Research On Recognition Of Intra-pulse Modulation Mode Of Radar Radiator Signal

The recognition of the pulse modulation mode of the radar radiation source signal is an important research technology in electronic warfare(EW).The recognition accuracy represents the advanced level of electronic countermeasures technology.To some extent,it determines the outcome of the war.With the rapid development of modern radar technology and the increasingly complex battlefield environment,the method of using traditional features to form pulse descriptor word(PDW)for radiation source identification has been unable to identify today's diverse signal types.The radar signal modulation methods used in electronic warfare are becoming more and more diverse,how to realize the recognition of a wide range of signal modulation methods and radar signals lacking in the radar database have become a new challenge in the current age.The main research contents are as follows:Firstly,the traditional pattern recognition idea is used to extract the features of the signal and then recognize it through the classifier.The specific process is to extract the frequency spectrum sequence of the radar emitter signal through Fast Fourier Transform(FFT),and then perform feature extraction on it to obtain the harmonic mean box dimension and the information dimension feature;through the Choi-Williams distribution(CWD)obtains the two-dimensional time-frequency diagram of the signal,extracts its singular value entropy and peak time-domain coordinate features,and then uses Support Vector Machines(SVM)for intrapulse modulation recognition.The simulation results show that this method can identify ten kinds of radar signals,and the performance advantages of this algorithm are demonstrated through comparative experiments.Secondly,in order to realize a wide range of intra-pulse modulation recognition,a radar source signal intra-pulse modulation recognition algorithm based on Convolutional Neural Network(CNN)is studied.The algorithm performs a series of preprocessing on the radar signal and then the time-frequency map,including normalization and binarization,morphological denoising and resizing,and finally builds a CNN model and inputs the preprocessed image into it.,in order to realize the recognition of the intra-pulse modulation mode of the radar radiator signal.Experiments show that this method can effectively identify 10 kinds of radar signals,and it has stronger anti-noise performance than traditional recognition methods.Finally,for the recognition of multipath radar signals that are not in the database,a multipath signal recognition algorithm based on transfer learning is proposed.When there are only direct path signals in the database,this algorithm can combine with newly received unlabeled multipath signals for model training,which effectively improves the recognition performance.In the design of the deep learning model,the transfer learning distance measurement loss function is added on the basis of the convolutional neural network,and a dynamic learning factor is designed to improve the generalization ability of the network's feature learning.Experiments show that the algorithm can effectively improve the recognition performance of unlabeled multipath signals.