Hi,
I am interested not in sound classification, but “infrasound”, very low frequency not audible sound files that are used for seismic activity detection. Do you think that a similar approach can be used for both sound and “infrasound” classification with Edge Impulse?
Regards
Interesting application!
I don’t think that edge impulse will cause you many issues in that regard.; it doesn’t really care if your audio data is audible or not. What kind of HW do you intend on using to capture your data?
Have woked with audio / spectrogram cnn detection of whales a few years ago. using hydrophones and a pretty standard audio card + rasperry pi setup. Never fully implemented the AI part of it; focus was more on long term recording and visualizing spectrogram data to visually identify relevant events.
I would recommend you to at least investigate that path. Using edgeimpulse to do spectrogram “object” detection instead of detecting raw data. But I’m not an expert on the topic; neither Cnn’s nor seismic activity.
tapping into higher / lower frequencies proved a bit challenging at the time but more because of the preamplifyers / hydrophones required for that ( I’m not a audio / HW engineer ). Also managing grounding / noise / interference was quite a relevant topic with lot’s of frustration potential.
Thanks for your kind response Adrian,
My plan is to use a diferential air pressure sensor as a “microphone” for infrasound. I would like to replicate this setup at first:
Regards,
Ivan