The basical idea is that one can use observations from different Different statistical tools are to be used according to the
sources, or coming from a variety of disciplines, able to define viewpoint of quality analysis: so, we think possible the use of
the position of an object. methods based upon sequence of observations, such as Hidden
Markov Models (HMMs).
ettlements,
ported by a BR TER FADE TBE 3 A - N N
: i ht
pt
LES fe mw à CE of " ^
locumentary
3 conditions
31]. Lets
3/B) = 0,35,
1ess of areas
orphological
, woodland,
refore equal : : : : : :
Fig. 88, 89 — Persistance of both toponyms morra in the IGM 1:25000 map (Morra di S. Martino — comune di Revello;
cen largely Castello della Morra — comune di Castellar)
that Bayes’
jguities; of
rol and help Hidden Markov Model is a powerful tool for sequence analysis. One can calculate:
find proper A Hidden Markov Model (HMM) is specified by a set of = the likelihood of a sequence of observations with
parameters: respect to a HMM by means of the so called “forward
‘ovement of = the set of states S = (sy, Sy, ...,SN) algorithm”
x siaie sequence Q — ( Gi, 4, ---» Gi) = the most probable sequence of corresponding states
of spatial, " the prior probabilities are the probability distributions (most probable path) by Viterbi algorithm
e, it is often of q; being the first state of a state sequence =" the HMM parameter estimation by forward-backward
vhose traces " the transition probabilities are the probability to go algorithm
an object, a from a state i to a state j, i.e P (qj| q;)
ed active or = the emission probabilities p(x|q;) are the probability of For instance, let's think that around a castle a settlement has
observing x when the system is in teh state q; . formed, which in other times has been abandoned: if one knows,
notice that the states (the Markov chain) are hidden. or can guess, or suggest probabilities of passing from a state to
4]