DocumentCode :
919578
Title :
Testing for harmonizability
Author :
Hurd, H.L.
Volume :
19
Issue :
3
fYear :
1973
fDate :
5/1/1973 12:00:00 AM
Firstpage :
316
Lastpage :
320
Abstract :
Let R(s,t) be a covariance function having the representation begin{equation} R(s,t) = int_{-infty}^{infty} int_{-infty}^{infty} exp (isx - ity)d^2 G(x,y) end{equation} where G(x,y) is continuous to the right in both variables and is of bounded variation in the plane; then R(s,t) is harmonizable in that G(x,y) is also a covariance. We give examples in which this result is used to determine the harmonizability of new processes and covariances that are formed by operations on old processes and covariances. Specifically, if X(t) is a real Gaussian harmonizable process, then X^n (t) is harmonizable. If X(t) is harmonizable, d^2 G(x,y) has bounded support and g(t) is a Fourier-Stieltjes transform, then X(g(t)) and X(t + g(t)) are harmonizable. If begin{equation} X(t) =int_{-infty}^{infty} f(t,u) dZ(u) end{equation} where f (t,u) = f (t - u) is a Fourier-Stieltjes transform and G(u,v) = E(Z(u)Z^{\\ast } (v)) has finite total variation, then X(t) is harmonizable. We also obtain a sufficient condition for Priestley\´s oscillatory processes to be harmonizable. We find that the Bochner-Eberlein characterization of Fourier-Stieltjes transforms, while not the only method, is particularly convenient for determining the harmonizability of these examples.
Keywords :
Covariance functions; Harmonic analysis; Spectral analysis; Stochastic processes; Fourier transforms; Frequency measurement; Frequency response; Maximum likelihood detection; Stochastic processes; Sufficient conditions; Testing;
fLanguage :
English
Journal_Title :
Information Theory, IEEE Transactions on
Publisher :
ieee
ISSN :
0018-9448
Type :
jour
DOI :
10.1109/TIT.1973.1055017
Filename :
1055017
Link To Document :
بازگشت