Research On The Subtle Feature Extraction Method Of Radar Radiator

How to determine the characteristic information of the enemy radiation source transmitter by intercepting the radiation source signal is one of the important topic for electronic countermeasures.For the problem of identification of individual radar emitters,the micro-feature extraction method of individual radar emitter signal is studied deeply.By analyzing the mechanism of the subtle characteristics of radar transmitters,the fine feature extraction method is studied and implemented from three aspects: time domain,frequency domain,and transform domain in problem in selecting eigenvectors for extracting subtle features of radar emitters.The main content and research results are as follows:1.The research status of the micro source feature extraction method is summarized,and the reasons for individual features generated by the source are analyzed.The phase noise of the source oscillator is elaborated,and the theoretical model of the individual source signal is constructed accordingly..2.In view of the conventional radiation source fine feature extraction method has carried on the simulation and implementation,including time-domain envelope extraction method to extract the signal envelope as an individual radiation sources on the subtle characteristics,the frequency domain of the revised Rife algorithm to extract the frequency deviation as a source of individual subtle features,EMD time-frequency transform domain reconstruction after logarithmic transformation SVD decomposition of singular vector valued as individuals microscopic characteristics of radiation source,through a certain classification of the reliability of the feature extraction method is verified by the experiments.3.A multi-dimensional feature extraction method based on bispectral analysis is proposed.The multidimensional feature vector is constructed by comprehensively using the box dimension,information dimension,LZ complexity of bispectrum diagonal slices,the waveform entropy and energy entropy of bi-entropy diagonal bispectrum,and the Simulation experiments verify the same radar emitter data.Compared with the previous feature extraction methods in time domain,frequency domain and transform domain,the classification effect is improved,which fully demonstrates that the multi-dimensional feature extraction method based on bispectrum analysis has better performance.4.A feature extraction method based on deep belief network is proposed.The sampled radiation source signal is sparse and sent to the deep belief network.The last layer of node value is directly learned as a fine feature of the individual signal of the radiation source through network training and learning.This method is applied to this method.Different radiation source data sets,and simulation experiments were performed to show the effectiveness of this subtle feature extraction method and good classification recognition performance.