[Home]
[Back Issues]
[Search]
[Calendar]
[Vol 1]
[Vol 2]
[Vol 3]
[Vol 4]
[Vol 5]
[Vol 6]
[Vol 7]
The Wavelet Digest, Volume 7, Issue 7 (July 21, 1998)
Today's Editor: Wim Sweldens
Lucent Technologies, Bell Laboratories
wim@bell-labs.com
Today's Topics:
1: Preprint: Test for Homogeneity of Variance
2: Preprint: Convergence rates of multiscale and wavelet expansions.
3: Preprint: Two preprints on Wavelet Transform in Chemistry
4: Preprint: Nonlinear Functionals of Wavelet Expansions
5: Preprint: Preprints concerning the shiftability problem
6: Thesis: The analysis of electromyography using wavelets
7: Contents: Numerical Algorithms 17: 1,2
8: Contents: Advances in Computational Mathematics 8-4
9: Answer: Wavelets and financial time series (WD 7.6 #26)
10: Question: experience with multiscale-edges-based image processing?
11: Question: Wavelet theory and field analysis.
12: Question: Definition of edges
13: Question: Wavelet Packets and Density Estimation
14: Question: Normalisation of the sinc wavelet filter bank
Current number of members: 12141
E-mail:
General help: help@wavelet.org
Add yourself as a member: add@wavelet.org
Remove yourself as a member: remove@wavelet.org
Publish in the next WD: publish@wavelet.org
Receive publishing policy: policy@wavelet.org
Visit us on the Web at http://www.wavelet.org/
Calendar of events:
Jul 30-Aug 7: Workshop on Self-Similar Systems, Dubna, Russia
Sep 7-11: Multivariate Approximation and Interpolation, Israel
Sep 28-30: Applied Computational Electromagnetics, Penn State
Oct 4-9: School on Wavelets in the geosciences, Delft, Netherl.
Oct 5-8: 10th GAMM-Workshop on Multigrid Methods, Bonn, Germany
Oct 27-28: Wavelets in Biology and Medicine, Hong Kong
Feb 24-25: Royal Society London, UK
Feb 24-26: IEEE Information Theory Workshop, Santa Fe, NM
---
Mar 22-26: II Symposium on Artificial Intelligence, Havana, Cuba
Jul 1-7: Curves and Surfaces, St-Malo, France
Date: Wed, 1 Jul 1998 21:27:52 -0700 (PDT)
From: Brandon Whitcher <brandon@stat.washington.edu>
Subject: #1 Preprint: Test for Homogeneity of Variance
A preprint is currently available on testing homogeneity of variance
in the presence of long-range dependence. It has been submitted for
publication.
Title: Testing for Homogeneity of Variance in Time Series: Long
Memory, Wavelets and the Nile River
Authors: B. Whitcher, S. D. Byers, P. Guttorp and D. B. Percival
Abstract:
We consider the problem of testing for homogeneity of variance in a
time series that has long memory structure. We demonstrate that a test
whose null hypothesis is designed to be white noise can in fact be
applied, on a scale by scale basis, to the discrete wavelet transform
of long memory processes. In particular, we show that evaluating a
normalized cumulative sum of squares test statistic using critical
levels appropriate for the null hypothesis of white noise yields
approximately the same null hypothesis rejection rates when applied to
the discrete wavelet transform of samples from a fractional difference
process. The point at which the test statistic, using the maximal
overlap discrete wavelet transform, achieves its maximum value can be
used to estimate the time of the unknown variance change. We apply our
proposed test statistic on a time series of Nile River yearly minimum
water levels covering the years 622 to 1284 AD. The test confirms an
inhomogeneity of variance at short scales and identifies the change
point around 720 AD, which coincides closely with the construction of
a new device in 715 AD for measuring Nile River water levels.
It is available via http://ftp.statsci.com/pub/longmem/lmwnr.ps
Brandon
Brandon Whitcher
University of Washington, Department of Statistics
Box 354322, Seattle, WA 98195-4322, USA (206) 685-3027
brandon@stat.washington.edu http://www.stat.washington.edu/brandon/
Date: Thu, 02 Jul 1998 12:34:37 +0200
From: Mark Kon <mkon@hydra.mimuw.edu.pl> (by way of Mark Kon <mkon@hydra.mimuw.edu.pl>)
Subject: #2 Preprint: Convergence rates of multiscale and wavelet expansions.
