Forums > MaxMSP

Predictive fft~


div
November 14, 2007 | 3:35 pm

Hi all,

I am investigating the posibilities of using ftt to predict sound. I would love your thoughts on the best way to do this. I guess it would be great to specify the analysis window size and then how far that moves into the future. I am pretty sure fft is the way to go.

Has anyone heard what this sounds like?

Best

Tim


November 14, 2007 | 3:40 pm

Wow!! I will be checking this topic! Predicting sound, amazing..

( maybe, one day, max will be able to bend the time space continuum.. ;) )


November 16, 2007 | 12:03 am

Timothy Devine schrieb:
> I am investigating the posibilities of using ftt to predict sound. I
> would love your thoughts on the best way to do this. I guess it
> would be great to specify the analysis window size and then how far
> that moves into the future. I am pretty sure fft is the way to go.

I am not sure what you are after, there is a philosophical problem with
predicting the future… ;-)

I think the Axxon pitch detectors used neural nets to predict the pitch
before it actually is detectable with other methods. I guess they used
both, the fft analysis and the timestructure of the envelope and matched
it to a database of guitar sounds…

Stefan


Stefan Tiedje————x——-
–_____———–|————–
–(_|_ —-|—–|—–()——-
– _|_)—-|—–()————–
———-()——–www.ccmix.com


November 16, 2007 | 9:03 pm

While "predicting" or at least "guessing" the future can be quite challenging, it’s still possible to first record something, and second do the analysis, so that you know exactly what you are analyzing. Then I agree it’s not realtime. Latency could be a possible trade-off …


Viewing 4 posts - 1 through 4 (of 4 total)