E4.4 HIGH PERFORMANCE SUD-LIKE PROCEDURE FOR SPECTRAL ESTIMATION USING

This work describes how Rayleigh estimates can be viewed as a method which performs as a SVD procedure without d oing it. As a short cut to get principal component reduction, Rayleigh quotients allows the resolution of frequency detectors yet preserving the asymptotic behavior to the actual power spectral density. In a filtering framework the estimate is extended to adaptive schemes a nd 2-D spectral estimation. The resulting estimate provides the way out for adaptive p rocessing with low computational complexity w hich, in general, are the two drawbacks associated to the principal component analysis. It avoids also the c rucial decision between signal subspace and noise subspace which promotes undesired distortion and false peaks in spectral estimation applicat ions.