Recognition of predetermined signals in the automated radio monitoring problems

Abstract

In practice, besides the signals predetermined in a probabilistic sense, the unknown signals for which the learning sampling cannot be obtained. In this case the classical recognition methods cannot be used, this results in the need for development of nontraditional signals recognition methods taking into account the presence of the unknown signals class. The distinctive feature of the given work is the signals recognition methods concretizing for the case of the signals description with probabilistic models in the form of auto regression processes and mixtures of random signals. The decision results of the typical recognition problems in automated radio monitoring with using this recognition methods are considered. When solving the problems of the specified radio transmission types recognition, the decision rule based on the signals’ auto regression model was used. Investigations were performed using the statistical simulation method with the samplings of radio signals for 10 different types of radio transmissions peculiar to the problems of the automated radio monitoring. The mean probability of correct recognition 0,95 was obtained. When solving another problem of radio monitoring - recognition of the type radio signals modulation - the decision rule, based on the type of the model of distributions’ mixtures was used. Investigations were performed with the sampling of radio signals from 5 different types of modulation typical for radio monitoring. The mean probability of correct recognition 0.9 was obtained.