dada.analysis energy & spread

[fcd]

Hello!
I am building a patch that analyses a roll in many different ways, and I am not sure to understand perfectly what [dada.analysis.energy] and [dada.analyisis.spread] relates to.

It is obvious that these objects are in the continuation of ircam descriptors, like in Zsa or CataRT.
https://discussion.forum.ircam.fr/t/sou ... nces/62803

I understand well what it means in an audio spectrum, but I am not sure to get the logic of the "symbolic" interpretation in a score (no sound / timbre = no spectrum), only with velocity and fondamental frequencies.

1) Energy
If I enter one note with velocity 100, it returns 100.
If I delay this note from the begining of the roll, the value decrease.
If I enter another note of velocity 100 on top of the first one, it returns 200, but if delay a bit the onsets, the value decrease.
If I enter two succedding notes with velocity 100, it returns 80, but if change the position of the second note, spacing the two more and more, the value decrease.

It appears that the energy is like a "musical" energy because it takes silence and time in consideration (if I got it well, it would be nice to add this in the helper!). Then, can we still consider the output value of this object as "velocity average" ?

2) Spread
The same for the spread, even with only two notes, the results are very different from an interval "ambitus". In spectral descriptors, the spread is linked to the centroid. There is dada.analysis.centroid, but still, how do you calculate a symbolic centroid ?
It would be great to have more informations about the calculation of it !

Thank you

[danieleghisi]

Hello, if memory doesn't trick me (I am away from the computer), energy is the average velocity, weighted by note duration. So, yes, silences will lower the total.
Spread should be either the variance or the standard deviation of pitches, perhaps also weighted by velocity/duration. I don't remember the ultimate details but if you double click on the module (it's an abstraction) it should be fairly easy to read them from the patch.

(I agree that the naming that recalls sound-based property may be puzzling. I wouldn't make the same choice now.)

[fcd]

Thank you Daniele, I didn't checked if it was an abstraction or not, I will explore it further.
About the naming, as long it is specified as "symbolic" in the helper I don't thinks it's a problem. But an additionnal line about the way it works could help!
++