In  this paper  we present  a database  correlation method  combined with  Bayes
filtering as a  pattern matching technique  for mobile terminal  localization in
GSM suburban environments. A 3D deterministic radio wave propagation  prediction
model  is used  for the  database construction.  Bayes filter  provide a  robust
technique to  deal with  multi-hypotheses situations  usually occur  in cellular
localization when the location dependent parameter considered is the radio power
level. Furthermore,  three methods  used to  yield a  location estimate from the
resulted hypotheses are presented  and compared according to  their performance.
Results  show very  acceptable location  accuracies relative  to the  considered
suburban  area,  which is  a  typical environment  in  many populated  parts  of
Germany.