THEORETICAL CONSIDERATIONS IN ADAPTIVE PROCESSING.
Abstract
Theoretical results on four logically independent problems arising in the application of adaptive multichannel processing are presented. Areas treated are multiconstraint adaptive maximum-likelihood methods, interaction of oversampling with rate of adaption, effect of local noise on multichannel filter design, and adaptive computation of high-resolution wavenumber spectra. Some theoretical questions on uniqueness of the adaptive maximum-likelihood method are considered, and the method is extended to multiple constraints. For oversampled data, MSE in predicting one channel from a set of independent channels is shown to monotonically decrease with k sub s to the point of instability; the severity of this effect depends on the degree of oversampling and the number of channels. Modeled local noise provides the covariance matrix necessary to determine the optimum Wiener filters. Theoretical MSE curves are given for the application of these curves to noise, signal plus noise, and signal plus noise plus local noise. The effect of this local noise model, when present in the design data and absent in the data to which the filters are applied, appears negligible. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 28, 1968
- Accession Number
- AD0832970
Entities
People
- Aaron H. Booker
- Chung-yen Ong
- Thomas E. Barnard
- Thomas Krile
Organizations
- Texas Instruments