Researches On Identification Key Technology For Radar Emitter

To obtain the dominance of the modern warfare, it is important to acquire the information of hostile radars, which makes radar emitter identification (EID) to be one of the most important modules in modern electronic reconnaissance systems. Based on the analysis of the requirements of EID, some subjects are deeply studied in this dissertation, including the modeling of radar emitters, the identification of emitters with interval and fragmentary feature parameters and the identification of multi-function radar (MFR) emitters. The main contributions can be briefly described as follows:In chapter 2, the modeling of radar emitters is studied. To solve the emitter identification error problem caused by the complex operations of hostile radars, a method of joint-parameter modeling (JPM) is proposed. This method is capable of modeling the scalar, interval, function, sequence and fragment parameters in both pulse-level and statistical parameter level. Based on the model of joint-parameters, a new identification algorithm is deduced by using JPM in identification processing and by constructing different identification methods for different models of the emitters.In chapter 3, the identification of radar emitters with interval feature parameters is investigated. Firstly, models of the interval uncertainty are described. Secondly, to deal with the error identification problem of EID caused by the interval uncertainty of measurement feature parameters, a new identification algorithm is proposed, which is based on the combination of vector neural networks (CVNN). This algorithm realizes the nonlinear mapping between the interval-value input data and the interval-value output emitter types. And, it requires less computational load in the training stage. The key idea of CVNN is to adopt a combination of multiple Multi-Input/Single-Output neural networks to construct an identification system. Thirdly, to treat with the emitter identification error problem caused by the vector neural network (VNN), which is incapable of processing the linguistic information and considering the reliability of the training samples in training phases, an identification method based on cloud model and vector neural network (CMVNN) is proposed. The new method can not only process the linguistic input data, but also process numerical input data. Finally, to solve the emitter identification error problem caused by the feature measurement uncertainty of multi-sensors, an interval Dempster-Shafer Theory (IDST) based identification algorithm is introduced. This algorithm combines the upper bound and lower bound of the intervals respectively, and then deduces the emitter type according to the fused confidence interval.In chapter 4, the identification of radar emitters with fragmentary feature parameters is studied. Firstly, models of the fragmentary uncertainty are described. Secondly, to deal with the EID error problem caused by the incomplete feature parameters of the template radars and the emitter type missing, it proposes a vector neural networks based incremental learning (VNNIL) approach for emitter identification, which combines vector neural networks (VNNs) and the ensemble-based incremental learning (Learn++) algorithm. The key idea of VNNIL is to construct weak classifiers by VNN, and then ensemble them with Learn++. The VNNIL can treat the classifier design and the emitter database automatic updating as a single task, and realizes the emitter identification and parameters updating simultaneously. Finally, to resolve the EID error problem caused by the fragmentary feature parameters of the template radars, a missing data imputation (MDI) based method is advanced. This method utilizes the modified CVNN to substitute the missing feature parameters, and takes use of the substituted training samples for training CVNN to obtain the structure parameters of the network. The advantage of the MDI algorithm is that it depends only on the training samples in the imputation processing, and requires no transcendent knowledge about the samples'distribution.In chapter 5, the identification of MFR emitters is discussed. Firstly, to dispose the radar words error extracting problem caused by dropped pulses and false pulses, a three-level matching (TLM) based extracting method is proposed. The TLM can not only correctly extract the radar words, but also have better adaptability to noisy environment. Secondly, to deal with the MFR emitter identification error problem caused by ignoring probability distribution of the production rules, a new identification method based on stochastic grammar (SG) is put forward. This method is deduced from the syntactic modeling of multi-function radars. And, this method constructs stochastic automaton for identification according to the radars'grammar. Simulations are presented to demonstrate the identification capabilities of the SG algorithm. Finally, to solve the EID error problem caused by the radar words'uncertainties of measured MFR emitters, a new identification method based on stochastic syntax-directed translation schema (SSDTS) is proposed. The SSDTS is deduced from the syntactic modeling of multi-function radars. And this method can not only correct the error radar words by using the stochastic translation schema, but also identify the work mode of measured emitters at the same time.