Type recognition of the digital modulation of radio signals using neural networks
С. С. Аджемов, Maxim V Tereshonok, Denis S. Chirov · Moscow University Physics Bulletin · 2015
The results of the application of artificial neural networks for the problem of type recognition of the digital modulation of radio signals are presented. The cumulants of the second and fourth orders, as calculated according to the values of recorded signal samples, are proposed as a recognition sign. The choice of informative features (in this case, a set of cumulants) and the formation of the decision rules are carried out in an expert way from a set of types of signal modulation that are required for recognition. Studies in this field show that the formation of distinction rules (classifiers) may be carried out using various intelligent analytical tools, such as artificial neural networks. The use of a multilayer perceptron as a classification (recognition) device enables one to automate the process of the formation of the decision rules for type recognition of the digital modulation of radio signals. This method provides a sufficiently high (0.7–0.99) probability of the correct recognition of signals with frequency (FSK), phase (PSK), amplitude (ASK), and quadrature phase (QAM) shift keying in the absence of synchronization of the receiving system at the carrier frequency.