Title: Convergence rates of multiscale and wavelet expansions, parts I and II
Authors: Mark A. Kon and Louise A. Raphael
Abstract: Several results are proved which characterize the rate at
which wavelet and multiresolution expansions converge to functions in
a given Sobolev space in the supremum error norm. Some of the results
are proved without assuming existence of a scaling function in the
multiresolution analysis. Necessary and sufficient conditions are
given for convergence at given rates in terms of behavior of Fourier
transforms of the wavelet or scaling function near the origin. Such
conditions turn out in special cases to be equivalent to moment
conditions and other known conditions determining convergence rates.
Preprint: Available at http://math.bu.edu/people/mkon/
Email: mkon@math.bu.edu
Mark A. Kon
Department of Mathematics
Boston University
Boston, MA 02215
USA
Tel: 617-353-9549
Fax: 617-353-8100
Email: mkon@math.bu.edu
Street Address: 111 Cummington St., Room 259
1996-7: Fulbright fellow at University of Warsaw (above email address
can be used - mail is forwarded).
Date: Wed, 15 Jul 1998 00:43:28 +0900
From: Zhang Maoshen <mszhang@dchp.chp.ustc.edu.cn>
Subject: #3 Preprint: Two preprints on Wavelet Transform in Chemistry
Dear Editor,
I'd like to submit two peices of preprints on Wavelet Transform in
Chemical application.
Paper one:The determination of the shift of binding energy of XPS by
the aid of Wavelet Transform
Authors:Ding Hong-bo Pan Zhong-xiao and Cai Wen-sheng(Department of
Applied Chemistry,University of Science and Technology of
China,Hefei,230026)
Abstract:Wavelet Transform was successfully applied to aid in solving
the binding energy shift of XPS,so as that it's easier to
qualitatively analyse the component of lithium-containing conversion
coatings on aluminium alloy.Based on the characteristic of Wavelet
Transform,we expect that it can also be used in the quantitative
analysis of XPS.
Paper two:The analysis of "Electrochemical Noise"
signal by the use of Wavelet Transform
Authors:Ding Hong-bo Pan Zhong-xiao(address:as that of Paper one)
Zheng Wei-xi Yu Xing-zeng(Fujian Institute of Research on the
Structure of Matters,Chinese Academy of Sciences,Xiamen,361012)
Abstract:"Electrochemical Noise"method is easy to appy in the
measurement of corrosion of metals,while it's difficult to extract
information from the testing results because of the unsteady
phenomenon.We used Wavelet Transform to filter out the instrumental
noise from the original signal,and used "Noise Resistance" method to
determine the rate of metal corroding from the treated signal.
Keywords:Wavelet Transform,XPS,shift of binding energy,Electrochemical
Noise,Noise Resistance Email:hb_ding@bbs.ustc.edu.cn or
mszhang@dchp.chp.ustc.edu.cn phenomenon.
Date: Wed, 15 Jul 1998 17:30:26 +0000
From: "Angela Kunoth" <kunoth@igpm.rwth-aachen.de>
Subject: #4 Preprint: Nonlinear Functionals of Wavelet Expansions
Nonlinear Functionals of Wavelet Expansions
- Adaptive Reconstruction and Fast Evaluation
Wolfgang Dahmen, Reinhold Schneider, Yuesheng Xu
Abstract:
This paper is concerned with the efficient evaluation of nonlinear
expressions of wavelet expansions obtained through an adaptive
process. In particular, evaluation covers here the computation of
inner products of such expressions with wavelets which arise, for
instance, in the context of Galerkin or Petrov Galerkin schemes for
the solution of differential equations. The central objective is to
develop schemes that facilitate such expansions at a computational
expense exceeding the complexity of the given expansion, i.e., the
number of nonzero wavelet coefficients, as little as possible. The
following issues are addressed. First, motivated by previous
treatments of the subject, we discuss the type of regularity
assumptions that are appropriate in this context and explain the
relevance of Besov norms. The principal strategy is to relate the
computation of inner products of wavelets with compositions to
approximations of compositions in terms of possibly few dual
wavelets. The analysis of these approximations finally leads to a
concrete evaluation scheme which is shown to be in a certain sense
asymptotically optimal. We conclude with a simple numerical example.
