ESTIMATION OF THE PARAMETERS OF HIDDEN MARKOV-MODELS OF NOISE-LIKE SIGNALS WITH ABRUPTLY CHANGING PROBABILISTIC PROPERTIES .2. ESTIMATION OF THE STRUCTURAL PARAMETERS OF THE MODEL

Methods minimizing the Akaike information criterion and validating the statistical hypotheses in order to estimate the structural parameters of a signal model, such as the number of classes of quasistationary fragments and the orders of autoregression of each class, are discussed as applied to long records of signals with varying probabilistic properties. Some results of simulation are presented.