Development of methods of classification of processes in discrete time by means of direct and indirect measurements
Abstract
New recognition algorithms operating in conditions of constraint on the observation period and under jamming are proposed. Sequential principle of information processing including the combined system of filtration and classification of signals as they enter is used as the developed approach basis. The proposed approach peculiarity is the use of dynamic trajectorywise description of a signal based on differential equations and optimal rules of stopping allowing to minimize the sum of probabilities of classification errors. Results of simulation on spectrum width classification of signals in conditions of noise interference are listed.