IGPM-Report 160, May 1998, RWTH Aachen
http://www.igpm.rwth-aachen.de/~dahmen/rep_98.html
Date: Mon, 20 Jul 1998 10:10:34 +0800
From: qiaowang <qiaowang@seu.edu.cn>
Subject: #5 Preprint: Preprints concerning the shiftability problem
We have three preprints concerning the shiftability problems of
wavelet bases and more general configuration. The following are the
abstracts of them. If you need the preprints, please send me an e-mail.
Qiao WANG e-mail: qiaowang@seu.edu.cn
[1] The shiftability of some wavelet bases. by Qiao WANG (submitted)
The wavelet orthonormal bases lack the shiftability. We construct in
this paper a metrical functional for the shiftability of wavelet
bases, where we say $\psi$ is shiftable if for every real $s$,
$\psi(x-s)\in Span \{ \psi(x-n);n\in \Z\}$, and the functional is $$
r(\psi)={1\over 2\pi}\int^{+\infty}_{-\infty}|\widehat \psi(\xi)|^4
\,d\xi $$ where the wavelet bases is orthonormal, and the functional
is easy to be generalized to non-orthonormal case by Wigner procedure.
The main result is:
(1) The range of $r(psi)$ is $(0,1]$.
(2) the wavelet base is shiftable iff $r(\psi)=1$.
(3) It is nearly shiftable iff it approximates $1$.
Depending on this functional, we researched the shiftability of
B-spline wavelets and Y.Meyer's wavelet, in particular the relation
between the shiftability of scaling function space and that of wavelet
subspace. Some interesting results are proved.
[2] Translation invariance and sampling theorem of wavelet,
by Qiao WANG and Lenan WU (submitted)
The sampling theorem is used at least for two goals: one is to present
a signal, the another is to recover the waveform. The wavelet sampling
theorem systematically researched by Walter and Janssen is applicable
for the first purpose, but not for the second unless the wavelet is
shiftable (Shannon's scaling function) or weak translation invariant
(e.g. Meyer's wavelet). In this correspondence we eneralize our metric
functional, and define the degenerate shiftability. We give a family
of functionals $$r_k(\phi)={1\over
2\pi}\int^{+\infty}_{-\infty}|\widehat \phi(\xi)|^2 |\widehat
\phi(2^{-k}\xi)|^2 \,d\xi $$ We proved that: $0<r_0(\phi)\le
r_1(\phi)\le r_2(\phi)\le \cdots \le 1$ and $\lim_{k\tends +\infty
}r_k(\phi)=1$. Thus if $r_n(\phi)\approx 1$ very closely, one can
recover the waveform by the formula $$ f(x+s)\approx \sum_k f({k\over
2^n}+s) S(x-{k\over 2^n})$$ provided $f(x)\in V_0=Span\{
\phi(x-k);k\in Z\}$ is a finite sum. Using this functional, we found
that $r_1(\phi_M)=1$ for Meyer's wavelet which is known by Walter in
another way. And we found that all the B-spline scaling function of
order $n>2$ satisfy $r_1(\phi_n)\approx 1$ and we construct a group of
shiftable real orthonormal wavelet bases other than Shannon's ones.
[3] A unified approach to shiftability problem,
by Qiao WANG and Lenan WU
For the translation invariance problem concerning translation, dilation
and other unitary operators, we build the metric functional based on the
Fourier analysis on local compact abelian groups. We also give the tile
characterization of the spectrum measure.
