Akadémia Kiadó

      The purpose of this book is to investigate modeling and representation approaches of stationary stochastic phenomena. These approaches have their origins in the theory of multivariate stochastic processes, time series analysis and in the algebraic geometric theory of stochastic systems. The stochastic representations most frequently used are the auto-regressive-moving-average (ARMA), matrix-fraction-descriptions (MFD) and the state-space representations. It is shown how to derive these – forward and backward – representations and their dual forms from the analytic and co-analytic spectral factors. These representations are parametrized by system invariants reflecting the four basic Kalmanian principles of controllability, observability, reachability and reconstructibility.       Detailed structure of the state-space realizations is provided using geometric (Hilbert-space) principles including the analysis of the zero structure and balanced realizations.       The structure of generalized Wiener-Hopf factorization is studied in details, first using geometric consideration, then computing the system matrices.       The perspective provided by this can be interesting for those who are doing research in signal processing, stochastic modeling and system identification, or in control system design. Parts of the text can also be useful in courses on stochastic systems, filtering, prediction or on realization theory.