@InProceedings{Supelec762,
author = {Matthieu Geist and Olivier Pietquin},
title = {{Kalman filtering \& colored noises: the (autoregressive) moving-average case}},
year = {2011},
booktitle = {{Proceedings of the IEEE Workshop on Machine Learning Algorithms, Systems and Applications (MLASA 2011)}},
pages = {4 pages},
month = {December},
address = {Honolulu (USA)},
url = {http://www.metz.supelec.fr//metz/personnel/geist_mat/pdfs/supelec762.pdf},
abstract = {The Kalman filter is a well-known and
efficient recursive algorithm that estimates the state
of a dynamic system from a series of indirect and
noisy observations of this state. Its applications range
from signal processing to machine learning, through
speech processing or computer vision. The underlying
model usually assumes white noises. Extensions to
colored autoregressive (AR) noise model are classical.
However, perhaps because of a lack of applications,
moving-average (MA) or autoregressive moving-average
(ARMA) noises seem not to have been considered before.
Motivated by an application in reinforcement learning,
the contribution of this paper is a generic approach to
take MA and ARMA noises into account in the Kalman
filtering paradigm.}
}