Date: Wed, 8 Jul 1998 12:20:35 -0700
From: "Jae S. Lee" <leejs@earthlink.net>
Subject: #6 Thesis: The analysis of electromyography using wavelets
Hello, I'd like to share my Ph.D. work with you. I've got my degree
at the University of Iowa (Biomedical Engineering) and currently
working at the California State University at Long Beach, California
as a post-doc fellow in Electrical Engineering Dept. I've mainly
worked on the analysis of Surface Electromyography (EMG) signals using
both discrete and continuous wavelet transforms. If you have specific
questions, like the detailed informations about my publications,
please e-mail me, or simply visit my home page
http://home.earthlink.net/~leejs
Title: THE ANALYSIS OF ELECTROMYOGRAPHY USING WAVELETS
Abstract: Electromyography (EMG) is a well-established technique that
provides an indication of a level of muscle activity. The analysis
used in my work is mainly based on wavelet transform, a recently
developed advanced mathematical technique. Three criteria, wavelet
filtering, window width and scale comparison index, were used to
detect onset time from sudden load experiment. The wavelet based
onset detection algorithm showed precision, consistency and
reproducibility not only in clear signals but also in signals that
have considerable background activity. The onset detection error in
wavelet based method was about 5 ms while those of traditional methods
were about 20 ms. The reaction time of low back muscles to sudden
loads has been measured and compared for various lumbar supports. No
statistically significant effect in reaction time was detected as a
result of wearing various lumbar supports. The relative muscular
activity, however, did show apparent differences when different lumbar
supports were used. The properly worn Chase Lumbar support produced
the least overcompensation. The effect of lumbar supports in fatigue
has also been demonstrated using median frequency shift measure during
the Sorenson test. Wide, circumferential compression applied around
the pelvis by wearing lumbar support snugly was shown to reduce muscle
fatigue. In the whiplash study, it was demonstrated that cervical
muscle activity exists shortly after the initial motion of the sled
using two different techniques: continuous wavelet transform and
multiresolution analysis. Both techniques gave approximately the same
results for the times and frequency of occurrences of muscle activity.
The reaction time measured in this study was shorter than those
reported previously by other researchers. This finding is significant
because it suggests that such muscle activity may be a contributing
factor in the creation of a more accurate whiplash biomechanical
model. Multiresolution analysis has been successfully applied for the
automatic onset time determination from low back surface EMG signals.
The results indicated that the reaction time from the erector spinae
muscle after rehabilitation was faster than the reaction time before
rehabilitation for about half of the cases.
Jae S. Lee, Ph. D.
California State Univ. at Long Beach
1213 E. Cordova St. Apt. 6
Pasadena, California 91106
Tel: 626)844-4520
E-mail: leejs@earthlink.net
Date: Thu, 2 Jul 1998 15:56:24 +0200
From: Baltzer Science <mailer@ns.baltzer.nl>
Subject: #7 Contents: Numerical Algorithms 17: 1,2
Numerical Algorithms 17 (1998) 1,2
Jean Piquet and Xavier Vasseur
Multigrid preconditioned Krylov subspace methods for three-dimensional
numerical solutions of the incompressible Navier--Stokes equations 1-32
D.E. Roberts
A vector Chebyshev algorithm 33-50
Tony F. Chan, Lisette de Pillis and Henk van der Vorst
Transpose-free formulations of Lanczos-type methods for nonsymmetric linear
systems 51-66
C. Brezinski and M. Redivo-Zaglia
Transpose-free Lanczos-type algorithms for nonsymmetric linear systems
67-103
Faezeh Toutounian
The stable ATA-orthogonal s-step Orthomin(k) algorithm with the CADNA
library 105-119
J.C. Butcher and M.T. Diamantakis
DESIRE: diagonally extended singly implicit Runge--Kutta effective order
methods 121-145
Frauke Sprengel
Periodic interpolation and wavelets on sparse grids 147-169
Vu Hoang Linh
On some questions arising in numerical realization of amplitude-phase
methods 171-191
Date: Thu, 2 Jul 1998 16:10:49 +0200
From: Baltzer Science <mailer@ns.baltzer.nl>
Subject: #8 Contents: Advances in Computational Mathematics 8-4
Advances in Computational Mathematics 8 (1998) 4
Marie-Laurence Mazure
Vandermonde type determinants and blossoming 291-315
F. Diele, L. Lopez and R. Peluso
The Cayley transform in the numerical solution of unitary differential
systems 317-334
David R. Willé
Experiments in stepsize control for Adams linear multistep methods 335-344
