Recognition Technology For Remote Sounding Radar Emitter

Remote sounding radar emitter recognition is one of the key procedures for satellite defense warning system of ground laser weapon. The results of recognition not only reflect the real-time performance and reliability of electronic countermeasures information but also determine whether the electronic warfare is success or not. On the basis of neural network recognition technology, the rough set theory is used in this dissertation in order to analyze the remote sounding radar emitter signal recognition to obtain a simpler and more rapid, accurate radar emitter recognition method.Firstly, the basic structure of the Electronic detection system is introduced in this paper. Then the common signal of remote sounding radar for single and chirp signal is analyzed. Regular features and intra-pulse characteristics are combined as recognition characteristics. The regular features are estimated of signal with FFT with window noise filter. The common intra-pulse characteristics analysis methods are compared from their advantages and disadvantages. Short-time Fourier transform method is selected to extract the intra-pulse characteristics of radar emitter to estimate the pulse repetition frequency characteristics.Secondly, the structure, algorithm and parameter selection method of Error Back Propagation (BP) and Radial Basis Function (RBF) are analyzed based on emitter characteristics extraction. Five kinds of sounding radar emitter are recognized under different SNR with BP and RBF respectively, and the results are compared also.In the end, according to the fact that the time of training and learning is too long when use traditional neural network with multi-dimensions input characteristic, the rough set of RBF neural network identification system is constructed on the basis of rough set theory. The feature parameters are pretreated and the simplest rule table is produced to generate RBF neural network based on rough set theory. Five types of sounding radar which has multi-working modes have been recognized. The simulation results showed that this method has the advantages of smaller input dimensions and network structure, faster training speed and better recognition results in contrast with the traditional RBF neural network method, which has the practical significance to improve the warning efficiency space-borne electronic reconnaissance system.