| Specific emitter identification has always been a key part of electronic countermeasures.With the development of science and technology,the electromagnetic environment has become more and more complex.The method of identifying emitter according to conventional features has been unable to adapt to the current complex electromagnetic environment.Aiming at this problem,in this thesis,by comparing biological gene,introducing signal gene features,extracting methods of signal gene features in exploratory research,and combining VGG stacking and ResNet Inception,a method based on VGG-ResNet network combined with gene feature enhancement for radar emitter is proposed.The whole process can be divided into three parts: the construction of radar transmitter,the research and extraction of electromagnetic signal gene characteristics,and the identification of radar emitter.The main work of this thesis is as follows:1.The mechanism of electromagnetic signal generation of the radar emitter was analyzed.The main vibration amplification radar transmitter was taken as the research object,and the influence of internal module parameters on the signal was studied.First,theoretically analyzed what happens to the signal when the signal passed through the module,followed by the analysis on the influence of the module on the signal according to the actual chip parameters provided by the company,and used the S-parameter model to construct a complete main vibration radar transmitter.2.The model of electromagnetic signal gene feature extraction was studied.First,the characterization model of the biological gene was studied.Combined with the deterministic model of biological gene characterization,it is clear that after the model being built,there would be no randomness in the model,and the characterization model of electromagnetic signals could be constructed accordingly.The 20 components' parameters of which were analyzed,and the five most obvious parameters affecting the signal were selected as the gene features.The genetic characterization model of the electromagnetic signal was derived based on the five genetic characteristics and the transmission coefficient matrix and coefficient matrix of the characterization model is derived by approaching the signal continuously via the Weierstrass approximation theorem,which led to the solution model of the electromagnetic signal gene characteristic.Through a large number of simulation experiments,the experimental results show that there are genetic characteristics inside the radar transmitter,and the genetic characteristics have a unique influence on the signal.3.Using the aforementioned radar signal gene feature model,the gene features of the four different signals generated by the radar emitter are extracted.Using VGG stacking and ResNet's Inception idea,the VGG-ResNet network is constructed.The degree of aggregation of the network is 32,and the network can improve the accuracy without increasing the parameter complexity,and also reduce the number of hyper parameters.Based on the extracted gene characteristics,a VGG-ResNet-based method for individual identification of emitter is proposed.The method can improve the accuracy of the deep learning network without increasing the complexity of the parameters.The experiment results show that all the three deep learning networks can get good identification performance,and the performance of VGG-ResNet is the best.In addition,using the VGG-ResNet,the proposed feature enhancement method can improve the recognition performance 2 dB on average compared with the method without feature enhancement. |
PDF Full Text Request