Bert Jüttler
The dual basis functions for the Bernstein polynomials 345-352
Mohammed Laghchim-Lahlou
The Cr-fundamental splines of Clough--Tocher and Powell--Sabin types for
Lagrange interpolation on a three direction mesh 353-366
Yuan Xu
Minimal cubature formulae for a family of radial weight functions 367-380
Carsten Franke and Robert Schaback
Convergence order estimates of meshless collocation methods using radial
basis functions 381-399
Date: Thu, 02 Jul 1998 13:00:07 -0400
From: Paul Costa <pc@mathworks.com>
Subject: #9 Answer: Wavelets and financial time series (WD 7.6 #26)
The question was:
> Question: My interest in wavelets is as a tool in helping to forecast
> financial time series. My primary question is whether there is
> software available to automate the calculation of wavelets for a long
> time series on a walk-forward basis, one day at a time, such that the
> wavelet calculation takes advantage only of current and past data and
> no future data. I've experimented with TimeStat at bit, which can
> perform the basis functions of detrending, normalizing, "mirroring"
> the endpoint data to remove distortions, and wavelet calculations, but
> only manually - one day at a time. Can anyone steer me to pertinent
> software to perform this analysis. I've read that the Morlet wavelets
> are most appropriate for such time series, but TimeStat does not offer
> this feature. Again, any recommendations?
>
> N. Avery
> Bryn Mawr, PA
> email: Noyes_L_Avery@email.mobil.com
The answer is:
There is an article titled "Making Wavelets in Finance" written by
Mark J. Jensen, Ph.D which maybe of interest. The article is
in a newsletter called "Financial Engineering News" (August 1997
Volume 1 Number 1) and maybe obtained from their web site:
http://www.fenews.com/
Best Regards,
Paul Costa
The MathWorks, Inc.
pc@mathworks.com
Date: Thu, 2 Jul 1998 10:01:41 +0800 (WST)
From: Fabio Boschetti <fabio@ned.dem.csiro.au>
Subject: #10 Question: experience with multiscale-edges-based image processing?
I have been playing with image reconstruction from multiscale edges,
using Mallat's algorithm (see Mallat and Zhong, 1992) and it works
very well.
My next step is trying to remove/enhance local features in an image by
removing/enhancing the corresponding multiscale edges. For my
knowledge, this technique has been used to denoise and enhance medical
images. For denoising purpose, the small edges are eliminated and the
image then reconstructed. For image enhancement, all the edges are
multiplied by some factor and then the image reconstructed. At least
this is my understanding.
What I am trying to do is to remove or enhance specific features,
whilst leaving the remaning mutliscale edges untouched. My first
results are a bit contradictory and in general not too successfull. It
appears that 1) manipulating some multiscale edges and not others
impose a strong distorsion on the wavelet representation and
consequently the recovered image look quite different from the
original one even in the untouched features. 2) the multiscale edges
corresponding to some features contain information related to adjacent
features as well, and, as such, can not be considered 'local'. This is
an obvious observation but I did not expect it to affect the
reconstruction too much.
I was wondering whether any of you had experience with this problem
and can send me any comments..
Thank you very much in advance.
Fabio
Dr. Fabio Boschetti
Research Scientist
CSIRO - Division of Exploration and Mining
39 Fairway, Nedlands 6009
PO Box 437, Nedlands 6009
Western Australia
email - fabio@ned.dem.csiro.au
Web page - http://www.ned.dem.csiro.au/BoschettiFabio
Tel (w) ++61 8 9284 8421
(h) ++61 8 9284 3282
Fax (w) ++61 8 9389 1906
Date: Thu, 2 Jul 1998 00:58:49 -0500
From: "Pete" <vrlpete@surf-ici.com>
Subject: #11 Question: Wavelet theory and field analysis.
a quick question if i may. i am interested in applying some wavelet
theory toward sound field analysis. i am not looking to set the world
on fire or redefine any universal constants, rather just attempt to
apply some number crunching to otherwise subjective parameters. since
my math is a tad rusty being some 8 years since my EE study days, can
you point me towards some tools/techniques that might prove usefull if
not for implementation, for usefull background and 'boning up'?
best regards,
pete maudlin
Date: Thu, 16 Jul 1998 17:35:12 +0800 (WST)
From: Fabio Boschetti <fabio@ned.dem.csiro.au>
Subject: #12 Question: Definition of edges
We have developed a multiscale edge detection algorithm that is
applied to the analysis of geophysical images. Details can be found at
the web page at the bottom of the message.
