Study On Time-Frequency Atoms Features For Advanced Radar Emitter Signals

The feature extraction of radar emitter signals is a critical process for ELectronic INTelligence (ELINT), Electronic Support Measures (ESM) and Radar Warning Receiver (RWR) systems. As the precondition and foundation of deinterleaving and recognizing radar emitter signals, the feature extraction technique would determine the performance of the electronic reconnaissance equipment directly and influence the war strategy subsequently. Along with the countermeasure activities in modern electronic warfare are becoming more and more drastic, the advanced modern radars with complicated systems are playing a major role in the field. Due to the high-density, complex and variable electromagnetism signals environment which destroyed the signal original rules, the normal five-parameter feature extraction method, radio frequency (RF), time of arrival (TOA), pulse width (PW), pulse amplitude (PA) and direction of arrival (DOA) cannot meet the requirements of modern electronic warfare. An advanced method of feature extraction of radar emitter signals, especially for which with complex systems is a new challenge for electronic warfare. The low level of theoretic research about radar emitter signal in our country has seriously restricted the further development of modern electronic equipment. The only way to improve present technique is to enhance nature feature study of radar emitter signals, investigate new valid feature parameters.Aiming at the key theoretical issues in signal processing of electronic warfare, A novel feature analysis method based on the time-frequency atom for radar emitter signals is presented in this dissertation. Theoretical fruits are as follows:1 A feature extraction method of radar emitter signals based on Gabor atom is proposed. Firstly, an over-completed Gabor atom dictionary suitable for decomposing radar emitter signals is built and an improved quantum genetic algorithm (IQGA) is introduced to effectively reduce the time complexity at each search step of matching pursuit (MP), and thus the radar emitter signals are decomposed into a linear expansion of optimal Gabor atoms representing the signal feature. Then, the dissertation studies the Gabor atom feature parameter difference among radar emitter signals with various modulation types, and analyses the Gabor atom feature performance by comparing the difference. The experiment result shows that even a small number Gabor atoms can express the major feature of original signals. The extracted Gabor atom feature not only has good noise-suppression ability, but also has the non-cross interference time-frequency distribution.2 Based on the over-completed Chirplet atom dictionary with superior gathering ability in time-frequency domains, a feature extraction method for radar emitter signal is presented. At the same time, a novel DNA evolution algorithm (NDEA) is proposed. Comparison of DNEA with other algorithms for typical complex functions demonstrates the algorithm has good characteristics of of rapid convergence, short computing times and strong search capability. Then, the fast MP algorithm based on NDEA is applied to effectively reduce the complexity of searching calculations, and thus some best-matched Chirplet atoms representing features of typical radar emitter signals are obtained. Furthermore, the dissertation compares and analyses the performance difference between Gabor and Chirplet atom dictionaries. The experiment results show that the smaller number Chirplet atom can represent more accurate feature of radar emitter signals compared to Gabor atom.3 In order to effectively measure the spectrum difference among radar emitter signals, a feature extraction method based on Spectrum atom is presented. At the beginning, the advantages and disadvantages of current main spectrum feature extraction methods are concluded. Based on it, the Spectrum atom expressing the local spectrum structure of signals is designed and a method of multi-scale Spectrum atom dictionary is proposed. Meantime, the fast FFT algorithm is applied to extract the best-matched Spectrum atom for all kinds of traditional radar emitter signals. The experiment results show that the extracted Spectrum atom parameters have a certain physical meaning which not only can express the spectrum difference among various modulation types, but also can recognize the radar emitter signals with same modulation type and different modulation parameters.4 The time-frequency atom feature performance of radar emitter signals on the noise condition are studied further. Firstly, the Chirplet atom derived feature extraction algorithm is presented and a signal classification method based on hierarchy strategy is designed to analyse the classification and noise-suppression ability of Chirplet atom. Then, the method to achieve the Spectrum atom characteristics vector according to the parameters of the atoms is proposed. Finally, the hierarchy strategy and the kernelized clustering algorithm are applied to realize the signal automatic classification and parameter estimate. The experiment results show that the extracted atom feature has good property of clustering the same and separating different radar emitter signals, and has good noise-suppression ability, which further confirms that the time-frequency atom analysis method is effective and feasible for feature extraction of radar emitter signals.This work is supported by the National Natural Science Foundation of China (No. 60572143) and the National Electronic Warfare Laboratory Foundation (No. 51435QT220401).