Research Of Specific Radar Emitter Identification Based On Characteristics Of Unintentional Modulation On Pulse

With the conflict between the electronic counter measures and anti-electronic counter measures becoming fiercer day by day in modern warfare,the increasing noise in electromagnetic environment and the development of counter-measures.,the identification of enemy radar emitters is becoming more and more complex.And the identification relying only on the traditional parameters is becoming less and less ideal.To solve the identification of signals in new environment,this paper use ambiguity function,research the extraction,optimization,fusion and classification algorithm of the radar signals intra-pulse unintentional modulation.The main research results are as follows:1.In view of the problem that the identification accuracy of signals under traditional Gaussian noise is not ideal.An ambiguity function(AF)based feature extraction and optimization method for unintentional modulation recognition of radar emitters is proposed,which utilizes the radial-cut.The algorithm studies the relationship between the radial-cut of AF and fractional Fourier transform(FRFT),gets the cuts by FRFT.Then,two kinds of cut-concatenation schemes are designed to construct two different pairs of feature vectors.Finally,Canonical Correlation Analysis(CCA)and kernel Canonical Correlation Analysis(KCCA)are used to optimize and fuse the extracted feature.The experimental results show that the algorithm can effectively improve the identification accuracy of signals,and realize the signal identification and recognition under the condition of low SNR.2.In view of the problem that the time-frequency analysis algorithm fails due to the fact that the noise signal does not have limited two-order statistic under the stable distribution,a fractional low-order AF based on four nonlinear transforms is proposed.The algorithm extracts signal unintentional modulation feature under stable distribution noise by FRFT without the future knowledge of distribution,and is more practical than fractional low-order covariance.Then,the cut-concatenation schemes and the CCA,KCCA are also used to optimize and fuse the extracted feature.The experimental results show that these four nonlinear transforms can all identify LFM and BPSK signals with ? stable distribution.The fractional low-order AF based on Sigmoid transform is the best.It can still have better performance under low MSNR.3.In view of the problem that the traditional least squares support vector machine(LSSVM)is based on all the eigenvectors,which leads to the decrease of the generality and sparseness,and becomes slow in large sample number problems.PCP-LSSVM is proposed based on the primary LSSVM using the Cholesky decomposition iteration.At the same time,the chi-square kernel has better performance than the Gaussian kernel under the condition of sparse solution,so the Chi-Gaussian combined kernel function is used as the kernel function of the LSSVM.The genetic algorithm is used to obtain the kernel function of the optimization parameters.The experimental results show that the sparse LSSVM with combined kernel function have better classification effect and classification speed on the smaller sample set UCI and the identification of the actual transmitter signals.It also has good performance in the signal identification of intra-pulse unintentional modulation.