Some remarks on filtering and prediction of stationary processes (Q814145)

The aim of the paper is to present some remarks on filtering and prediction theory of stationary processes. The first remark is a negative answer to a purely probabilistic question about some equalities of random variables in some given condition about the corresponding distributions. The second remark concerns the sum between a stationary real-valued stochastic process which represents the signal and a second process which represents some kind of additive noise. Introducing the notion of disjoint stationary processes, and assumption of integrability of these processes, one can almost surely filter out the noise. Here, the possibility of carrying this out in a unilateral way arises. The third remark concerns some questions about an unknown stationary process after observing the first \(n\) outputs. Expressing the ergodicity of a process as a series of conditions on the empirical distributions, some nice results on classifying some processes are obtained.