-->
Page 1 of 1

Sound classifier using NodeMCU

PostPosted: Wed Dec 23, 2020 7:52 am
by Jake Duningham
As the title states, wondering if this is possible using the ESP8266 microcontroller, since this was given as a group project by my teacher.

The basic idea is, a microphone (connected to the esp8266 (i don't know if this is possible)) records and sends the recorded audio to an AI (such as Teachable Machine) and classifies the source of the sound, if it's a dog/cat/human etc... then displays it on a website.

Any help or feedback is appreciated if this is something that can be done, or any alternatives, I'm fairly new to the whole IoT and electronics. Thank you!

Re: Sound classifier using NodeMCU

PostPosted: Fri Dec 25, 2020 6:29 am
by schufti
it should be possible but you won't be able to connect a analog microphone, you have to use a "digital" on i2s bus instead. there should be threads on this, use board search.

Re: Sound classifier using NodeMCU

PostPosted: Fri Dec 25, 2020 9:51 pm
by davydnorris
You can definitely connect a microphone and capture audio - this project should help you.

https://hackaday.com/2016/07/15/baby-monitor-rebuild-is-also-esp8266-audio-streaming-how-to/

The problem you're going to have is getting sufficiently high sample rate and depth to get the accuracy you require.

I use an I2S mic to capture 48kHz/24bit audio but that's too much to stream - I end up processing that on the ESP8266 for what I need to do. You may end up better off building a dedicated DSP circuit that does some of the processing and then sends a subset of data back