Looking for Suggestions for Voice Recognition with ESP32-S3

@huckohio I can't help Mike, but I will point out that the ESP world is in some disarray at the moment with lot's of broken code. The board level has changed from R2.x to R3.x to make use of the S3 series unique features. Unfortunately that broke a lot of existing code and libraries.

Maybe as a first sanity check, update your boards and then re-compile all your code to make sure nothing is broke. If it is, roll your boards back to 2.0.17.

I recall someone on the forum spoke of using some sort of cloud based VR but I think that is a poor solution. I also have a vague

Good luck.

p.s. I just did a quick check of existing libraries, and I don't see anything that could be considered 'ready for prime time'.

p.s. In case you missed it, see these links

https://randomnerdtutorials.com/esp32-migrating-version-2-to-3-arduino/

https://docs.espressif.com/projects/arduino-esp32/en/latest/migration_guides/2.x_to_3.0.html

Hi Mike @huckohio,

  I hadn't even heard of the ESP32-S3-EYE board before, but out of curiosity, did a 5 second Google.

As you have no doubt already seen, the Getting Started guide at

https://github.com/espressif/esp-who/blob/master/docs/en/get-started/ESP32-S3-EYE_Getting_Started_Guide.md#13-block-diagram

says:

Except for GPIO3 that can be used to configure LED statuses, all GPIOs of the ESP32-S3-WROOM-1 module have already been used to control specific components or functions of the board. If you would like to configure any pins yourself, please refer to the schematics provided in Section Related Documents.

Which confirms that the board designers have not made life easy for you.

A further 5 seconds, looking at one of the schematics suggested that there MIGHT be opportunities for reconfiguring one or more GPIO pins, so that communications with another processor becomes feasible, but that will require surgery to the board, and probably, also surgery to the supporting software from Espressif, both of which are likely to be tricky at best, and possibly, practically impossible. (Bear in mind, some of the processor pins and wiring shown on the schematics might be hidden under the rectangular lid ... check before deciding on any hardware surgery.)

So whilst, I would not rule out the possibility for an experienced practitioner of both hardware and software being able to use the board as you suggest, it looks like a lot of skilled work, with no guarantee of success.

If you are looking for a challenge, then you might consider purchasing one, and start with the 'out-of-the-box' functionality, to evaluate whether the voice recognition capability would meet your needs, before worrying about the problems of interfacing it to a second processor. With a little luck, you may get some useful experience, but be prepared for hitting a brick wall or two.

Ron (@zander) has already pointed out other contemporary issues with the Espressif family, which he is more familiar with than I am, so also be prepared for a 'rocky ride' on the 'out of the box' support situation.

Good luck, with whatever you decide.

Best wishes, Dave

@huckohio Does that board have WiFi or Bluetooth? If so, use that (maybe esp-now) to communicate to the other esp32 to do the VR.

I think you will do a lot better using a Pi4, it has the VR (search on this forum for 'voice recognition'). Maybe start there then back port the ESP32-S3-EYE code to the Pi4?

I am not a fan of VR and have disabled it on all the devices I own that have it (except Alexa), I don't even use SIRI. I had several bad experiences with my car listening to me and then doing something stupid.

Good Luck.

EDIT: Just stumbled on this 

Espressif’s speech recognition SDK

in https://www.espressif.com/en/products/devkits/esp-eye/overview.

@zander Thanks Ron.  I compiled all three systems (rabbit shed, chicken coop, and house) with the new library with no issues.  I am not a fan of the ESP32 solutions because they primarily use the ESP IDF - which I have no experience.  

I am still researching.

@davee Thanks Dave.  I am not that "skilled practitioner" so I've eliminated the ESP32-S3-EYE as a possibility.

@zander

Posted by: @zander

↑

If so, use that (maybe esp-now) to communicate to the other esp32 to do the VR.

That is one possibility that I am looking into.

Thanks Ron.

Hi @huckohio,

   Probably a wise decision to give the ESP32-S3-EYE a miss at present, as I think there is a significant risk that you will hit some pretty hard brick walls, which can be very challenging, especially if you are working on your own.

  Of course, the more you do, the more skills you will acquire, and may become a 'skilled practitioner' in your own right, so don't give up.

Best wishes and good luck, Dave

@huckohio Maybe a Pi Zero 2?

@zander I'll look at it.

@huckohio Here are a few links.

https://dronebotworkshop.com/hacking-google-aiy-voice-kit-part-1/

https://dronebotworkshop.com/hacking-google-aiy-voice-kit-2/

This camera may be better in terms of pins, and it has an onboad Mic I think. I just got a couple in the mail last week. https://dronebotworkshop.com/xiao-esp32s3-sense/#XIAO_ESP32S3_Sense_Pinouts

Hope this helps.

@huckohio Here is another possibility. It's still an esp32-s3, but it appears to have quite a few free pins, and an on-board mems mic. I have 2 but have not yet messed with them. Price is also reasonable.

https://www.seeedstudio.com/XIAO-ESP32S3-Sense-p-5639.html

@zander Thanks Ron.  As I was researching options/links last night I decided on the ESP32-S3-Box-3 (Link).  It's a little overkill, but it works with the arduino IDE and has GPIO pins that I can use. 

I did look at Bill's videos with the Seeeduino, but I was under the impression that the mic was for video audio, not for voice commands.  I may have misunderstood. 

While this started out as a fun thing to add to the chicken coop, I am seeing other possible usages.

Thanks for all you comments/recommendations.  I appreciate you (and @davee) and the help you offer everyone.

Mike

@huckohio That looks interesting, but it appears to need ESP-IDF which you earlier were not happy with. Am I misunderstanding something I may have missed on a very quick look?

On this page

Link may be not showing, so here it is again

I screen gabbed the following

@zander Ron, I hope I didn't misunderstand the marketing material, but here is what espressif listed under features:

The last bullet lists development frameworks as ESP-IDF, Arduino, PlatformIO, Circuit Python and more.