Research On Emitter Fingerprint Identification And Fine Feature Extraction

Electronic support within the context of electronic warfare is concerned with how tocommand the enemy’s radiation characteristics and judge the threat level by detection,interception, location, analysis and identification of radiated electromagnetic energy. Themodulation type in the radar emission system including IMOP (Intentional Modulation onPulse) and UMOP (Unintentional Modulation on Pulse).Aiming at identifying radar signals foremitters with the same type and the same IMOP, fingerprint features for radar emitter signalswere studied. The main study content included four aspects.The fingerprint features generated mechanism for radar emitters were analyzed in thispaper. According to the definition of phase noise, this paper proposed a method for phase noiseexpress in the base band, and single side band power spectral density paten was built, whichmade it available to getting the sample data for radar emitter fingerprint identification system.The in-time features and the in-frequency features were studied for UMOP signals. Due tothe traditional Fast Fourier Transform Algorithm (FFT) had some drawbacks in the frequencyestimation field，this paper proposed an amplitude comparing algorithm based on FFT，whichimproved the estimation precision and made it possible to take the estimated frequency as oneof the feature vectors in radar emitter fingerprint recognition system.For the problem that it’s hard to select the proper feature vectors in radar emitterfingerprint system，this paper studied with the in-time signal and the in-frequency signal, andfigured out several radar emitter signal parameters which were able to represent the naturalsubtle characters, and made comparison among these feature parameters.Facing the problem that the feature vectors are too limited for sampled signals，accuraterecognition for fingerprint signals with the same type and the same IMOP was realized by usingfuzzy clustering means (FCM) and support vector machine (SVM) in this paper. Theexperimental results showed that combined with IMOP, the method for emitter fingerprintfeature extraction based on phase noise feature parameters and classifier design based on FCMand SVM have preferable accuracy and reliability performance.