Dynamical methods for random processes recognition

AbstractThe problem of multialternative recognition of nonstationary random processes is considered. To describe the classes of the processes, stochastic dynamical models are used. On the basis of recursive and differential equations for sufficient statistics, the methods and algorithms of a sequential recognition of classes, change detection and estimation of the change moment in the properties of random processes are obtained. The methods for the recognition of composite, interval-stationary and local-stationary sequences are worked out. The algorithms of current and group classification are considered