A. V. Kuzovnikov

Prospects of development of systems of radiomonitoring consist in introduction of new methods of digital processing, which will allow in an automode (without participation of the operator) to spend processing for the purpose of determination of structural parametres of input signals. In this article the question of automatic determination of type of modulation of parametrically undefined signal is considered. The analysis of possibility of use of neural networks for the assigned task decision is passed. The analysis of various types of neural networks has allowed to reveal the most effective structure of a neural network for classification of modulation of an input signal. Results of modelling of a neural network of counter propagation have shown high efficiency of use of the given type of a neural network which is expressed in high probability of correct detection of an input signal. For determination of a terminal conditions of applicability of a neural network of counter propagation numerical modelling has been spent. Results of modelling have shown, what even for low- energy BPSK signal with a signal/noise ratio the minus 50 dB at increment the number of neurons with 10 to 200 probability of correct detection is increment from 20 % to 70 %. Imperfection at use of neural networks is big enough demanded operation speed of digital system of processing