Research On Modulation Recognition Technology Of Communication Signal Based On Deep Learning

In the non cooperative communication scenario of electronic warfare,the modulation mode of received communication signal is unknown,and modulation recognition is a key step before signal demodulation.In the civil field,the rapid development of intelligent receiver in recent years requires more and more adaptive demodulation.How to effectively identify the modulation becomes an urgent problem in the military and civil communication field.The electromagnetic environment is becoming more and more complex,which makes the signal-to-noise ratio range of the received signal larger.At the same tim e,there are many signals to be identified in the same channel in the time domain,frequency domain and spatial domain.At this time,the traditional recognition method after signal separation fails,so it is necessary to find a direct recognition method.Aiming at these two problems,this paper studies the modulation recognition algorithm of communication signal based on deep learning,which mainly includes the following contents:First of all,based on the wideband if software radio receiver with modulati on recognition function,a signal model of wide range SNR and aliasing is built and the data set used in this algorithm is generated according to the actual communication signal format,parameters and channel conditions.Secondly,the convolution neural network is used to automatically extract features,extract time-domain signal features and realize signal modulation classification using the last layer of the network,so as to solve the problem of poor noise robustness caused by the failure of traditional features when the signal-to-noise ratio of a single signal of interest changes widely.Thirdly,based on the co-existence signal modulation recognition algorithm of capsule network,two kinds of routing methods,dynamic routing and EM routing,are used to solve the problem of time-frequency spatial overlapping of signals in the same channel,the signals to be recognized are not fixed,and they are interested in their modulation methods.The purpose is to improve the recognition accuracy,reduce the missed detection and false alarm rate,and improve the noise robustness and generalization ability.At present,there are two signal aliasing in the related research.In the work of this paper,the maximum number of coexisting signals is three.