Julien Bayle

I'd need one value per frequency band with 3 basic frequency bands.

biquad + average~/snapshot is basic and nice.

could we do better with zsa.descriptor for instance ?

I need only 3 bands (and very basically some kind of low/mid/high)

Hi there,
I'd need one value per frequency band with 3 basic frequency bands.

could we do better with zsa.descriptor for instance ?
I need only 3 bands (and very basically some kind of low/mid/high)


analysis-per-frequency-band-with-zsa-or

what about svf~? even cheaper. Or go the FFT route?

I'd probably use biquad though, most simple and flexible. I'd usually do my own evelope follower though: abs~+slide~(or some other filter)+aToDb

what about svf~? even cheaper. Or go the FFT route?
I'd probably use biquad though, most simple and flexible. I'd usually do my own evelope follower though: abs~+slide~(or some other filter)+aToDb


filtering actually gives approximations (filter aren't perfect and have a x db/oct)

fft route is be the way I guess... digging.

filtering actually gives approximations (filter aren't perfect and have a x db/oct)
fft route is be the way I guess... digging.


Well FFT certainly isn't perfect either, right? What are you trying to achieve?

using fft at least gives normalized value of energy which is ... cool considering the other methods.

using fft at least gives normalized value of energy which is ... cool considering the other methods.
right?


There's also analyzer~. You could take the peak, or the average, of a few of the bark scale outputs.

analysis per frequency band with zsa or ?