Research On Radar Signal Sorting Based On Intrapulse Features

Radar signal sorting is a key technology of electronic intelligence reconnaissance,which has an important impact on electronic warfare.With the increasingly complex electromagnetic environment and increasing pulse density,the traditional five-parameter signal sorting can no longer meet the requirements.How to extract effective features from radar pulses is the focus of current radar signal sorting research.The research content of this thesis is as follows:1.Aiming at the problem that traditional parameters cannot meet the signal sorting requirements,this thesis proposes a signal sorting method based on the combination of inter-pulse features and intra-pulse intentional modulation features.In this thesis,the frequency spectrum fractal box dimension feature of the pulse signal is extracted,which is combined with the inter-pulse carrier frequency and pulse width features as the input features for pre-sorting,and an improved density peak clustering method is proposed in combination with K-means clustering and density peak clustering.The clustering presorting is realized,and finally the main sorting is realized by combining the improved PRI transformation method,so as to achieve the purpose of radar signal sorting.2.Aiming at the completely different interpulse characteristics and intrapulse intentional modulation characteristics of different modes of multimode radar,a signal sorting method based on intrapulse unintentional modulation characteristics is proposed.Firstly,based on the phase noise model and the nonlinear model of the power amplifier,different radar emitter individuals are constructed,carrying different unintentional modulation information,and then the fractal box dimension and energy characteristics of the decomposed signal are extracted according to the variational mode decomposition.Bispectral estimation extracts features such as skewness,energy entropy,and contour integral waveform entropy.A total of five-dimensional features are used as unintentional modulation features.Different classification algorithms are used to realize the sorting and identification of radar individuals.The performance of different classifiers is simulated and compared,and Indicates the effectiveness of the extracted features.3.In view of the complex process of manual extraction of unintentional modulation features and the possible omission of information,the convolutional neural network is used to automatically extract unintentional modulation features to realize the sorting and identification of radar individuals.It is proposed to realize time-frequency analysis based on variational mode decomposition-Hilbert-Huang transform to obtain a time-frequency graph,which is input into the VGG model and ResNet model for signal sorting research,and the simulation shows that the automatic feature extraction of convolutional neural network is realized Sorting is better than manually extracting features to achieve sorting.