Hello,

I need to find a good way to detect a specific sound in a room of a museum. People will be given an object that makes a percussive noise (like a clapperboard), and I have to detect it and trigger sounds in response to that noise.

So far I’ve been playing with bonk~, but I can’t get good results using the learn mode. I got bonk working very well to distinct a finished number of various percussive sounds, but not with a specific one between hundreds of noises that could be done in the museum environment.

Has anyone done something similar? any ideas on how I could tweak bonk to do that?
thanks for your advices!

J.

some ideas: use ann to create a neural network and train it to recognize the right sound http://puredata.info/downloads/ann
use TimbreID and/or bark~ http://williambrent.conflations.com/pages/research.html
use correlation http://www.dspguide.com/ch7/3.htm (or matched filtering) http://comments.gmane.org/gmane.comp.multimedia.puredata.general/25715

or a mix of them

TimbreID, bark and bonk get really effective when I want to distinct a sound between others already 'known'.
I'll give a try using correlation and matched filtering.

thanks,
J.

something else from him http://williambrent.conflations.com/papers/features.pdf