I'm working on a project, where we're recording sound with a piezo-disc which looks a little something like this:
Now, unless we're doing something horribly horribly wrong, I've discovered that we're only recording the magnitude of sound, since a piezo-disc emits a response to pressure, regardless of direction.
This is a sound-recognition project and previously, we'd been working with smartphone microphones. Back then, we could MFCC to perform feature extraction, yet now it looks like some things will break:
I fear that the pre-emphasis step in particular might lead to faulty results.
What kind of an approach could we follow?
- Is there a way to "guesstimate" the amplitude from the magnitude?
- Assuming that I do skip pre-emphasis, I don't expect the Fourier transform of the magnitude over time to differ from that of the Fourier transform of the amplitude over time. But am I better of skipping it, or fine-tuning it?
- I should just ignore all of these concerns and just continue to take MFCC's as usual.
Any other ideas?