Paper available: An On-line Method for Segmentation and Identification of Non-stationary Time Series

[Jens Kohlmorgen]

Readers of the connectionists list might be interested in the
following paper:


Kohlmorgen, J., Lemm, S. (2001),

      "An On-line Method for Segmentation and Identification 
                of Non-stationary Time Series"

in: Neural Networks for Signal Processing XI, IEEE, NJ, pp. 113-122. 



It is available from

   http://www.first.gmd.de/~jek/Kohlmorgen.Jens/publications.html


Abstract:

We present a method for the analysis of non-stationary time series
from dynamical systems that switch between multiple operating modes.
In contrast to other approaches, our method processes the data
incrementally and without any training of internal parameters.  It
straightaway performs an unsupervised segmentation and classification
of the data on-the-fly.  In many cases it even allows to process
incoming data in real-time.  The main idea of the approach is to track
and segment changes of the probability density of the data in a
sliding window on the incoming data stream.  An application to a
switching dynamical system demonstrates the potential usefulness of
the algorithm in a broad range of applications.


==========================================================================
   Dr. Jens Kohlmorgen                     Tel.(office) : +49 30 6392-1875
                                           Tel.(secret.): +49 30 6392-1800
   Intelligent Data Analysis Group                  Fax : +49 30 6392-1805
   Fraunhofer-FIRST (former GMD FIRST)                                    
   Kekulestr. 7                            e-mail: jek at first.fraunhofer.de
   12489 Berlin, Germany               http://www.first.fraunhofer.de/~jek
==========================================================================