How to collect I2S microphone data effectively without SD card

[ujur007]

I have a custom board with ESP32. I have forgotten to add SD card slot to the board. But the board has two microphones.
I can get the I2S data but in order to process it I need something. I am not that familiar with this topic.
One idea is to use the USB interface to stream/send this data to the host or use something else such as BLE/Wi-Fi to do this.
Anyone has any idea or example work?
Thank You for the support.

[Josverl]

In what format do you want to export the data?

If you can just save it as files to the flash, say in a folder /captures then you can copy the off the board using mpremote cp -r :/captures ./captures

[ujur007]

I want to have it as .wav format.

[JackCode0011]

Since I'm working on something similar, here's how I'm solving it:

1. Write I2S data to a byte array.
   Option 1: Write the byte array to a file and convert it to a WAV file on a PC.
   Option 2: Convert the byte array to a WAV file on the microcontroller and save it as WAV file.
   Option 3: Stream input via USB to PC and collect it there with a Python script for example.

The decisive factors here are how much memory you have and how large your audio files will be? Can you give us any numbers?
I myself work with SD cards and convert the file on the microcontroller itself. What is special about my approach is that I continuously write the audio to a ring buffer and, when a defined signal is received, copy the past 5 seconds and the next x seconds to the byte array. The x seconds are defined by the disappearance of the triggering signal plus a defined additional time of, for example, 10 seconds.

[ujur007]

Thanks this is nice feedback.

I have ESP32-S3 device, and it has enough Flash memory to store data. My audio files will be just some seconds long for keyword spotting.
I will try to do this over a socket connection.

[jonnor]

For USB, the standard type to show up as a soundcard/microphone on the PC, would be Audio Device Class
However, there is no example code for that in micropython-lib, yet. https://github.com/micropython/micropython-lib/tree/master/micropython/usb
I also think that USBDevice requires ESP32-S3 ?

So the easiest is likely to use Wi-Fi and send the data to a web server. For example, send a few seconds at a time, using either HTTP or MQTT. You will want to implement some queuing/buffering between the reading of audio data and transmission over Wi-Fi. I hope you have a ESP32 board with PSRAM, as this will be a bit RAM hungry to get working reliably.

[ujur007]

Yes, thanks for this explanation. I have ESP-32S3 and it has PSRAM on it.

[ujur007]

Hello people,
I am back with the solution to the problem. Now I am doing the audio transmission manually. So I record the audio and then use mpremote to send the recorded audio to the host PC.
This works nice. I can hear the words that I spoke. It is low volume, but I can hear it clearly.

I think this topic is not that much discussed openly on internet. Would be a nice idea to create some examples from my side.!!