I'm doing a research project involving phase vocoders and looking at
09.pvoc.pd in the example patches as a reference implementation. I was
wondering if anyone could help me to understand it. Specifically, how
is the "instantaneous frequency", obtained by comparing phases in
successive frames, used? What is the relationship between the two
segments of the source array that are read at the same time and the
phase-real, phase-imag arrays?
I basically understand the procedure outlined in Computer Music Tutorial
(Roads) on pages 1105-6:
1. read M samples of input signal
2. window
3. zero-pad the windowed frame (this seems to be absent from the patch)
4. FFT
5. convert from rectangular to polar, differentiate the phase to obtain
instantaneous frequency
6. apply any modifications (time-stretch, pitch-shift, etc)
7. IFFT
8. window again if phase spectrum was altered
9. overlap-add the output frames
but step 6 is sketchy, and I don't understand how the instantaneous
frequency is actually *used* in resynthesis. Any help would be
appreciated.
Thanks,
Ben