Principle of guaranteed result in a problem of stationary linear smoothing with uncertainty of spectral densities

The problem of linear smoothing is discussed for a stationary random process. An observed process is an additive blend of the uncorrelating useful component and noise whose spectral density moments are constrained.