Maximum likelihood estimation of quantized Gaussian autoregressive processes with Particle filters with resampling

András Horváth; Miklós Rásonyi

Summary

Proceedings of the 2012 International Symposium on Nonlinear Theory and its Applications

2012

Session Number:B3L-B

Session:

Number:427

Maximum likelihood estimation of quantized Gaussian autoregressive processes with Particle filters with resampling

András Horváth, Miklós Rásonyi,

pp.427-430

Publication Date:

Online ISSN:2188-5079

DOI:10.15248/proc.1.427

PDF download (325.3KB)

Summary:

In this paper we propose a method for the calculation of the maximum likelihood estimator for the autoregression coefficient of a stable quantized Gaussian autoregressive, AR(1) process. Our method uses particle filters with resampling and suits ideally on manycore architectures and can be implemented in a parallel way, this way yields fast processing speed. The extension to multidimensional autoregressivemoving-average (ARMA) systems is straightforward.

References:

[1] A. Horvath, M. Rasonyi Topographic Implementation of Particle Filters on Cellular Processor Arrays. ELSEVIER Signal Processing Submitted

[2] D. Crisan, A. Doucet. A Survey of Convergence Results on Particle Filtering Methods for Practitioners. IEEE Transactions on Signal Processing 736-746, 2002

[3] D. Magill. Optimal adaptive estimation of sampled stochastic processes. IEEE Transactions on Automatic Control 434-439, 1965

[4] H. Boche, U. J. Monich. Behavior of the Quantization Operator for Bandlimited, Nonoversampled Signals. IEEE Transactions on Information Theory 2433-2440, 2010

[5] L. Finesso, L. Gerencsér, I. Kmecs. A randomized EM-algorithm for estimating quantized linear Gaussian regression. Proceedings of the 38th IEEE Conference on Decision and Control, Phoenix, Arizona, USA. 5100-5101, 1999.

[6] L. Finesso, L. Gerencsér, I. Kmecs. A recursive randomized EM-algorithm for estimation under quantization error. Proceedings of the American Control Conference, Chicago, Illinois., 790-791, 2000.

[7] L. Gerencsér, I. Kmecs, B. Torma. Quantization with adaptation - estimation of Gaussian linear models. Communications in Information and Systems, 8, 223-244, 2008.