Effects of Oversampling on Time-Adaptive Filters
Abstract
Independent time-series data, consisting of white noise with different degrees of oversampling, have been used to study the effect of oversampling on the time-adaptive prediction filter. If the data are oversampled, false gain occurs in the adaptive prediction results. The false gain depends on the degree of oversampling, the number of channels used in making the prediction, the filter length, and the convergence parameter. Two adaptive algorithms -- one having a constant convergence parameter and the other having a variable convergence parameter -- are discussed in this report. Particular cases of the prediction mean-square-error function of the time- adaptive filter are derived and compared to the empirical results. Although the false gain can be quite severe for high rates of adaption, rates of adaption can be selected in terms of theoretical maximum rates of adaption so that the false gain is not significant -- even for oversampled data cut at 1/20 of the folding frequency.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 29, 1968
- Accession Number
- AD0849741
Entities
People
- Aaron H. Booker
- Chung-yen Ong
Organizations
- Texas Instruments