The accuracy of estimating the parameters of stationary processes

Addresses the general problem of establishing bounds on the achievable estimation accuracy of the parameters of a regular stationary process, from a single observed realization of this process. The special case of establishing bounds on the estimation accuracy of the parameters of harmonic components in the presence of noise with a known spectrum or covariance function has been studied quite extensively, especially for the case in which the noise is white. However, relatively little work seems to have been done for the case of harmonic components in unknown noise. The authors concentrate on a solution to the problem of the achievable accuracy in jointly estimating the parameters of the harmonic (deterministic) and purely indeterministic components of the process, based on a finite length, single observed realization of this process