Hello All,
I may not be posting in the proper location... could not find a 'good match' .
Anyway, I am describing an ongoing project that I am working on... it is for a close relative who had a large stroke, and is quite disabled, now.
He has not lost his ability to think, but has difficulty communicating ; has only use of one hand, slightly.

I am building a joystick interface, with software so that he can select, On-Screen, letters, or numbers. The alphabet appears as five rows of LARGE
characters ( A - F, G - L, etc). By moving the joystick, an individual letter, or number is hilighted. He cal then 'Click' the joystick to select the character.
It is added to a word 'being built' at an area at bottom of screen. I am working on a word search algorithm so as he enters characters, matching
possibilities appear at right side of screen. The display is a large screen TV display, driven via a VGA to HDMI converter. ( $12 On Line !! ).
I am using a joystick like used for the Arduino-type projects. It has X and Y potentiometer outputs, and when pushed down, connects a switch.
This goes to a custom board in a PC (wirewrapped) which uses an ADC0820 analog to digital converter. I made a GAL to do PC bus logic
decoding to enable 74HC4066 analog switch(es) for X, Y, Button scanning values. May also use an interrupt driven subroutine with Button press
to record screen location, thus the character to be selected. The library for word lookup is being built by scanning multiple text documents,
grabbing unique words, and placing them in a lookup table ( indexed by char 1, then 2, etc). i.e. sort of alphabetised, with indexes into
table. The goal is to allow him to 'speak' via the screen, and perhaps set it up so he can write a book that he had been planning on doing.
I am trying to give him a method of having 'purpose' in life, now. He had previously been a smart, vital person. He still is, but now is limited.

Anyway, this post is mostly for interest sake. Any feedback or direction is appreciated.

SESMD

 

I never really did any speak via screen projects like the one you have described. But I can at least say that Arduino is not fit for this task. You may need to switch to Raspberry Pi. There is a python library for speech to text conversion. I guess you will need to work with those.

 

This sounds like a simple $3 (ok, today $6) microcontroller can handle. It looks like you have thought his out pretty well and are on the right track.

Have you given any thought to mass storage, even if in flash of EEPROM. You are going to need to a way to store the eventual text and get it off the machine onto a computer.

 

Hi sesmd,
You could consider an SD card for audio/data storage, there are a number of Arduino projects using this method.
E

 

Welcome to AAC.

Two things come to mind.

First, there are many extremely capable MCU dev boards available today. You seem very comfortable with discrete logic and μP things, but to fall back on that without investigating the current options may be a mistake. Something like the Teensy 4.1 could do what you are doing and more. Whichever dev board you might choose, though, you gain a huge advantage of all the peripherals you might need being built-in and what would previously have been hardware changes now being software which is a lot easier to do.

Second, you might want to consider using the network to communicate with bigger, hungrier hardware software that could do more running on bugger platforms. This could be a strictly local connection to a computer that is running the search algorithm(s) or expanded to use APIs from resources on the Internet. The edge computing business has boomed and this local/network hybrid can be very powerful.

This also includes the possibility of using ML (Machine Learning) which seems very relevant here. It could learn about the patterns of the user and anticipate them. This sort of thing is a solved problem and easy to use in various frameworks. The boom in Edge AI would allow you to do some processing on the local device and ship off the harder stuff for a larger computer to deal with.

(Just a note, the one thing the Teensy is lacking is WiFI but that's easily added on if needed. Alternatively there are the ESP32 and ARM architectures to explore which are more "modern" than the Teensy in terms of things like wireless and ML).