Thesis: Receiver Structures based on SOMs

[Kimmo Raivio]

The following Dr.Tech. thesis is available at

http://www.cis.hut.fi/~kimmo/papers/thesis.ps.gz (compressed postscript, 217K )
http://www.cis.hut.fi/~kimmo/papers/thesis.ps (postscript, 797K )

Most of the articles that belong to the thesis can be accessed through
the page http://www.cis.hut.fi/~kimmo/papers/

----------------------------------------------------------------------

Receiver Structures Based on Self-Organizing Maps

Kimmo Raivio

Helsinki University of Technology       
Lab. of Computer and Information Science
P.O.BOX 5400, FIN-02015 HUT, FINLAND    
Email: Kimmo.Raivio at hut.fi

Abstract

New adaptive receiver structures are studied to allow a more efficient
compensation of the disturbances of the communication channel.  This
study concentrates on the use of the Self-Organizing Map (SOM)
algorithm as a building block of new adaptive receivers.  The SOM has
been used both as an adaptive decision device and to follow up error
signals.
  
When the SOM was used as an adaptive decision device, comparisons with
conventional equalizers such as the linear equalizer and the decision
feedback equalizer were performed.  The new structures were also
compared with other neural methods like radial basis function networks
and multi-layer perceptrons.  The performances of the neural
equalizers and especially the SOM have been found to be better in
nonlinear multipath channels and about equal in linear channels.

When the SOM was used to follow up error signals, the actual idea was
to cancel interference. This task was divided between following up the
error distribution and finding out the error estimate. The error was
approximately the same as the interference. Other sources of error
were noise, intersymbol interference, wrong error estimates and
detection errors due to the reasons mentioned before. The error
distribution can be followed up, but the problem is how to predict the
error.  Some solutions are presented in this thesis, but they do not
provide satisfactory results.  The performance has been compared with
a pure detector without any kind of interference cancellation and with
a receiver based on the radial basis function network.  However, it
was discovered that these neural receivers designed for
interference cancellation perform better when nonlinear distortions
are compensated.

The receivers based on the SOM are slightly more complicated than
conventional ones, but when a channel has nonlinear disturbances in
particular, they offer one possible solution.


-- 
* Kimmo Raivio, Dr. of Science in Technology | email: Kimmo.Raivio at hut.fi 
* Lab. of Computer and Information Science   | http://www.cis.hut.fi/~kimmo/
* Helsinki University of Technology          | phone +358 9 4515295 
* P.O.BOX 5400, FIN-02015 HUT, FINLAND       | fax   +358 9 4513277