have a much closer look at your project as i suspect you have already solved problems that I will come across.

The 1062 in the T_4.0 and 4.1 can be clocked lower to use ~10-15% more power than a T_3.2, but offers 2X better performance - thanks to cache and other MCU features. Also the added RAM and onboard flash size.

Looking here gives graphs of relative coremark perf and current used across the PJRC ARM MCU's: github.com/manitou48/teensy4
near 200 Mhz seems like a sweet spot could be arranged in setup(), or the 396 MHz build option is ~5X faster at ~2X the current of a T_3.2 - and 528 MHz is much faster without the extreme current used at 600 MHz where the MCU voltage is pushed up:
View attachment 26950
View attachment 26951

I have just checked the arduino IDE and it allows the user to set a lower clock speed there under the tools menu. I have a couple of teensy 3.2 on the way for testing and comparison to the 4.0 i already have.
It maybe a case that he has several battery packs that can be easily changed and charged so that he is never without it when he really needs it.