What Peak should I use for DynamicRange?

[davies-w]

I'm trying to characterize relative energy of sections of music. One method that has been suggested to me would be to look at the Dynamic Range of each section (although how I should aggregate that I'm not certain either, but thats a seperate issue).

There is an existing RMS function, that operates over a window-size and hop-size, but there doesn't appear to be a corresponding Peak method. Now I realize that in theory one could just take max(y[x:x+w]), but I have a feeling that for music, that the notion of Peak is a little more subtle - and would look at the max variation between some successive troughs and peaks, but maybe not in the same window size as we compute RMS for.

Any advice would be great - I saw there's a Peak-Pick function, but that doesn't seem to have the same parameters as the RMS, and is related to beat/onset detection as far as I can tell.

Apologies if this is not the right category, but it seemed neither a Bug nor a Feature Request!

Happy Thanksgiving!
Winton

[davies-w]

FWIW, I came across PyLoudNorm which does what I wanted (well it identifies the loudest sections at least!).