Comments on "Linearization method for finding Cramer-Rao bounds in signal processing" [with reply]

The authors comment that an interesting attempt was made to simplify the derivation of the Cramer-Rao bound (CRB) for the principal parameters in the so-called superimposed-signals-in-noise models. Here, we streamline the derivation in question and then go on to show how it relates to other possible derivations of the CRB. We show that the new derivation can be neatly interpreted as performing a block diagonalization of the CRB matrix, which is a sensible thing to do in the presence of nuisance parameters. Gu (see ibid., vol.48, p.543-545, Feb. 2000) replies that the interesting problem of de-coupling in Cramer-Rao bounds is algebraically and neatly approached in this article, whereas the linearization method is geometrical, with statistical interpretations.