🚀 go-pugleaf

In article <32n6qk$mj7@darum.uni-mannheim.de>,
Waldmann <thinklab@rummelplatz.uni-mannheim.de> wrote:
:Student trying to confer a doctorate on viszera-evoked brain potentials
:needs some methodological advice. These kinds of signal aren't very well
:behaved (bad signa-to-noise-ratio, ill-defined shape, varying onset etc.),
:so ordinary digital filters like moving averages or MMSEs won't do the
:job when extracting the signal from noisy background EEG. Some say that
:Kalman filtering some some state-space-represented filter model might
:help. I understand the basics and the frames of application, but driving a 
:car is quite different from actually constructing and manufacturing it,
:isn't it ? Consequnetly, I'd be very thankful for some hints as regards:
:
:      * introductory books / papers on Kalman filtering and its
:        applicability in the field of signal extracting, perhaps
:        with an emphasis to EEG analysis
:
:      * who has ever worked in the field before what filters did they use
:
:      * papers discussing the links between state-space-models or general
:        systems' representation and biosignal analysis
:
:Maybe some researchers' group working on the same field wants to establish
:a regular exchange of ideas over the net. Please send e-mail to
:
:				thinklab@rmlplz.uni-mannheim.de
:or contact
:				Hans-Christian Waldmann
:				Falkenstr. 21 A
:       				68307 Mannheim 
:				Germany

Kalman filtering, especially in the context you seek, should be
considered as some quasi-linear modeling of your data, with all the
benefits and dangers implied therein.  Some nice papers which
illustrate how Kalman filtering is related to other linearized systems,
e.g., ARMA models, are
	A K. Kishida
	%T Physical Langevin model and the time-series model in systems
	   far from equilibrium
	%J Phys. Rev. A
	%V 25
	%D 1982
	%P 496-507
	 
	%A K. Kishida
	%T Equivalent random force and time-series model in systems
	   far from equilibrium
	%J J. Math. Phys.
	%V 25
	%D 1984
	%P 1308-1313

If you believe, and there is good reason to do so, that EEG contains
more nonlinear structure in its basic deterministic "drift" as well as in
its "noise," then you likely will do better modeling this data with a
model better suited to describing your particular system, thereby
increasing your chances of getting a better "signal to noise" ratio.

Unfortunately, as soon as you leave the "canned black-box" quasi-linear
world of most software, you enter a world requiring more work on more
details of many stages of your work.

In the archive ftp.alumni.caltech.edu, in directory /pub/ingber, I have
a series of papers, smni..., which give my own work leading up to
modeling EEG.  The most recent papers, smni91_eeg.ps.Z and
smni95_eeg.ps.Z likely are the most readable.  These can be retrieved
via anonymous ftp.

There are quite a few other approaches to modeling EEG in more
realistic ways than by Kalman filtering.  One text I recommend is
	%A P.L. Nunez
	%T Electric Fields of the Brain: The Neurophysics of EEG
	%I Oxford University Press
	%C London
	%D 1981
and a newer version should be out this year:
	%E P. Nunez
	%B Neocortical Dynamics and Human EEG Rhythms
	%I Oxford University Press
	%C New York, NY
	%P (to be published)
	%D 1994

Lester
-- 
\\ Prof. Lester Ingber                                                //
\\ Lester Ingber Research                                             //
\\ P.O. Box 857                     E-Mail: ingber@alumni.caltech.edu //
\\ McLean, VA 22101       Archive: ftp.alumni.caltech.edu:/pub/ingber //