FILTERS FOR ESTIMATING MARKOV MODULATED POISSON PROCESSES AND IMAGE-BASED TRACKING

Authors
KRISHNAMURTHY V ELLIOTT RJ
Citation
V. Krishnamurthy et Rj. Elliott, FILTERS FOR ESTIMATING MARKOV MODULATED POISSON PROCESSES AND IMAGE-BASED TRACKING, Automatica, 33(5), 1997, pp. 821-833
Citations number
19
Categorie Soggetti
Controlo Theory & Cybernetics","Robotics & Automatic Control
Journal title
Automatica → ACNP
ISSN journal
00051098
Volume
33
Issue
5
Year of publication
1997
Pages
821 - 833
Database
ISI
SICI code
0005-1098(1997)33:5<821:FFEMMP>2.0.ZU;2-I
Abstract
We present algorithms for state and parameter estimation of Markov mod ulated Poisson processes (MMPP). We first derive finite-dimensional in novations and Zakai filters for various statistics of a MMPP. Using th ese filters, a tilter-based expectation-maximization algorithm is deri ved for computing maximum-likelihood parameter estimates. Finally, we present an application of the techniques in image-based tracking by es timating the state of a Markovian jump linear system. (C) 1997 Elsevie r Science Ltd.