adamiam

I recently posted on this forum asking about building a MIDI low-pass/high-pass filter, which can sweep through MIDI notes, lowering the ones that it filters like a audio filter does with audio. Simple enough.

The problem is that isn't what an audio filter really does. I've now realised that what I'm really trying to build is a MIDI filter which looks at how frequently MIDI notes occur, filtering the ones that occur more/less often, but is that even possible to create?! I know that an audio filter performs a fast Fourier transform on an audio frame to separate how much of each frequency is occurring, before filtering each accordingly and then reconstructing an audio signal with an inverse FFT, but I wouldn't even know where to begin doing this with MIDI data, or if it's an absurd thought to do so in the first place.

What I'm asking is if it's possible to take a frame of MIDI data, perform something like an FFT on the timing data, filter the data so that more/less frequently occurring keys have more/less velocity when reconstructed via something like an inverse FFT, and if this can be done in real time like a standard audio filter.

I'd greatly appreciate any thoughts or advice on this!

I recently posted on this forum asking about building a MIDI low-pass/high-pass filter, which can sweep through MIDI notes, lowering the ones that it filters like a audio filter does with audio. Simple enough. 

The problem is that isn't what an audio filter really does. I've now realised that what I'm really trying to build is a MIDI filter which looks at how frequently MIDI notes occur, filtering the ones that occur more/less often, but is that even possible to create?! I know that an audio filter performs a fast Fourier transform on an audio frame to separate how much of each frequency is occurring, before filtering each accordingly and then reconstructing an audio signal with an inverse FFT, but I wouldn't even know where to begin doing this with MIDI data, or if it's an absurd thought to do so in the first place. 

What I'm asking is if it's possible to take a frame of MIDI data, perform something like an FFT on the timing data, filter the data so that more/less frequently occurring keys have more/less velocity when reconstructed via something like an inverse FFT, and if this can be done in real time like a standard audio filter. 

I'd greatly appreciate any thoughts or advice on this!
Thanks,
Adam


a-true-midi-fft-filter-impossible

Sounds like you want to make a histogram of played MIDI notes, and to process the MIDI data according to this histogram. All of this is possible to do, actually more easy with MIDI data than with audio. I don't think you need an FFT algorithm to do what you describe.

Here is an excerpt from the object [histo]'s reference. It could be a first step for your project.

Sounds like you want to make a histogram of played MIDI notes, and to process the MIDI data according to this histogram. All of this is possible to do, actually more easy with MIDI data than with audio. I don't think you need an FFT algorithm to do what you describe.
Here is an excerpt from the object [histo]'s reference. It could be a first step for your project.

Ah, that's very exciting indeed, thank you.

A True MIDI FFT Filter Impossible??