Research On Intelligent Recognition Of Modulation Mode For Communication Radiation Source

Communication reconnaissance can provide intelligence support for precise and efficient military strategic decision-making,and is one of the important means of intelligence reconnaissance.The recognition of the modulation mode of the communication radiation source plays a very important role in the fields of communication reconnaissance,cognitive radio,satellite measurement and control,etc.The correct recognition of the modulation mode of the radiation source signal is the basis for the accurate demodulation and decoding of the subsequent signal.The rapid development of communication technology has brought great challenges to communication reconnaissance.At the same time,with the rapid development of computer technology in recent years,the speed of graphics computing has been unprecedentedly improved,artificial intelligence technology has been widely used.As an "end-to-end" learning framework,deep learning does not require too much manual intervention in its feature extraction process,and can achieve sufficient data mining.Significant research results have been achieved in the fields of signal processing and target detection and recognition.In the field of communication reconnaissance,the introduction of artificial intelligence technology has greatly improved the performance of identifying the modulation mode of communication radiation sources.Therefore,it is of great significance to study the intelligent recognition of the modulation mode of the communication radiation source.However,the existing deep learning-based communication radiation source modulation method recognition methods have limited feature extraction ideas,limited network feature representation capabilities,and unclear physical meanings,resulting in unsatisfactory network recognition performance.Therefore,how to make full use of the essential characteristics of radiation source signals,combined with the powerful nonlinear mapping ability of deep learning to improve the performance of existing radiation source modulation methods,has important research significance.In addition,with the complexity of the battlefield environment and the continuous development of modern communication technology,the modulation mode of the communication radiation source has developed from a single mode to a composite mode,and has been gradually applied to various modern communication systems.It makes up for the weakness that a single modulated signal is easy to be intercepted and interfered,and improves the reliability and communication quality of the communication system.Existing modulation recognition methods for complex modulated radiation sources mostly focus on extracting characteristic parameters or adding signal demodulation techniques.Such methods rely heavily on the types of complex modulation radiation sources to be identified and lack universality.Moreover,the data point distribution characteristics of the composite modulation mode are relatively similar,and it is difficult to directly use the traditional recognition algorithm of the modulation mode of the communication radiation source to identify.Therefore,how to use the deep learning method to recognize the modulation mode of the complex modulated radiation sources is an urgent problem to be studied.This dissertation is funded by the Innovation Special Field project of KJW and the Horizontal project.It focuses on the recognition of radiation source modulation based on SERes Net＿IRNN-SF(Squeeze and Excitation Residual Networks Identity Initialized Recurrent Neural Networks Statistical Features)and the complex modulation radiation source modulation based on Time Frequency Deep Convolutional Neural Networks(TF-DCNN).The main research contents are summarized as follows:1.The signal models of different communication radiation sources are studied and their characteristics are analyzed.There are many modulation methods for communication radiation source signals,and different modulation methods of signals show different characteristics in multi-dimensional domains such as time domain and frequency domain.Therefore,the research on the signal model and modulation characteristics is the basis for the recognition of the modulation mode of the communication radiation source.First,the analog modulated signal and digital modulated signal models are constructed,and the modulation characteristics of different signals are studied through the time-domain waveform,spectrogram,and constellation diagram of the signal.The signal model of the composite modulated radiation source is constructed,the modulation principle of the composite modulated radiation source signal is studied,the spectral characteristics of the composite modulated radiation source are analyzed,and the correctness of the conclusions obtained is verified by the spectrogram of the simulated signal.The recognition of the modulation mode of the source lays a theoretical foundation.2.Aiming at the problems of limited signal feature extraction ability and weak interpretability of the existing intelligent recognition method of modulation mode of communication radiation source,a modulation mode recognition method of radiation source based on SERes Net＿IRNN-SF is proposed.Firstly,the physical mechanism of the instantaneous characteristics and High-order Cumulants(HOC)characteristics of the signal is studied,and the statistical characteristics of the radiation source signal under different modulation methods are extracted;secondly,the SERes Net＿IRNN-SF network is constructed,and through fusion Residual Networks(Res Net)and Identity Initialized Recurrent Neural Network(IRNN),pay attention to the spatial and temporal characteristics of signals,and add SENet(Squeeze and Excitation Networks)between modules,introducing The channel attention mechanism adds weight to each feature map and maximizes the use of the feature map extracted by the network;at the same time,the features extracted by the neural network and the statistical features of the signal are integrated,and the multidimensional modulation information of the radiation source signal is fully utilized to realize radiation.Intelligent recognition of source modulation mode;finally,the effectiveness and robustness of the proposed method are verified on the simulation data set and the public Radio ML2016.10 a data set,which provides theory and technology for the subsequent research on modulation mode recognition of complex modulation radiation sources support.3.Aiming at the problems that the existing modulation recognition method of complex modulation radiation source relies on expert knowledge and accurate demodulation technology,and has poor universality,a modulation recognition method of complex modulation radiation source based on TF-DCNN is proposed.Firstly,based on the modulation principle of the complex modulation signal and the instantaneous phase of the signal as the cut-in point,the time-frequency characteristics of different kinds of complex modulation signals are studied by using Short-time Fourier Transform(STFT).Secondly,the modulation difference of the complex modulation radiation source is converted into the difference of time-frequency images.In view of the powerful image processing capability of Convolutional Neural Networks(CNN),a TF-DCNN is built to fully extract the time domain and frequency domain of the complex modulation signal.Finally,the effectiveness and robustness of the TF-DCNN algorithm are verified by using the simulation data set.The results show that the proposed method has better recognition performance.