Research On Radar Signal Sorting Using Multi-dimension Information Features

Radar reconnaissance is a significant means to acquire intelligence information in modern electronic battlefield.As a key technology of radar reconnaissance system,radar signal sorting deinterleaves the signals that belong to different emitters from interleaved radar pulse sequences according to the correlation of signals from one emitter and the difference of signals from different emitters,which provides reliable sorting information for the feature analysis,feature matching and recognition of emitter signals as well as the judgment of the working states and purposes of emitters.With the rapid development of modern radar technology,the radar with multiple working modes appear in modern electronic battlefield,and emitted signals adopt complex interpulse and intrapulse modulations,which complicate the signal environment in battlefield and bring severe challenges for radar signal sorting.Aiming at some practical issues confronted by radar signal sorting in complex electromagnetic environment,the dissertation analyzes diversity,variety and low probability interception of complex radar signals.On this basis,the research processes the intercepted radar signals in different dimensions of information,digs out novel signal features and proposes new sorting methods,which are applicable to the signals from multiple advanced radars as well as the signals with complex modulation.Achieved research achievements are as follows:1.A method for blind separation and reconstruction of MIMO radar signal in underdetermined condition is proposed.The method provides strong supports for feature extraction and signal sorting in waveform dimension by separating and reconstructing intercepted overlapped signals.Existing radar signal sorting methods can hardly separate multi-sensor radar?e.g.Multiple-Input Multiple-Output radar?signals and extract their feature parameters because the signals emitted from different channels overlap severely in spatial,time and frequency domain.Especially when the number of reception channels is less than that of transmission channels,underdetermine problem inevitably causes multiple solution.Utilizing the time-frequency sparce characteristics of MIMO radar signals,the dissertation proposes a underdetermined blind separation for Discrete Frequency Coding Waveform MIMO radar signals.First,the proposed method combines time-frequency independent complex-argument point detection and guidance clustering based on field strength of data field to estimate the mixing matrix.Then reconstruct the transmitted signals using l₁ norm minimization algorithm and frequency coding curve filtering.The proposed method starts signal sorting process at receiving end,and estimate the mixing matrix and frequency codes with high accuracy.The signals transmitted from different channels can be effectively separated and recovered,which ensures the correctness of extracted feature paramters and improves the effectiveness of signal soring in waveform dimension2.For the intercepted radar full pulse sequence,the dissertation regards the variation of pulse paramters as an information dimension and analyzes the changing characteristics of interpulse parameters.And a method is proposed for extracting the periodic inter-pulse features of radar signals.The method preprocesses the 2-dimentional sequence comprised of Direction of Arrival and Radio Frequency parameters.Then Singular Spectrum Analysis is introduced so that the extraction of interpulse changing characteristics is transformed into the separation of weak signals.The periodic feature in radar pulse sequence is extracted by separating equivalent signal subspace and spectrum analysis.It is verified through simulation experiments that the presented method can correctly extract the slippage frequency of slippage-frequency signal when lost pulse and slippage bandwidth are within a certain range.3.Regarding that the classification accuracy of complex-modulation radar signals remains low in case of low SNR,the dissertation digs out novel feature parameters in intra-pulse information dimension and proposes a classification method based on distribution features of ambiguity energy and fractal features of main ridge slice of ambiguity function?MRSAF?.On account that ambiguity function fully describes the internal structural information and highlight individual features of signals with different modulations.Besides,we can dig out the features used to distinguish different types of signals since the signals with different modulations correspond to different MRSAF shapes.The dissertation researches the distribution features of ambiguity energy and irregular degree of MRSAF for complex-modulation signals,and extracts rotation angle and fractal dimension of MRSAF as as the feature parameters forming feature vector.Derived feature space has good aggregation within cluster and separability between clusters.Experimental results have proved that higher correct classification rate can be derived when classifying different complex-modulation signals using extracted feature parameters.4.Multi-mode radar signals tend to be classified as multiple emitters when using traditional signal classification methods.To address the“extension”problem,the dissertation adopts the idea of spatial data mining and proposes a multi-mode signal classification method based on data field and cloud model.The method employs theories of data field and views the distribution of potential field formed by data radiation in feature space as an information dimension.The distribution features of parameter samples of radar signals are sufficiently revealed by potential field,and cluster the samples under the guidance of field force.On the basis of membership analysis of clustering results,the final signal classification is accomplished according to the established belonging judgment criterion.Simulation results prove that the proposed method can effectively avoid different working modes of one emitter being classified as different emitters.