ComfortableInClouds

Hi - for an installation, I would like to create a patch where the audio manipulation parameters are determined largely by formant frequencies. i was thinking i could perform an FFT analysis within formant ranges, and use the values of the bins as input to my processing parameters. is it possible to perform an fft where the # of bins is spread over a smaller frequency range than 1/2 the sampling rate?

vocal-analysis-fft-within-limited-frequency-range

No, it's impossible, it's the nature of an FFT analysis. You have fft bins spread all over the frequency range, and all frequency bins have equal width.

One idea you could try would be to put your FFT analysis into a Jitter matrix, and to sum (in a weighted way or not) the frequency bins corresponding to each formant zone you are interested in.

No, it's impossible, it's the nature of an FFT analysis. You have fft bins spread all over the frequency range, and all frequency bins have equal width.
One idea you could try would be to put your FFT analysis into a Jitter matrix, and to sum (in a weighted way or not) the frequency bins corresponding to each formant zone you are interested in.


hmm. why is it that you can't have the 512 bins dividing the frequencies from 200-3000 hz? I am not very familiar with mathematics, so I don't understand how the FFT works, but couldn't you offset k and then limit the range, thus have smaller bins within a smaller range?

Vocal Analysis -> FFT within limited frequency range?