recognition of images
the determined and random radio signals
Intensive development of means of a radio communication leads to necessity of system engineering of the control of radio-electronic environment. Perfection of methods of modulation allows to form low-energy signals which are difficult for finding out and classifying. All it makes more hard demands to methods of classification of random low-energy signals. The analysis of tendencies of development of methods of detection and definition of parameter’s of radio signals has shown that the greatest efficiency and autonomy the intellectual methods of processing based on use neural network algorithms possess. In article the general classification of available neural networks is presented and the greatest efficiency of use of a neural network of counter distribution for the decision of a problem of classification of radio signals is proved. Various variants of updating of the given network are shown at the decision of a problem of classification of the determined and casual radio signals. The analysis of the received results has shown that use of neural networks allows to lower considerably a threshold of authentic classification of the casual and determined radio signals. At modeling samples ФМ-2, ФМ-4, ФМ-8, 16-KAM, ЧМ and АМ signals are used. Research of dependence of probability of correct classification from number of neuron’s in a competing layer of a neural network of counter distribution is conducted.