Research On Signal Feature Extraction For Passive Radiation Sources Detection On Satellite

In modern warfare, electronic surveillance is a key electronic warfare technologies.By analyzing the intercepted target radio signals, passive radar extract the valid targetparameters such as carrier information and the modulation information, then find out thekey features of the target to identify and locate it.This thesis mainly study aboutspaceborne passive radar related questions, such as the channel effect, the intra-pulsefeature extraction, unintentional modulation feature extraction and signal sorting.First, how the channel impact signal were analyzed as a reference to designsimulation environment and set parameters. According to the research of the specificneeds of this study, a specific simulation signal source was designed so that thesimulation signal source may accurately reflect the fine features, as the basis for futureresearch.Then, how to apply fractional Fourier transform on extracting radar pulseCharacteristics was studied. The anti-noise performence of the method under simplechannels as well as feature extraction performence under satellite-ground channels weretested through simulation experiments. The simulation experiments shows thatfractional Fourier transform method has strong anti-noise performance, and one of theextracted pulse characteristics, the modulation slope of the linear frequency modulationsignal, is still very accurate and stable under satellite-ground channels.Next, how to apply higher order statistics on extracting signal unintentionalmodulation characteristics, namely, fine features were studied. Respectively, simulationwere carried out to see how it performed when signal features were based on the thirdorder cumulant and slice of fourth order cumulant characteristics or based onbispectrum and trispectrum features, or based on bispectrum and spectrum of fourthorder cumulant slice features,under simple channels as well as satellite-ground channels.Experimental results show that higher-order statistics has great application prospects,and the extracted features achieved better classification results.Finally, some signal sorting methods were introduced, and PRI transformationbased signal sorting was studied. Through simulation compared traditional andimproved PRI transformation based sorting method in anti-jitter performance. On thisbasis, with the previously extracted pulse characteristics, pulse similarities andimproved PRI transformation were combined for signal sorting.