keep tracking of attacks using spectral data with analyzer~
Nov 19 2011 | 3:37 pm
hello everyone
I am making histogram analysis of brightness and loudness values using analyzer~.. the curves contain some noise.. which might derive from the noise parameter itself of the spectral data information.. so I need to subtract the noise of the analyzer~ output to brightness and loudness.. and then cooking an ad-hoc algorithm to determine, within variable ranges, whenever I have an ascending/descending curve.. I need this to make analaysis of attacks of extremely complex sounds in a spectral domain, like a violin or viola.. which contain very complex and rich spectral information and usually very linear amplitude curves, mainly if you are talking about detache or legato bowed articulatons.. so If I can do this.. probably my problem will be solved.. if you can help me with this.. I would thank you a lot, guys if you can help me.. I already implemented a dynamic treshold.. coming from some math operations on top of amplitude measuring, but still not as accurate as I need..
the patch I've done until right now follows bellow
and the histogram tool that I am using follows bellow
thanks in advance
tiago morgado
- Nov 19 2011 | 6:00 pmI was thinking about to possible situations..one would be compressing the audio signal a little bit, in order to give more punch to the audio signal itself and make it easier to track..another one would be using gestural following to decode the oscillatory movements, but there must be easier ways of accomplishing it..
- Nov 19 2011 | 7:31 pmI think it would be better if I could make tracking of the oscillation movement.. any ideas about that?
- Dec 07 2011 | 2:48 amsolved it with inertia+energy tracking only with analyzer~ (which means that I didn't keep track of the energy with gbr.yin from ircam, and implemented it from the ground) - the object is quite unstable.. and the good version of it.. yin.. is paid and very expensive if you buy it with ircam forum stuff