Study On The Interception Of Low Probability Of Intercept Radar Signals

To reduce the probability of intercept of the reconnaissance receiver, radardesigners try their best to spread the energy of radar signals to multidimensionaldomains, including time, frequency and space domains, etc., making radar signals havelow probability of intercept (LPI). The reconnaissance receiver, which is designed forthe LPI radar signals, should have a broad instantaneous monitoring bandwidth to keepa high probability of intercept in the frequency domain, and be able to deal with the LPIradar signals with complex signal forms. First, when the instantaneous monitoringbandwidth reaches up to tens of GHz, we design a new reconnaissance receiver toreceive the LPI radar signals with a high probability of intercept in the frequencydomain using a low sampling rate analog-to-digital converter. Second, under thecondition of a low or even negative signal-to-noise ratio (SNR), this dissertation focuseson the intra-pulse modulation recognition of some conventional radar signals andclassical LPI radar signals based on the instantaneous and non-instantaneous signalfeatures. The main contributions of this dissertation are as follows:1) When the monitoring bandwidth is up to tens of GHz, first, we propose aNyquist folding receiver (NYFR) based on the look-up table of the time of zero crossingrising to realize the local oscillator signal (LOS) synchronization, solving the problemof the synchronization difficulties when estimating the Nyquist zone (NZ) of the outputsignal of the NYFR. Second, because the nature of the NYFR is that it can modulatedifferent information in different NZs, we present the synchronous NYFR (SNYFR).2) When the input signals are the monopulse (MP), linear frequency modulation(LFM) and binary phase shift keying (BPSK) signals, the corresponding parameterestimation algorithms are proposed. For the MP signal case, we first transform theestimation of the NZ into the detection of the LFM signals with a fixed slope and anunknown carrier frequency when the LOS is a periodic LFM signal, and then thedemodulation is performed according to the NZ by utilizing a fast frequency estimationalgorithm. For the LFM signal case, we synchronize the LOS and transform the synchronized signal into the MP signal for detection of the NZ. Moreover, we take intoaccount the existence of the NZ of the output signal of the SNYFR and present theparameter estimation of the LFM signal intercepted by a left sideband SNYFR. For theBPSK signal case, we compute the square of the received signal and transform theparameter estimation of the BPSK signal into the case of the MP signal. In addition, wealso study the parameter estimation of multi-component input signals.3) We use the phase difference, the generalised time-frequency representation ofZhao-Atlas-Marks (ZAM-GTFR) and the singular value decomposition (SVD) of therotated ZAM-GTFR to realize the intra-pulse modulation recognition of several kinds ofsignals. First, the presented phase difference algorithm takes into account the effect ofunwrapping frequency ambiguity caused by the SNR threshold on the recognition. Theprobability of successful recognition can reach90%when the SNR is above4dB.Second, we derive the ZAM-GTFRs of different kinds of signals, and find that there is asharp negative peak in the time-frequency plane with respect to the phase change of thecoded signal. This algorithm works well when the SNR is above-2dB. Finally, toreduce the SNR requirements, we rotate the ZAM-GTFR time frequency image andperform the SVD, and the different meanings of different singular values are used toachieve the signal recognition when the SNR is above-6dB for most signals.4) The ambiguity function (AF) and the Rihaczek distribution (RD) are able torepresent the characteristics of the signal. We use the AF and RD to extract a number ofintra-pulse features of the received signal. Two one-dimension functions are defined toreduce the computational complexity of the AF, and some non-instantaneous featuresare derived from these two functions. The algorithm using the AF works well when theSNR is above-1dB. The algorithm using the RD takes advantage of the cross terms ofthe RD and the excellent line detection ability of the Hough transform, and extracts twonew features to realize the recognition when the SNR is above-4dB.