| Need help with Wavelet Workbench [message #14836] |
Wed, 07 April 1999 00:00  |
jkbishop
Messages: 1
Registered: April 1999
|
Junior Member |
|
|
I'm trying to use Wavelet Workbench on a long (48000 pt) signal. I
think that two separate problems are occuring.
I upsampled the data set to 65536 points (by zero padding in frequency
space). I hacked wreaddat, wdyadlng, wdyad, and wfwtpo to use long
integers in some places. The result is that I can now plot the
scalogram for my data set (wreaddat, wintwave, wdoscog are the
programs I'm calling). However, the plot of the scalogram looks like
only the first half of the data set is being used. The coarser scales
have some variation just beyond the half-way point (bleed-over from
the convolution process?), but the more detailed scales show a solid
color in the upper half of the time axis. Anyone have any ideas what
is going on?
So far, I have tried upsampling again to 2*65536 points (whatever
that is). The result is that the convolution-by-FFT process in
wmfilt takes forever (I didn't wait for it to finish; it was taking at
least 10 times as long as the 65536 point set, as verified by
printed status statements). I don't understand why, but would
the FFT process be the problem with the 65536 data set?
To get an answer, my next approach was to downsample to
32768 points. This brings me to the second problem, which is a
type conversion problem in wintwave. For the 65536 point data
set, this part seems to work fine (with the adjustment of 2 to 2L in
the line that finds n_work). When I put the 32768 point set in,
the data set gets truncated to 16384 points because
fix(alog(n_elements(x_work))/alog(2))) evaluates to 14 instead of
15. alog(n_elements(x_work))/alog(2)) is given as 15.0000. Can
someone explain this so even a mechanical engineer can
understand? To get around this one, I have just inhibited the
length check; I just have to be careful to only use dyadic length
sets.
Any help would be appreciated.
--
Jonathan Bishop
jkbishop at frontiernet dot net
jab7981 at rit dot edu
|
|
|
|
| Re: Need help with Wavelet Workbench [message #14966 is a reply to message #14836] |
Thu, 08 April 1999 00:00  |
Amara Graps
Messages: 24
Registered: June 1996
|
Junior Member |
|
|
Jonathan Bishop wrote:
>
> I'm trying to use Wavelet Workbench on a long (48000 pt) signal. I
> think that two separate problems are occuring.
>
> I upsampled the data set to 65536 points (by zero padding in frequency
> space). I hacked wreaddat, wdyadlng, wdyad, and wfwtpo to use long
> integers in some places. The result is that I can now plot the
> scalogram for my data set (wreaddat, wintwave, wdoscog are the
> programs I'm calling). However, the plot of the scalogram looks like
> only the first half of the data set is being used. The coarser scales
> have some variation just beyond the half-way point (bleed-over from
> the convolution process?), but the more detailed scales show a solid
> color in the upper half of the time axis. Anyone have any ideas what
> is going on?
This scalogram is from the Matlab Wavelab code. I translated it
as is. It is pretty strange code to me, and I never went back and did
anything with it to make more sense of the results and display.
The idea is that you should see another representation of the values
of your wavelet coefficients at each 2^n scale.
The discrete wavelet transform results are much more intuitive to
me, and that's what I rely on, instead.
Feel free to hack up that scalogram code .... (I don't know when
I will get around to working on that part.)
Also the scalEgram (note the "e" difference) calculates the
absolute power of the wavelet coefficients, (but I have
to use a Haar wavelet for integrating over the scale, see the
notes to the software about why I (and the WaveLab folks) do it that
way). So perhaps that scalegram would be a more useful representation
of your transformed coefficients.
Hope this provides some insight,
Amara
--
************************************************************ *****
Amara Graps | Max-Planck-Institut fuer Kernphysik
Interplanetary Dust Group | Saupfercheckweg 1
+49-6221-516-543 | 69117 Heidelberg, GERMANY
Amara.Graps@mpi-hd.mpg.de * http://galileo.mpi-hd.mpg.de/~graps
************************************************************ *****
"Never fight an inanimate object." - P. J. O'Rourke
|
|
|
|
| Re: Need help with Wavelet Workbench [message #14968 is a reply to message #14836] |
Thu, 08 April 1999 00:00 |
Amara.Graps
Messages: 11
Registered: November 1998
|
Junior Member |
|
|
In article <370b984c.55748962@news.rit.edu>, jab7981@nospam.rit.edu wrote:
>
> The result is that WWB only works for datasets < 32768 in length
> rather than <= 32768 in its original form.
>
Dear Jonathan Bishop,
I can answer this, and I think t's time to say something about Wavelet
Workbench (WWB).
I'm the author of WWB. About 2/3 of WWB is code that I translated
with (D. Donoho's) permission from the Matlab WaveLab algorithms
and software, and the rest of WWB is my algorithms and software.
I devloped WWB in IDL v4.0.1 on my old Macintosh.
Yes, WWB is old, and yes I have access to IDL version 5 on
a Sun at my Institute, but my time is severely limited for me
to spend much time away from my PhD while I'm at my Institute, in
order to develop that software for the public.
Yes, I understand that that public version of WWB is crippled in many ways.
I was hoping that many folks would see the code of the open routines and
upgrade and develop it themselves. I know of some people that have
made the necessary changes to have WWB run for arrays > 32767 and on
newer version of IDL. (It's not a trivial change, but it is doable.)
In the last 1 1/2 - 2 years that I released that Wavelet Workbench
publically, I've added a fair bit of new code to my own copy of that
software, but mostly to support my own research work on Galileo data.
The extra bit of work for me to update the WWB manuals and to make more
examples has been prohibitive enough to prevent me from making
a new release of Wavelet Workbench. And I've had substantive changes
in my real world that have stopped me from finding time to do such
things. (Try moving to a foreign country, where everything is well..
"foreign" and becoming a graduate student.. that will put a major crimp
on one's life) My free time every day is on the order of minutes ...
Yes, I have made the changes to WWB to work on >32 767 size arrays.
Yes, I've doubled the number of wavelet bases available.
Yes, I've added a continuous wavelet transform. (I'm adding
wavelet-ridge-finding for intantaneous frequencies now.)
Yes, I've fixed quite a few little bugs here and there.
Yes, I have a number of very nice new examples and excercises, from
my own work, as well as the result of other people's use of WWB, but
they are not written up in the manual yet.
No, I have not worked on any "speed" issues, to speed up WWB.
Since I'm in the middle of a large number of big changes with the software,
I don't wish to make a public release now.
Would you be interested in a version from last summer that solves
that "32,767" problem? That older version of WWB was "stable" (for me,
that is) for some time before my current frenzy of modifications of
the code. That's the best I can offer right now.
I'll put it at my personal server this evening:
ftp://ftp.amara.com/wwb_winter.tar.gz
And I will leave it up until April 12. Anyone else that wants a little bit
newer version of the WWB code can grab it too. (Sorry I can't leave it
up longer, but every Internet usage costs me, and I'm living on a PhD
stipend you see...)
Sincerely,
Amara
P.S. I'll be on travel for a large part of April. Sorry I can't do more!
--
************************************************************ *****
Amara Graps | Max-Planck-Institut fuer Kernphysik
Interplanetary Dust Group | Saupfercheckweg 1
+49-6221-516-543 | 69117 Heidelberg, GERMANY
Amara.Graps@mpi-hd.mpg.de * http://galileo.mpi-hd.mpg.de/~graps
************************************************************ *****
"Never fight an inanimate object." - P. J. O'Rourke
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
