That's a wonderful explanation and exploration of wavelets Raul. In particular I think your diagrams of the analysis and resynthesis processes are very good.
While the idea of wavelet resynthesis isn't new the transformations you're showing in chapter 3 seem insightful and original. And they're in Puredata! :) Your patches make this process much more accessible to the rest of us. Thankyou very much!
A couple of refs I think you might have missed and may want to look at are
Grossman and Morlet 1984 - decomposition of Hardy functions Gianpaolo Evengelista 1991 - Wavelet transforms we can play (although I think that one is already in De Poli)
A question regarding one part I still don't understand. Are you saying it's only possible to manipulate time in powers or two, like ... 1/8, 1/4, 1/2, 1, 2, 4, 8 ...? Or is that a limitation of the implementation (your stretch function).
Also, what are the possibilites for interpolating between wavelet coefficients taken from different sounds as a kind of cross-synthesis? (rather than merely mixing grains taken from different sources). It seems well suited to this, but I'm guessing because I've never worked with this method practically.
Great work.
Andy
On Fri, 26 Jan 2007 13:58:21 +0100 "raul diaz" raul.lete@gmail.com wrote:
Hi list,
I want to present some information at my PD site about my Diploarbeit at IEM.
[ The aim of this work is to research the possibilities of use the discrete wavelet transform in real-time with Pure Data for analysis / resynthesis of audio signals. With this intention I approached to a new sort of granular synthesis with Pure Data based on wavelet transform: resynthesis of an audio signal from its wavelet coefficients by means of an additive synthesis of wavelet streams (a temporal succession of 'wavelets grains' for each scale or frequency level which are enveloped by an amplitude factor obtained from wavelet analysis coefficients). Another audio applications by means of wavelet coefficients manipulation have been developed to study the possibilities of pitch shift, time stretch, equalization and audio randomization. ]
I have thought it could be interesting for people who is interested in wavelet transform and audio processing. Any criticism and feedback is welcomed. I seized the opportunity to thank Winfried Ritsch, for all their help and support during the course of my research at IEM, and also I thank Johannes Zmölnig and Thomas Musil for their help in PD programming.
--> https://puredata.org/Members/raul
Regards.
-- Raul Diaz Poblete
raul.lete@gmail.com Ciudad Real [Spain]