© The Institution of Electrical Engineers
Estimated spectral densities may often take on negative values in some frequency bands, and hence need be corrected to become positive for all frequencies. The Letter describes a most natural approach to enforcing the positiveness condition on an estimated spectral density, which is shown to lead to a semi-infinite optimisation problem. In the numerical example reported, the latter problem is solved by using the Matlab Optimization Toolbox.
References
-
-
1)
-
R.L. Moses ,
D. Liu
.
Optimal nonnegative definite approximation of estimated moving average covariance sequences.
IEEE Trans. Sig. Process.
,
9 ,
2007 -
2015
-
2)
-
P. Stoica ,
R.L. Moses
.
On the unit circle problem: The Schur-Cohn procedure revisited.
Signal. Process.
,
95 -
118
-
3)
-
J.A. Cadzow ,
Y. Sun
.
Sequences with positive semidefinite Fourier transforms.
IEEE Trans.
,
6 ,
1502 -
1510
-
4)
-
Feyh, O., Mullis, C.T.: `Moving average separation', Proc. Int. Conf. Acoust., Speech, Signal Processing, 1988, p. 2280–2283.
-
5)
-
Jackson, L.B.: `Approximate factorization of unfactorable spectral models', Proc. Int. Conf. Acoust., Speech, Signal Processing, 1985, p. 324–326.
http://iet.metastore.ingenta.com/content/journals/10.1049/el_19931340
Related content
content/journals/10.1049/el_19931340
pub_keyword,iet_inspecKeyword,pub_concept
6
6