Using sound classification to trigger a camera?

Hi All,

I’m a bit new to all of this, and working with my kid to try and figure something out.

We’ve followed the Coding Train Teachable Machine Sound Classification tutorial, and got it done while having a ton of fun. She asked that if when she said “boo” instead of the ghost emoji showing up, could it take a picture of her?

Is this possible? If so, can we save the photo instead of displaying it and having it disappear when the next sound is identified? Any guidance in the right direction would be appreciated.

We’re using a Raspberry Pi with the new HQ camera.

Thanks for helping a dad out.

You should be able to take the picture, for example here is a simple sketch:

You should be able to save it if you check this reference:

Could you clarify how are you interacting with your raspberry-pi? I have to admit I am not familiar with “Coding Train Teachable Machine Sound Classification tutorial”. I understand you are using p5js, a raspberry-py and the camera is attached to it (?).

Let us know if the links above help or if you have any other question.


You might consider audio fft analysis, + the use of convolutional non. The fft will produce a spectrogram or an image with values from 0-255 across all frequencies then you can analyse that with a neural network. Reason I’m suggesting convolution nn, is that its likely to produce better results as the standard one will be inflexible to variations.