The edge detector works very well, enabling the user to discrimate
features even in areas of the image that look completely flat. However
recently we encoutered some problems, where reasonably visible
features did not generated any edge. Always these feature appear close
to major features in the image.
The reason for this to happen is that the definition of 'edge', as I
found it often in the literature, is 'a local maxima of the horizontal
derivative in the direction of the local gradient'. Close to main
features in an image, some features may lie parallel to the local
gradient and as such they do not generate edges even if quite clearly
visible. This, of course, can be overcome by searching for edges
independently from the direction of the local gradient, but in this
case many spurious edges appear.
I was wondering if any of you have encountered this problem on real
world images. Also, I would like to have pointers to references where
the definition of edges is discussed, especially with regard to
multiscale analysis and image reconstruction from the edges
themselves.
Thank you very much in advance.
Fabio
Dr. Fabio Boschetti
Research Scientist
CSIRO - Division of Exploration and Mining
39 Fairway, Nedlands 6009
PO Box 437, Nedlands 6009
Western Australia
email - fabio@ned.dem.csiro.au
Web page - http://www.ned.dem.csiro.au/BoschettiFabio
Tel (w) ++61 8 9284 8421
(h) ++61 8 9284 3282
Fax (w) ++61 8 9389 1906
Date: Thu, 16 Jul 1998 14:47:36 +0100
From: "Gary S. Collins" <g.s.collins@exeter.ac.uk>
Subject: #13 Question: Wavelet Packets and Density Estimation
Hello,
My question is regards Wavelet Packets, and was wondering (if
possible) if anyone can tell me if there exists a method/algorithm to
calculate the wavelet packet function, w_{j,b,k} at an arbitrary
point, maybe in a similar fashion to that of the Daubechies-Lagarias
algorithm which calculates the scaling function by the infinite
product of suitably chosen matrices calculated from the filter coefs.
Another (related) question is whether there exists some relationship
between the wavelet packet function and scaling/or mother wavelet in
ordinary wavelet analysis. The questions are with regards to Density
Estimation where empirical coefs are needed to be calculated from the
data.
Many thanks in advance
Gary Collins.
Gary S. Collins
Dept of Mathematical Statistics and Operational Research
Room 222
Laver Building
University of Exeter
Exeter
EX4 4QE.
Tel: (01392) - 264465
FAX: 0044 01392 264460
Email : G.S.Collins@Exeter.ac.uk
http://msor.ex.ac.uk/PostGrads/gary/HomePage.html
Date: Fri, 17 Jul 1998 09:51:20 +1000
From: Gary Mc Darby <gmcdarby@ee.unsw.edu.au>
Subject: #14 Question: Normalisation of the sinc wavelet filter bank
To the wavelet mathematicians,
I am presently trying to develop a theoretical backround on how
wavelets are designed. I am still in the early stages of this work and
am looking at how the Haar and sinc wavelets and their associated
filter banks are related.
Could somebody explain to me (or point me to some relevant literature)
the reason why, in the case of the sinc wavelet, one needs to
introduce a phase shift into the high pass filter in order for the
basis to be complete? Secondly, if the gain of the analysis filters =
sqrt(2), why is the gain of the synthesis filters not the inverse of
this in order for perfect reconstruction? It seems to me that if the
gains in both the analysis filters and synthesis filters are sqrt(2)
then the overall gain of the filter bank is 2. Is this correct?
Thanking you in advance
Best regards
Gary Mc Darby B.E. MEngSc CEng
Department of Electrical Engineering
University of New South Wales
Sydney 2052
Australia
Email: gmcdarby@ee.unsw.edu.au
Phone: 0061-2-93855315
<webmaster@wavelet.org>