Hi
I am currently working on a project that requires me to convert user given speech input to text in real-time using an arduino(either locally or send the voice wirelessly to my laptop for recognition in my laptop). Is there a way I could implement this only by using an arduino uno and a microphone module as we have a tight budget. Any expensive solutions will also be appreciated(for e.g. using raspberry pie). Would be thankful if someone helps me as this is my first project of this kind. Thank you in advance

 

Is there a way I could implement this only by using an arduino uno and a microphone module as we have a tight budget

No. Text to speech requires way more compute power than an Arduino.

 

No. Text to speech requires way more compute power than an Arduino.

Got it, but first thing I want to actually convert Speech-To-Text and not Text-To-Speech. Secondly, is it possible to send the speech data to my laptop and use my laptop's computational power(or any AI speech recognition tool api) rather than do it locally within the Arduino to accomplish that? And Thank you for your time!

 

Sorry, I knew that, just wrote it backwards, so the same answer.

 

Welcome to AAC.

Tensorflow Lite and a supported board might offer a solution, but it really depends on the capabilities you expect. If you want something like dictation, you will just have to use a much larger computer. If you are interested in particular voice commands, something like MicroSpeech would work.

Full blown TensorFlow on a Raspberry Pi 4 can do pretty well, like spchcat.

As far as sending the audio to a computer, you can just use any streaming solution.

 

Welcome to AAC.

Tensorflow Lite and a supported board might offer a solution, but it really depends on the capabilities you expect. If you want something like dictation, you will just have to use a much larger computer. If you are interested in particular voice commands, something like MicroSpeech would work.

Full blown TensorFlow on a Raspberry Pi 4 can do pretty well, like spchcat.

As far as sending the audio to a computer, you can just use any streaming solution.

Thank you for the reply. I found some solutions for implementing using raspberry pie 4. But I wonder if somehow I can stream the voice over to my laptop using some extremely cheap microcontroller like Arduino Uno or Raspberry pie Pico and process everything inside my laptop and then send the text over to the microprocessor/controller. Your solution involves processing the voice inside the raspberry pie then streaming to my computer. Is my method even possible? Thanks in advance

 

Welcome to AAC.

Tensorflow Lite and a supported board might offer a solution, but it really depends on the capabilities you expect. If you want something like dictation, you will just have to use a much larger computer. If you are interested in particular voice commands, something like MicroSpeech would work.

Full blown TensorFlow on a Raspberry Pi 4 can do pretty well, like spchcat.

As far as sending the audio to a computer, you can just use any streaming solution.

Basically I am currently working for a university project and trying to build something like the emo robot.

My take is that I will stream the Speech recognition of the user to an customized AI language model API for example ChatGPT. Then the response received will be read by the robot using a custom trained text-to- voice AI(All this processing will be done on my laptop and the communicated to the robot). For simulating eyes/facial expressions I thought of using an i2c LCD/OLED display. The thing is I completely new to this stuff so thought of seeking help online. Feel free to correct me if my ideas are too ambitious or silly to implement and that should I drop it. Thank you

 

.

 

I

Welcome to AAC.

Tensorflow Lite and a supported board might offer a solution, but it really depends on the capabilities you expect. If you want something like dictation, you will just have to use a much larger computer. If you are interested in particular voice commands, something like MicroSpeech would work.

Full blown TensorFlow on a Raspberry Pi 4 can do pretty well, like spchcat.

As far as sending the audio to a computer, you can just use any streaming solution.

I liked this idea though. Using Raspberry pie my 90% problems will be solved but the problem is raspberry pie and it's mockups are very expensive here in India and it is not worth to spend this much for a few marks. Hence I am looking for a solution that includes a cheap microcontroller for just streaming and receiving data from my computer while all processing will be done inside my laptop

 

I don’t think you need a microprocessor at all! Get a Bluetooth Microphone and send the audio directly to the laptop via Bluetooth.

 

O

I don’t think you need a microprocessor at all! Get a Bluetooth Microphone and send the audio directly to the laptop via Bluetooth.

Ofcourse that was my first idea but my professor may not be satisfied with that, atleast if I do some additional steps he could be convinced. Also we would still raspberry pie as recording audio is not the only thing the robot will do .Thanks for the reply!

 

If you use an RPi running Linux all kinds of things open up. As I mentioned above, you can use something like DarkIce to convert the microphone audio into a useful format, then stream it using something like Icecast.

This can be received and used on the PC as a normal audio stream. Alternatively, you could simply encode the audio as a bitstream of ~5kHz bandwidth and attach that to a UDP port, with a corresponding listener on the PC side.