Robust Linear Filtering for Multivariable Stationary Time Series.
Abstract
The problem of asymptotic, non-causal linear filtering for statistically contaminated multivariable stationary time series is considered. The spectra of both the signal and the noise components of the observation process are assumed to belong to certain convex and compact classes. The minimax criterion of optimality is adopted, and for some specific spectral classes the corresponding solutions are found. The performance of those solutions is studied, where the performance criteria used are efficiency, error variation within the classes and breakdown curves or points. Some examples are studied quantitatively. Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1983
- Accession Number
- ADA131209
Entities
People
- Haralampos Tsaknakis
- P. Papantoni-kazakos
Organizations
- University of Connecticut