Analysis global understanding and pca

Hi there,

flucoma is just… magic.
I already started to integrate it in some parts of my teaching with Structure Void (we got a lot of professional people wanted to learn Max here)


I understand the idea of fluid.pca~
Multidimensional datasets → reduced (projected) to 2-dimension space using the “possibly most significant part of the dataset”

In this example, there is a mfcc analysis. It renders 13 parameters, ok.

How could I do exactly the same but with for instance 3 or 4 parameters ?
The thing is not so related to fluid.pca here, but I’d like to have an example for doing exactly the same kind of sound’s fragments splitting according to other critera.

For instance, I want to segment my buffer according to loudness mean and pitch mean.
I’d have, in the end, a 2D-map (the plotter) with on x-axis the loudness, and on y-axis the pitch.

Or, for instance, I’d like to segment according to transients (with a defined level) and then to split these segments and to get them on the plotter like: x-axis noisiness analysis (spread, I mean) and y-axis centroid.

Would you help me to do just one of them ? I guess I could extend it to do the other one.

Any helps would be very appreciated here :slight_smile:

Hi Julien

I had the same question a while ago and found the answer in the FluCoMa video tutorials ( 2D Corpus Explorer) - the relevant one is Part3 (Building a 2D Corpus Explorer (Part 3) - YouTube) where James Bradbury demonstrates plotting loudness against central spectroid - see @12m39s and on.

I noticed that the version of the patch in the FluCoMa tutorials is slightly different than the one in the video so if you want to plot loudness against pitch, then watch the video to see the exact code.

Hope this helps


1 Like