E-catalog        

Библиогр.: 13 назв.

We introduce the technically simple approach to the determination of the abrupt change of the unknown dispersion of the high-frequency fast-fluctuating Gaussian random process against white noise with an unknown spectral density. For this purpose, we determine new approximations of the decision statistics for various hypotheses, we carry out their maximization on unknown parameters, and we develop the block diagrams for the corresponding detector and measurer in the form of the comparatively simple single-channel units. For the analytical analysis of the performance of the synthesized algorithms, the asymptotically exact expressions for their characteristics, specifically - type I and type II error probabilities (when an abrupt change point is detected) and conditional biases and variances of the estimates (when measuring the parameters of the analyzed random process), are obtained by means of local Markov approximation method. We also illustrate a new procedure for determining the distribution law and the central moments of the estimate of the discontinuous parameter (abrupt change point), with an allowance of anomalous effects. The experimental testing of the presented theoretical results is implemented by the methods of statistical computer simulation.