DSPRelated.com
Forums

spectral accumulation using phase vocoder

Started by Unknown August 10, 2017
Hi!

I've made a simple freeze effect with a phase vocoder that can 'freeze' the sound by repeatedly converting the same spectral frame of that sound to the time domain, incrementing the phases each time with the phase difference calculated from that frame and the previous frame.

This works fine. What I would like to do though, is take another freeze frame out the incoming sound and mix it in the frequency domain, calculate some sort of phase difference to add to successive frames and resynthesize the whole lot with one phase vocoder pass. 

Is this possible? I've tried summing the bins of both freeze frames and summing the phase differences but that definitely doesn't work, but it made sense to me. 

Any ideas?

cheers


Marek
On Thursday, August 10, 2017 at 6:22:55 AM UTC-4, Marek Bereza wrote:
> Hi! > > I've made a simple freeze effect with a phase vocoder that can 'freeze' the sound by repeatedly converting the same spectral frame of that sound to the time domain, incrementing the phases each time with the phase difference calculated from that frame and the previous frame. > > This works fine. What I would like to do though, is take another freeze frame out the incoming sound and mix it in the frequency domain, calculate some sort of phase difference to add to successive frames and resynthesize the whole lot with one phase vocoder pass. > > Is this possible? I've tried summing the bins of both freeze frames and summing the phase differences but that definitely doesn't work, but it made sense to me. >
what do you mean by "mix it in the frequency domain"? what are you mixing or intending to mix? r b-j
Hi

Thanks for your help! I have 2 phase vocoders that keep overlap-adding the same 2 spectral frames repeatedly, then mixing the two signals in the time domain to make it sound as if the 2 sounds are frozen

I would like to mix those 2 frames whilst they are in the frequency domain before passing it to the phase vocoder to convert back to the time domain so as not to have to do 2 passes of the overlap-add. So it would sound the same as having 2 phase vocoders doing the overlap-add but be more computationally efficient. Hope that makes sense.



On Friday, 11 August 2017 07:37:38 UTC+1, robert bristow-johnson  wrote:
> On Thursday, August 10, 2017 at 6:22:55 AM UTC-4, Marek Bereza wrote: > > Hi! > > > > I've made a simple freeze effect with a phase vocoder that can 'freeze' the sound by repeatedly converting the same spectral frame of that sound to the time domain, incrementing the phases each time with the phase difference calculated from that frame and the previous frame. > > > > This works fine. What I would like to do though, is take another freeze frame out the incoming sound and mix it in the frequency domain, calculate some sort of phase difference to add to successive frames and resynthesize the whole lot with one phase vocoder pass. > > > > Is this possible? I've tried summing the bins of both freeze frames and summing the phase differences but that definitely doesn't work, but it made sense to me. > > > > what do you mean by "mix it in the frequency domain"? what are you mixing or intending to mix? > > r b-j