Anastasis Panayi

I want to make a segmentation to a sound wave using these parametres

and then extract from each segment these sound features

I think that each max object makes it"s own segmentation and it is standard.

But i want the same segmentation for all objects.

My problem is that i cannot control the segmetntation parametres (winsize, step).

Hello,
I want to make a segmentation to a sound wave using these parametres
WinSize : 0.015ms
Step : 0.005ms
and then extract from each segment these sound features
1) mfcc
2) F0 (pitch)
3) Formants (f1, f2, f3, f4, f5)
4) Amplitude
5) Intensity
6) Energy

I think that each max object makes it"s own segmentation and it is standard.
But i want the same segmentation for all objects.

My problem is that i cannot control the segmetntation parametres (winsize, step).
I want to be able to control them.

segmentation-for-features-extraction

If you're not getting the response you want, maybe you're not asking the question the right way. Could you describe what you are trying to analyze, why you want to do it in Max, why you need these very specific settings, as well as what you've already tried? Have you investigated the objects that people suggested here the last time you posted this?

https://cycling74.com/forums/segmentation-and-feature-extraction/

I'm not going to tell you can't do this task with Max but it would be much easier to use one of the dedicated MIR toolkits such as

You might even be able to control them from Max via scripting, particularly in the case of Pymir. There are a couple of projects that you can find in with a forum search that will let you connect to Python.

Also, is there a particular reason why it has to be these particular settings? 0.015ms at 44.1kHz is less than a sample. 15 ms would still only be 661.5 samples, and would require padding in order to reach a valid FFT size.

If you're not getting the response you want, maybe you're not asking the question the right way.  Could you describe what you are trying to analyze, why you want to do it in Max, why you need these very specific settings, as well as what you've already tried?  Have you investigated the objects that people suggested here the last time you posted this?
https://cycling74.com/forums/segmentation-and-feature-extraction/

I'm not going to tell you can't do this task with Max but it would be much easier to use one of the dedicated MIR toolkits such as 

Marsyas: http://marsyas.info/about/overview.html (free)


MIRToolBox: http://www.mathworks.com/matlabcentral/fileexchange/24583-mirtoolbox (free, but requires MATLAB)


Pymir: https://github.com/jsawruk/pymir (free)

You might even be able to control them from Max via scripting, particularly in the case of Pymir.  There are a couple of projects that you can find in with a forum search that will let you connect to Python.

Also, is there a particular reason why it has to be these particular settings?  0.015ms at 44.1kHz is less than a sample.  15 ms would still only be 661.5 samples, and would require padding in order to reach a valid FFT size.


I am really sorry! I didn't mean to offend anybody and also i made a mistake. I wrote 0.015ms else 0.015sec (0.015 sec = 15 ms), my bad.

I checked the answers the others gave me and they were very useful but not exactly what i wanted.

Thank you very much for your respond. It was really helpful.

I am really sorry!  I didn't mean to offend anybody and also i made a mistake. I wrote 0.015ms else 0.015sec (0.015 sec = 15 ms), my bad.
I checked the answers the others gave me and they were very useful but not exactly what i wanted.
Thank you very much for your respond. It was really helpful.


Segmentation for features extraction