Post #1659180
2026-04-12 19:25 UTC
@nnirror @williamfields This is a good paper but I'm not sure I would have gone with a discrete compression algorithm for a continuous signal. Windowed RMS -> STFT of RMS signal -> spectral flatness or spectral entropy might be easier to implement and more CPU-friendly with similar results.
Replies (1)
-
@nnirror@post.lurk.org 2026-04-12 19:46
@ho @williamfields oh that’s interesting, i follow. yeah, i’ve been able to get a decent implementation in gen~ which I think compiles to c++ under the hood and can handle realtime with reasonable cpu load, but definitely having potentially scores of iterations per window is questionable on cpu usage. it’d be trickier to do the STFT in gen~ but i could see that being a good approach with python or something else where you don’t have to roll your own FT code