Title :
Approximate factorization of unfactorable spectral models
Author :
Jackson, Leland B.
Author_Institution :
University of Rhode Island, Kingston, RI
Abstract :
Promising MA, AR, or ARMA modeling techniques are sometimes discarded because they can produce non-positive or unfactorable spectral estimates. However, if the spectral information is otherwise reasonably preserved, approximate factorization can be obtained by Durbin´s method to give positive spectra. Several useful applications to AR and ARMA modeling are cited.
Keywords :
Autocorrelation; Mean square error methods; Measurement standards; Polynomials; Predictive models; Sampling methods;
Conference_Titel :
Acoustics, Speech, and Signal Processing, IEEE International Conference on ICASSP '85.
DOI :
10.1109/ICASSP.1985.1168405