Deinterleaving Technology For Radar Emitter Signals Based On The Intra-Pulse Features

Radar signal deinterleaving is to separate the pulse trains interleaved in time into individual emitter groups, where pulses in one group belong to the same emitter. It is basic and important function in the electronic intelligence system and electronic support measure systems and it directly determines performance of electronic reconnaissance equipement. The aim of signal deinterleaving is to provide corresponding pulse sequence to signal analysis, measurement and recognition. Nowadays conventional methods of signal deinterleaving use each pulse signal parameters, i.e., time of arrival (TOA), radio frequency (RF), pulse width (PW), pulse amplitude (PA) and direction of arrival (DOA). This technique works well when the signal is received in low noise, and the signal belongs to continuous wave radar emitter. However, as the countermeasure activities in modern electronic warfare are becoming more and more drastic, the density of electromagnetism signals environment arrives at the degree of mega and the application of new complex waveform modulations in modern radar, which has the ability to result in pulse loss, parameter overlapping heavily and destroy the regularities of signal sorting. Thus, the conventional inter-pulse techniques may not to be enough to distinguish them one from another in such environment. In recent years, with the development of modern digital signal processing technology and VLSI, the radar medium frequency receiver can obtain radar signals digital waveform. This seems to be enough to put them into practice using intra-pulse feature to achieve high-accuracy sorting radar emitters in high density and high interleaving environment. In this dissertation, we focus on the problems of deinterleaving based on unknown radar pulse signals and lucubrate from three facts when the classical pulse sorting methods are badly destroyed, i.e., the intentional and unintentional modulation feature extraction and performance analysis for radar signals, the cluster number estimation of radar emitter signals and the novel intrapulse-based sorting approach related to cluster algorithm. The main research results are described as follows:1. In order to research on intentional modulation characteristics of advanced radar emitter signals, the feature extraction algorithms based on symbolic time series analysis(STSA), wavelet transform in spectrum and autocorrelation function of first difference are proposed. These algorithms can provide new parameter vectors which are advantage to engineering application.Because the waveform of radar emitter signals have compact distribution and good shape in frequency domain, an approach for intra-pulse feature extraction of radar emitter signals based on STSA is proposed. It is efficient to obtain quantitative information from signals. Embedding time-delay and modified Shannon entropy are used as two-dimensional feature vector to sort the interleaving radar signals. The time-delay feature can determine the length of symbol series. The entropy feature can quantitatively reveal deterministic information and complexity of radar intra-pulse modulation signals. Experimental result and data analysis indicate that the features of seven typical radar emitter signals extracted by STSA have good characteristics of clustering and suppressing noise. Moreover, the algorithm is of advantage to engineering application and implementation because of its computational simpleness, efficiency and capability of simplifying sorter.In fact, radar emitter signals are always interfered with by plenty of noise in the process of transmission in the air and in the process of receiving and processing in scout. This paper presents an approach for intra-pulse feature extraction of radar emitter signals based on wavelet transform domain filtering. It is efficient to obtain quantitative information from signals. The energy entropy from approximation coefficients of wavelet transform and the other energy entropy from inner-scale correlations denoise of detail coefficients are used as two-dimensional feature vector. Experiment results demonstrate that the features of ten typical radar emitter signals extracted by wavelet transform have good performance of noise-resistance and clustering when SNR is OdB.An approach for intra-pulse feature extraction of radar emitter signals based on the first difference autocorrelation function is proposed. The envelop feature extraction of autocorrelation function involves the use of first difference operation of radar emitter signals, which can enhance the difference in modulation information and has good noise abatement. The criterion, defined as degree of separability, is used as two-dimensional and three-dimensional feature vectors selection. Experiment results demonstrate that the features of six typical radar emitter signals extracted by autocorrelation function have good performance of noise-resistance and clustering when SNR varies from-5dB to OdB.2. Specific emitter identification(SEI) technology extracts subtle and persistent features from received pulse signal to create a fingerprint unique to specific radar. Radar emitter signals have their individual feature due to inevitable transmitter phase noise. An approach based on surrounding-line integrated bispectrum is proposed to extract unintentional phase modulation features caused by oscillator. The quantitative features, i.e. bispectra entropy, mean and waveform entropy of surrounding-line integrated bispectrum, is further extracted from bispectrum to reveal the individual difference between emitters. Computer simulation results show that the quantitative features can classify emitters using individual difference under moderate SNR.Examining the unintentional shape variations in pulse envelope can provide insight into the transmitter type and may help to identify the signal. In this paper, the approach for estimating the number of emitters is proposed based on model selection of eigenvalues from principal component analysis(PCA) of pulse envelope vectors. A novel information theoretic criterion is formulated for determining the number of emitters. When compared with the other information theoretic criterions, computer simulations show that the effectiveness and feasibility of the proposed approach.3. Aiming at the issue of complicated characteristics distribution and undistinguishable boundary between clusters of radar signals, a deinterleaving models using clustering algorithm based on grey relational analysis is proposed. And the theoretical problems relative to these models are discussed in detail.A divisive hierarchical clustering algorithm based on grey relational measure is proposed. In the algorithm, grey relational analysis is used to measure the degree of similarity between data set. On the basis of generation of top-down density-based hierarchical partitions of data set and proposed clustering validity index, the extremum of the index curve is used to estimate the number of clusters. Computer simulation results on Benchmark data set and synthesis data set demonstrate that the proposed algorithm is feasible and has good clustering performance especially for arbitrary shaped clusters.In consideration of some factors, such as the number of radar emitters, the amount of intercepted pules, pulse loss rate and signal-noise-ratio etc, the performance of deinterleaving algorithm based on clustering are studied in deep. The experimental result on intra-pulse features vectors of radar signals and different combinations of features vectors show that the clustering algorithm based on grey relational measure has good performance in sorting radar emitter signals and has good stability under the above factors. Additionally, the experiment gives some useful conclusion and reference in chosing new effective deinterleaving feature vector.