Autoregressive Spectral Estimation, Log Spectral Smoothing, and Entropy.
Abstract
Two important methods of spectral estimation, autoregressive spectral estimation, and log spectral kernel estimation are derived from a minimum information divergence estimation principle. The fact that autoregressive spectral estimators are maximum entropy estimators is shown to be proved without the use of the calculus of variations using the properties of minimum information divergence estimation. Adaptive procedures for forming these estimators (and combining to form iterated estimators) are provided by order-determining and truncation point determining criteria, which are described. An estimated spectrum is given for Wolfer's sunspot data. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1981
- Accession Number
- ADA104940
Entities
People
- Emanuel Parzen
Organizations
- Texas A&M University