Although the wavelet transform conceptually doesn't fit a streaming
model (for exactly the reason Michael points out), lifting scheme
implementations of the transform do operate on streams.  If I were
working on this, I would consider an implementation in which the
analysis process produced several output streams, then have another
function that would take those output streams and buffer them to
create complete cells of the desired size.

In my experience, trying to modify an audio signal based on wavelet
analysis data is problematic because the resynthesis is very sensitive
to changes in the data.  I've done interpolations between different
sounds based on wavelet analyses, which worked very well, but pitch
shifting and time stretching were not satisfactory.  Although I
haven't tried this myself, I've read some articles that indicate a
dual-tree wavelet transform greatly reduces the artifacts present
after resynthesis.  I could try to dig those up if anyone is actually
interested in implementing this.

For the record, my wavelet experiments were all in Python.  If I
weren't swamped with other projects now, I'd be tempted to try
implementing this myself.
John Lato

On Tue, May 20, 2008 at 5:35 PM, Michael Gogins <gogins@pipeline.com> wrote:

> The wavelet case is much more difficult, and it does not so readily lend
> itself to streaming signal types. Notionally, the wavelet cells increase in
> duration as they decrease in frequency. So, either there is a uniform
> wavelet analysis with the lowest cell lasting the duration of a sample, or
> there is a longest cell. However, even in that case, there still may be
> avantages to the wavelet analysis/resynthesis because it potentially affords
> more control over tradeoffs between time precision and frequency precision,
> artifacts with transients, and so on.

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Csound-devel mailing list
Csound-devel@lists.sourceforge.net