Research On Multi Dimension Feature Extraction And Recognition Technology Of Radar Active Deception Jamming

The radar is regarded as the eyes of modern warfare. Its performance directly affects the consequence of the war. As the frequency spectrum becomes increasingly complicated, the emergence of alternative jamming seriously influences the performance of radar and poses a severe challenge to the living environment of radar. This is particularly serious in radar active deception jamming. Thus currently the radar anti-jamming technology to active deception jamming has been the focus of military study. The basic procedure of radar anti-jamming is to acquire the type of active jamming based on the extraction, classification and identification of feature parameters of interference signal, and adopt corresponding measures to suppress the jamming. This thesis focuses on the feature extraction and identification of radar active deception jamming in multi-dimensional transform domain and the main work is summarized as follows:1. The produce of common active deception jamming is analyzed and the simulation analysis is carried out in accordance with the working principles of radar, which makes further preparations for the feature extraction of interference signal.2. The differences of seven active deception jamming signals are analyzed in multi-dimensional transform domain such as frequency domain, time domain, image domain, etc. Based on these characteristics of interference signal, a series of feature parameters featured by obvious discrimination and low noise sensitivity are extracted by time-frequency transform, wavelet transform, waveform analysis, etc. the distribution characteristics of every feature parameter are summarized by simulating in different JNR.3. Relying on multi-dimensional feature parameters, backward cloud model, tree discriminator and Artificial Neutral Network(ANN) are applied to effectively identify the interference signal. The simulation result is also given to analyses merits and demerits of different identification algorithms. Specifically, under low JNR condition, tree discriminator has low identification rate; backward cloud model has a low identification rate with certain interference because the method of computing the least distance is employed; ANN has a good performance but needs abundant prior information. On the basis of different characteristics of identification algorithm and feature evaluation metrics like F parameters, inner and outer distance, specificity parameters, etc. feature selection is carried out to optimize identification rate of interference signal and a good result is obtained.4. After the jamming type is identified adequately, the RVGPO is taken as a study object of the anti-jamming. Firstly the differences between the real target’s echo signal and interference signal are analyzed in time-frequency domain. Secondly the mode decomposition method is used to decompose and reconstitute the interference signal. On this basis, the methods of waveform transform and PCA are resorted respectively to improve the anti-jamming performance of RVGPO. The analysis of simulating result is offered to demonstrate better anti-jamming effects.At last, the research contents and main work of the thesis are summarized, the problems existing in the research and the direction of the work are pointed out.