Preprocessing Techniques for the Estimation of Parameters from Noisy Transient Data.
Abstract
Two kinds of pole estimators for given noisy transient data are presented. One is a SVD-based methods and the other is the iterative preprocessing algorithm (IPA). The SVD method estimates the characteristic equation coefficients from the weakest eigenvectors when the system order is overdetermined. The IPA is shown to converge to the maximum-likelihood estimator for some range of SNR. It not only reduces the computational burden of the currently existing iterative algorithms but also improves stability. The approximate IPA (AIPA), which further reduces the computational burden, is described. The Cramer-Rao lower bound for the system characteristic equation coefficients is calculated. The IPA is extended to pole estimation for given multiple data sets. This technique is applied to estimating the natural frequencies of an electromagnetic scatter. An iterative scheme is described for estimating other parameters of the singularity expansion method (SEM) formalism. Keywords include: SEM; Noise reduction; Singular value decomposition; Maximum likelihood estimation technique; Multiple data set; and Transient data.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1986
- Accession Number
- ADA165764
Entities
People
- Sung-won Park
Organizations
- University of New Mexico