Optimal multilinear estimation of a random vector under constraints of causality and limited memory

A new technique is provided for random vector estimation from noisy data under the constraints that the estimator is causal and dependent on at most a finite number p of observations. Nonlinear estimators defined by multilinear operators of degree r are employed, the choice of r allowing a trade-off between the accuracy of the optimal filter and the complexity of the calculations. The techniques utilise an exact correspondence of the nonlinear problem to a corresponding linear one. This is then solved by a new procedure, the least squares singular pivot algorithm, whereby the linear problem can be repeated reduced to smaller structurally similar problems. Invertibility of the relevant covariance matrices is not assumed. Numerical experiments with real data are used to illustrate the efficacy of the new algorithm.

[1]  Harold W. Sorenson,et al.  Parameter estimation: Principles and problems , 1980 .

[2]  Colin J. Champion.  Empirical Bayesian estimation of normal variances and covariances , 2003 .

[3]  Tatsuya Kubokawa,et al.  Estimating the covariance matrix: a new approach , 2003 .

[4]  V. J. Mathews,et al.  Polynomial Signal Processing , 2000 .

[5]  Colin O. Wu.  LOCAL POLYNOMIAL REGRESSION WITH SELECTION BIASED DATA , 2000 .

[6]  Jan P. Allebach,et al.  Combating Speckle In SAR Images: Vector Filtering And Sequential Classification Based On A Multiplicative Noise Model , 1990 .

[7]  V. Gimeno,et al.  Obtaining the EEG envelope in real time: a practical method based on homomorphic filtering. , 1987, Neuropsychobiology.

[8]  Gene H. Golub,et al.  Matrix computations , 1983 .

[9]  Anatoli Torokhti,et al.  An optimal linear filter for random signals with realisations in a separable Hilbert space , 2003, The ANZIAM Journal.

[10]  Adi Ben-Israel,et al.  Generalized inverses: theory and applications , 1974 .

[11]  Thomas L. Marzetta,et al.  Estimating a covariance matrix from incomplete realizations of a random vector , 1992, IEEE Trans. Signal Process..

[12]  Keinosuke Fukunaga,et al.  Introduction to statistical pattern recognition (2nd ed.) , 1990 .

[13]  Lihong Wang,et al.  Asymptotics of estimates in constrained nonlinear regression with long-range dependent innovations , 2004 .

[14]  Andrew W. Moore,et al.  Efficient Locally Weighted Polynomial Regression Predictions , 1997, ICML.

[15]  V. Yohai,et al.  A Fast Procedure for Outlier Diagnostics in Large Regression Problems , 1999 .

[16]  Phil Howlett,et al.  Best approximation of operators in the modeling of nonlinear systems , 2002 .

[17]  Anatoli Torokhti,et al.  An optimal filter of the second order , 2001, IEEE Trans. Signal Process..

[18]  Anatoli Torokhti,et al.  The modelling and numerical simulation of causal non-linear systems , 2001 .

[19]  Olivier Ledoit,et al.  A well-conditioned estimator for large-dimensional covariance matrices , 2004 .

[20]  Wanquan Liu,et al.  Generalized Karhunen-Loeve transform , 1998, IEEE Signal Process. Lett..

[21]  R. Carroll,et al.  A Note on the Efficiency of Sandwich Covariance Matrix Estimation , 2001 .

[22]  Alan S. Willsky,et al.  A Krylov Subspace Method for Covariance Approximation and Simulation of Random Processes and Fields , 2003, Multidimens. Syst. Signal Process..

[23]  Keinosuke Fukunaga,et al.  Introduction to Statistical Pattern Recognition , 1972 .

[24]  Anatoli Torokhti,et al.  Constructing fixed rank optimal estimators with method of best recurrent approximations , 2003 .

[25]  Anatoli Torokhti,et al.  Optimal Mathematical Models for Nonlinear Dynamical Systems , 2003 .

[26]  Vladimir N. Fomin,et al.  Abstract Optimal Linear Filtering , 2000, SIAM J. Control. Optim..

[27]  Yukihiko Yamashita,et al.  Relative Karhunen-Loève transform , 1996, IEEE Trans. Signal Process..