Research on network communication signal processing recognition based on deep learning

With the popularization of wireless communication technology, the modulation of wireless signal not only improves the information transmission, but also can realize encryption and anti-interference processing. For the unknown signal, it is necessary to determine its modulation type before demodulating the real signal, so as to determine whether the signal is legal. This study introduced back-propagation (BP) neural network and convolutional neural network (CNN) and applied them to the modulation type recognition of wireless communication signals. In order to improve the recognition accuracy of CNN model for modulation signals, the steps of drawing signal constellation diagram were added on the basis of original CNN. Then the simulation experiments were carried out on the BP, traditional CNN and improved CNN models by using MATLAB software. The results showed that the constellation could effectively reflect the modulation type characteristics of the modulation signal; in the model training process, the improved CNN model had the fastest convergence and the smallest training loss when the convergence was stable, followed by the traditional CNN model, and the BP model had the slowest convergence and the most loss when the convergence was stable; with the increase of the signal-to-noise ratio of the detection signal, the average accuracy of the three recognition models showed a tendency of stable after increasing; under the same signal-to-noise ratio, the improved CNN model had the highest recognition accuracy, followed by the traditional CNN model and BP model. © 2020 Begell House Inc.. All rights reserved.