Research On Unintentional Modulation Feature Analysis And Identification Of Specific Pulsed Emitter

In order to make strategic judgments and tactical decisions accurately in modern electronic warfare(EW),it is necessary to obtain accurate battlefield information,which desires a clear demand for emitter recognition.Conventional emitter recognition can provide basic information,such as the type,band and modulation parameters of the electromagnetic target,but it cannot solve the problem of distinguishing different individuals of emiters with the same,and even cannot identify the same emitter operated with wartime parameters.The present emitter recognition is unable to meet the increasingly sophisticated demands of recognition.To solve the above problems of emitter recognition,the thesis studies on unintentional modulation feature analysis and identification of specific pulsed emitter,proposes a new technical framework of specific emitter identification(SEI)to recognize different emitters of the same type with same modulation parameters.Some theoretical analysis and simulation,semi-physical experiments and real data verification are exploited to prove the practical performance of the proposed algorithms.The contents of this dissertation consist of the following four parts:(1)This thesis focuses on the modeling of multi-component signal,which is defined by a mixed signal of multiple emitter individual pulsed signals,and the time frequency analysis.After the signal modeling according to real radar signals,to solve the time-frequency information cross-interference and fuzzy problems of a multi-component signal,a instantaneous frequency estimation algorithm based on the adaptive fractional spectrogram is proposed,which is the foundation of time-varying filtering and feature extraction presented later.To solve the mode mixing problem which always occurs in convenient empirical mode decomposition(EMD),a mode mixing suppression algorithm is proposed based on self-filtering method using frequency conversion.Comparing with the most advanced masking signal method for EMD(MS-EMD),the suppression performance can be increased by about 26%,which helps the feature extraction as foundation.(2)The intra-pulse feature extraction technique for SEI is developed to characterize the individual differences of unintentional modulation on pulse(UMOP).The UMOP features are mainly focused on the individual differences in pulse envelopes and the weak individual characteristics caused by phase noise.Then an intrinsic mode function distinct native attribute(IMF-DNA)feature extraction algorithm and a joint feature selection(JFS)algorithm based on majority vote algorithm are proposed to extract envelope features,which together constitute the final proposed SEI technique with better generalization ability for modulation parameters.Compared with radio frequency DNA(RF-DNA),IMF-DNA had a obvious improvement of correct emitter identification rate.For the phase noise feature extraction,a method using fractal features based on box-counting dimension(BCD)and variance dimension(VD)is presented.Generally speaking,fractal feature based on BCD or VD has better performance than RF-DNA feature and convenient surrounding-line integrted bispectra(SIB)with lower feature dimension.A real data verification method was developed to verify the performance of IMF-DNA for specific emitter identification,which achieves a correct identification rate of 85.3% and 83.7% for VD in comparision.(3)To characterize the inter-pulse UMOP features,this thesis studies the specific emitter identification using the unintentional modulation characteristics of frequency drift during the emitter start-up process.Frequency drift models based on physical phenomenon is established and the frequency drift parameters are modulated into the pure signal of a typical radar.Then the thesis focuses on the common phenomenon of frequency drift,defines geometric features of frequency drift curve and finally proposes a practical algorithm of specific emitter identification using the geometric features,which has anti-t-flexibility ability.The proposed feature extraction algorithm consists of signal segmentation,feature geometric point selection and feature composion.Simulation results and real data experiment verifies the practical performance of the proposed algorithm.(4)The UMOP features used for specific emitter identification are usually weak comparing with primary signal,which will worse the performance of SEI seriously if no further measures taken to remove the fluence from primary signal.To solve the above problems,a feature extraction enhancement algorithm is exploited in this thesis.A time-varying filtering(TVF)algorithm based on order time-varying short-time fractional Fourier transform(OTV-STFr FT)is proposed separate the primary signals from a multi-compoent signal,which is a preprocess before feature extraction.To solve the poor robustness of SEI caused by the weakness of extracted features,a feature extraction enhancement algorithm based on OTV-STFr FT is proposed by the thesis.Simulation results show that the improvement of correct identification rate after feature extraction enhancement is significant.In addition,a multiple kernel learning(MKL)algorithm is developed to fuse the multi-domain features of envelop features from time domain,BCD or VD features of phase noise from fractal domain,SIB features of phase noise and geometric features of frequency drift curves from frequency domain,which achieves a better performance of final SEI.The simulation results show that the improvement of correct identification rate is more than 5%.The real data verification proves that that the feature extraction enhancement can improve the correct identification rate of SEI and the feature fusion with MKL can achieve a further 4% improvement.