Title :
Particle Filter as A Controlled Markov Chain For On-Line Parameter Estimation in General State Space Models
Author :
Poyiadjis, George ; Singh, Sumeetpal S. ; Doucet, Arnaud
Author_Institution :
Dept. of Eng., Cambridge Univ.
Abstract :
In this paper we present a novel optimization method for on-line maximum likelihood estimation (MLE) of the static parameters of a general state space model. Our approach is based on viewing the particle filter as a controlled Markov chain, where the control is the unknown static parameters to be identified. The algorithm relies on the computation of the gradient of the particle filter using a score function approach
Keywords :
Markov processes; maximum likelihood estimation; particle filtering (numerical methods); state-space methods; controlled Markov chain; general state space models; on-line maximum likelihood estimation; on-line parameter estimation; optimization method; particle filter; score function approach; Filtering; Hidden Markov models; Maximum likelihood estimation; Optimization methods; Parameter estimation; Particle filters; Sliding mode control; State estimation; State-space methods; Statistics;
Conference_Titel :
Acoustics, Speech and Signal Processing, 2006. ICASSP 2006 Proceedings. 2006 IEEE International Conference on
Conference_Location :
Toulouse
Print_ISBN :
1-4244-0469-X
DOI :
10.1109/ICASSP.2006.1660657
