Forums > MaxMSP

Advanced audio analysis?


dtr
June 11, 2009 | 12:11 pm

Hi

I’m working on a project where audio analysis (not necessarily real-time) feeds parametric 3d models. I’m wondering if anyone has done or knows about audio analysis which goes further than the straightforward per-sample frequency and amplitude data.

I’m thinking of pitch, rhythmicallity (does this word exist?), texture (as in harsh/noisy versus smooth/soft), etc.

I’m aware of the fiddle, bonk and pitch externals which are candidates for developing a more complex analysis system.

I think a lot can be done just by analyzing frequency and amplitude over time. Any hints on algorithms are most welcome.

Tanx! Dieter


June 11, 2009 | 1:00 pm

http://web.media.mit.edu/~tristan/maxmsp.html

a few more here… not sure if it’s what you’re after.



dtr
June 11, 2009 | 1:09 pm

nice! gonna check m out


June 11, 2009 | 1:20 pm

Check out our MEAPsoft package:

http://www.meapsoft.org/

not directly max/msp, but a lot of interesting analyses. Java source code is downloadable.

brad
http://music.columbia.edu/~brad



dtr
June 11, 2009 | 2:00 pm

tanx man, a chunk-wise analysis like that is one of the ideas spinning in my head at the moment


June 11, 2009 | 3:14 pm

These are new, and I have no idea if they will apply to you…. but they might!

http://www.e–j.com/sphpblog/static.php?page=zsa_overview



dtr
June 11, 2009 | 4:18 pm

hey that s some cool new ej’s!


June 12, 2009 | 1:12 pm

You might also want to look into FTM and Gabor, which have some tools you might find useful:

http://ftm.ircam.fr/index.php/Main_Page

M



dtr
June 13, 2009 | 7:42 pm

anyone managed to get pitch data out of more complex compositions (eg. pop song)? fiddle and analyzer only seem to work with very clean sounds. (i do realise that s easier said than done…)


June 13, 2009 | 9:09 pm

check my thread on the jitter forum, theres some useful things on there, ones I made and a link someone posted to an amazing one that works with polyphonic material.

Pop will be touch because its rhythm based, but you could probably adapt my fffb based method to work quite easily.

(its mainly just an fffb~ with frequencies for each area you want to analyze and cranked up reso and gain.)


Viewing 10 posts - 1 through 10 (of 10 total)