A non-probabilistic framework for signal spectra
Lennart Ljung · 1985
Spectra for sequences of deterministic vectors are defined without reference to any probabilistic framework. It is shown that these spectra transform under linear filtering operations according to the same formulas as spectra for stationary stochastic processes. The results are applied to convergence and consistency analysis of parametric identification methods. It is also shown how a probabilistic framework can be imposed "afterwards" as a separate option.