Research On Reconnaissance Of Frequency Hopping Signals

Frequency hopping (FH) communication has been widely applied in military communications, due to its anti-jamming performance, low probability of interception and detection, and networking capability. And it's a great challenge to carry on radio reconnaissance. Now, the research of FH signals reconnaissance, such as signal interception, parameter estimation, and signal sorting is one of the major tasks in communication reconnaissance. A series of key technologies of FH communication reconnaissance in complex electromagnetic environment (CEE), including FH signals detection, parameter estimation, and signal sorting, etc., were mainly researched in this thesis.Firstly, to find the preferable time-frequency representations (TFRs) for FH signals detection, the performance of various TFRs was simulated and quantificationally compared in time-frequency concentration and suppression cross-term interference. Then, the mathematical model of FH signal was build, and the definitions of FH signal's parameters were introduced. The applications of some linear TFRs, bilinear TFRs, reassigned TFRs, and synthesized TFR for FH signals detection were emphatically researched. The representation performance of various TFRs was evaluated quantificationally utilizing information entropy. Moreover, the compute complexities of some typical TFRs were given.Secondly, using single-antenna wideband receiver, a series of new algorithms of blind parameters estimation in CEE based on time-frequency analysis were proposed. In FH signals reconnaissance field, a new quantificational evaluation of electromagnetic environment complexity named'synthesized information entropy', which consists of type-entropy, density-entropy and distribution-entropy, was presented. And then, the denoise processing and signal preselection based on TFRs and channelization limits were accomplished. When there was only one FH signal in the interception band, the blind estimation of hop period, hop timming, and carrier frequency would be gained by spectrum and conventional method. The blind parameters estimation algorithm based on synthesized TFR (SP&SPWVD) when there were multicomponent FH signals was proposed, and the performance estimation was simulated numerically.Thirdly, based on time-frequency analysis, spatial spectrum estimation, integrating digital channelization and time-frequency focusing technologies, the directions of arrival (DOA) of FH or FH/DS signals were estimated accurately in underdetermined case. The mathematical model of spatial time-frequency distribution (TFD) was built based on conventional spatial-time array model and time-frequency analysis. Exploiting STFD to estimate DOAs of multicomponent FH signals can gain SNR enhancement, and the factors that can affect the enhancement value were discussed. Furthermore, an algorithm to estimate DOAs of multicomponent FH signals based on linear STFD was proposed in non-frequency-collisions case. The DOAs estimation can be achieved in underdetermined case, but when N /M is large, the estimation will not be precise because of the interferences among multicomponent signals. This problem is solved solved by using digital channelized receiver. We also proved that the accurate estimation of DOAs of FH/DS signals could be achieved in underdetermined case, and the method was based on STFD and direction finding technologies of wideband signals.At last, the thesis investigted the FH signal sorting technologies. One novel sorting method for non-orthogonal FH signals was presented. In contrast with some conventional techniques, the proposed method has a lower compute complexity, which is convenient for the real-time sorting and appropriate for fast FH signals reconnaissance. And then, an improved K-Means clustering algorithm with optimal initial clustering center and estimation of K value was proposed. Because the initial centers are optimal, the clustering has less iterations, and clustering in the local minimum is also avoided. The improved K-Means clustering algorithm was applied in the clustering of HDW aggregation, which was with lower compute complexity but more robust than conventional K-Means algorithm. To obtain the optimal parameterσof gaussian kernel function, an improved method including coarse searching and precise estimation was presented, which can obtain preciseσopt with less iterations. The optimization of initial clustering centers and the estimation of K value of kernel K-Means (KKM) clustering were achieved by using density distribution, region radius, and adaptive limits. The clustering sorting of FH signals with time-variable hop periods and DOAs was achieved by using gaussian KKM algorithm, and the simulation results demonstrated that the clustering sorting was effective.