In-pulse Characteristic Parameters Extraction Method Of Radar Signal In Complex Environment

Intra-pulse features analysis based recognition of radar emitter signal is a key link of theelectronic reconnaissance technology area, its target is to make feature analysis and judgmentof the pulse signals belong to different radar emitters. With the development of radartechnology, new system radar is emerging continuously, modulation modes of radar signalsare more flexible, signal parameters are more variable, all these result in the more complexelectromagnetic environment of the war. Identify the type of radar signal safely, effectively,fast, accurately, and make an reasonable judgment, especially in a complex situation whereexist dense electromagnetic signals, will be essential to the electronic warfare, evendetermines the results of the future wars. Therefore, this paper extracts the feature parametersof radar emitters of complex electromagnetic environment, analyses its performance, andmakes a deeply discussion and research on the theories which relate to complex system radaremitter signal recognition. The main work and results are as follows:A key point of radar emitter recognition is to extract the feature parameters of the radarsignals. At the present time, there has been many mature achievements both at home andabroad, which mainly include: instant autocorrelation method, wavelet transform method,fuzzy function ridge feature method, wavelet packet, entropy feature method and so on. Thesemethods have effectively improved the recognition quality from different angles and differentaspects. Especially in the signal-to-noise ratio SNR5dB, recognition accuracy reachedabove99%. But when the SNR is less than this value, the recognition accuracy of most ofthese methods will serious decline. Aiming at this problem, fuzzy function ridge featuremethod using fractional order, extending the recognition domain to fractional domain, andextracting the moment features of the main ridge section as recognition parameters toeffectively overcome the noise impact. This method can steadily reached the recognition rateabove80%when SNR2dB, but when SNR is less than2dB, its recognition effect willappear serious error. Aiming at such problem, based on the extraction of fuzzy function ridgefeature, this subject proposes the extraction method of three feature parameters: fuzzyfunction main ridge section resemblance coefficient, Holder coefficient, and cloud modelsimilarity. Considering the problem that when SNR is low, the shape of main ridge section is liable to be influenced by noise, this paper employs the singular value decomposition (SVD)and empirical mode decomposition (EMD) two denoising method to depress the noiseinterference of the section envelope curve. The simulation analysis shows that, after denoisingprocess, the main ridge section can keep good whole shape. In recognition of several signals,compare with traditional moment feature parameters, these three feature parameters havesimilar inter-class separation degree and intra-class aggregation degree, and relatively stable,so they could be effective feature parameters of signal. According to such results, based onridge feature recognition, this paper proposes fuzzy function main ridge section-basedmultiple-dimensional feature space data fusion recognition method. Takes the new featureparameters of main ridge section and associate them to traditional ones, formmultiple-dimensional feature recognition vector, and employs fuzzy kernel C-meansclustering algorithm on both traditional feature recognition vector and the proposedrecognition method. The experimental results show that, compare with traditional fuzzy mainridge feature extraction method, the proposed method has an improvement on2dBrecognition accuracy, even when the SNR is0dB, it still has an recognition accuracy whichover90%.