Research On Modulation Patterns And Interference Type Recognition Based On Deep Learning And Time-frequency Analysis

Electronic warfare,especially communication countermeasures,plays an important role in "network-centric warfare" and has a prominent position.Among them,modulation pattern recognition and interference recognition are the key basic problems faced by communication countermeasures.Although from the perspective of technological evolution,these two issues have been deeply studied since they were put forward,and the research history has been more than 70 years,but now,with the breakthrough development of artificial intelligence technology represented by deep learning,while further research in the two fields injects new vitality,it also brings huge challenges.On the one hand,enabling wireless communication based on deep learning will enable both parties to gain a higher degree of intelligence;on the other hand,with the help of the powerful feature extraction and expression capabilities of deep learning,it is expected that more precise modulation patterns and interference type recognition can be achieved.Therefore,for the application field of communication countermeasures,the development of modulation pattern recognition and interference recognition based on deep learning has important practical significance and great practical value.Specifically,the main research content of this article is summarized as follows:?Research the blind modulation recognition algorithm of multi-antenna system.This paper proposes a blind modulation recognition algorithm based on short-time Fourier time-frequency transform,Alex Net transfer learning and decision fusion.Specifically,the windowed short-time Fourier transform algorithm is used to analyze the time-frequency characteristics of the modulated signal received by each receiving antenna of the multi-antenna systems,and the time-frequency spectrogram image of the modulated signal is converted into a colorized time-frequency spectrogram image.Then,the transfer learning is used to fine-tune the Alex Net to adapt to our classification problem,and insert the generated color spectrogram image into the neural network,extract the features and train the network to determine the modulation method of each receiving antenna.Finally,through the decision fusion module,the decision from each antenna of the multi-antenna system receiver is combined to make the final decision,and the final modulated type is obtained.The simulations show that our algorithm has high recognition accuracy no matter in single antenna or multi-antenna network.In addition,the proposed scheme also has a relatively robust performance in frequency selective fading channels.? Research the interference recognition algorithm of frequency hopping communication system.We propose an interference identify algorithm.Specifically,consider the presence of multiple and compound interference in the frequency hopping system,in order to more fully and more distinguishably extract the characteristics of various interferences,we design a composite time-frequency analysis algorithm to calculate the time-frequency distribution of the interference signal.The composite timefrequency analysis algorithm comprehensively considers the time and frequency information of interference signals extracted by linear time-frequency transformation and Cohen bilinear time-frequency transformation methods,and finally generates timefrequency spectrogram image containing two types of time-frequency analysis features as training samples.In the training step,the Siamese neural network is selected as the classifier.The Siamese network determines whether the two samples are the same type by calculating the distance vector of the input samples of the two sub-networks,and matching and identifying the samples by repeated training.The results show that the accuracy of the proposed algorithm is better in most cases,and it can still achieve higher recognition accuracy when the sample size is small.