Analysis and filtration of special discrete-time Markov processes. I: Martingale representation (Q2577387)

The paper deals with a class of discrete-time Markov type processes with finite or denumerable state space. This class arises when studying hidden Markov models. The goal of the authors is to derive properties of the transition probabilities, find martingale representations of such processes in forward and backward time and also study the generated stochastic measures. There are several results which are given with their proofs. The authors promise, in another paper, to apply these results to the filtering problem [see Autom. Remote Control 66, No. 7, 1125--1136 (2005); translation from Avtom. Telemekh. 2005, No. 7, 112--125 (2005; Zbl 1130.60310)].