Maximum-Likelihood Estimation of Delta-Domain Model Parameters From Noisy Output Signals

Authors:
V. Kadirkamanathan;S.R. Anderson
Affiliations:
Dept. of Autom. Control & Syst. Eng., Univ. of Sheffield, Sheffield;-
Venue:
IEEE Transactions on Signal Processing - Part I
Year:
2008

Citing 0
Cited 2

Predictive control of fast-sampled systems using the delta-operator

International Journal of Systems Science
Brief paper: Structure detection and parameter estimation for NARX models in a unified EM framework

Automatica (Journal of IFAC)

Quantified Score

Hi-index	0.00

Visualization

Abstract

Fast sampling is desirable to describe signal transmission through wide-bandwidth systems. The delta-operator provides an ideal discrete-time modeling description for such fast-sampled systems. However, the estimation of delta-domain model parameters is usually biased by directly applying the delta-transformations to a sampled signal corrupted by additive measurement noise. This problem is solved here by expectation-maximization, where the delta-transformations of the true signal are estimated and then used to obtain the model parameters. The method is demonstrated on a numerical example to improve on the accuracy of using a shift operator approach when the sample rate is fast.