Abstract (EN):
We are interested in a piecewise linear discrete time filtering problem with small observation noise. We present and compare different tests which enable us to compute the intervals of linearity of the observation function, specially when it is symmetric. Over such an interval, we can then approximate the optimal filter by the correspon-ding Kalman-Bucy filter. As in [FJSZ], we approximate discrete time processes by diffusions, in order to estimate the probabilities of error and the expected times for taking a decision.
Idioma:
Português
Tipo (Avaliação Docente):
Científica
Referência:
Rapport de Recherche 1451