Subject: Re: [linux-audio-dev] analysis/resynthesis environment?
From: Andrew W. Schmeder (andy_AT_e-molecule.com)
Date: Tue Aug 06 2002 - 06:15:43 EEST
<quote who="Vincent Touquet">
> On Mon, Aug 05, 2002 at 02:35:33PM -0700, Andrew W. Schmeder wrote:
>>Loris (open source): http://www.cerlsoundgroup.org/Loris/
>> --andy
>
> Has anyone tried Wavelets / Wavelet Packets for
> better time / frequency accuracy ?
My understanding of the theory is that there is a mathematical restriction
which says all analysis methods must admit a certain degree of uncertainty in
time and/or frequency space. Certainty in one space is inversely proportional
to certainty in the other space. This is also known as the heisenberg
uncertainty principle.
Wavelet packets and local cosine methods use an adaptive process to select the
best balance of accuracy depending on the local structure of the signal. This
is good for compression but cannot actually improve the simultaneous
time-frequency accuracy.
The authoritative source on all such matters is Stephane Mallat's "A Wavelet
Tour of Signal Processing".
If you want to play with such things, try Lastwave.
(http://www.cmap.polytechnique.fr/~bacry/LastWave/). In particular the gabor
atom transform is interesting from the analysis/resynthesis perspective.
--andy
-----------------------
Andrew W. Schmeder
andy_AT_e-molecule.com
This archive was generated by hypermail 2b28 : Tue Aug 06 2002 - 06:21:55 EEST