Methods And Applications Of Random Modulated Radar Signal Processing Based On Compressed Sensing

In traditional coherent pulse-train radar, there are mainly two key problems such asambiguity and coupling, both of which could be resolved by transmitting the randommodulated signal. However, due to the randomness, the high sidelobe pedestal incorrelation processing will decrease the performance of target detecting and imaging,which prevents the development of random modulated signal radar. Recently, theappeared compressed sensing theory expatiates on the relationship between randommeasuring and signal sparsity in signal reconstruction, which would provide new ideafor random modulated radar signal processing with low sidelobe. Based on combinationof the compressed sensing theory and the random modulated signal radar, this thesismainly analyzes some new parameter design criterions and corresponding algorithmsfor target detection and imaging by exploiting the sparsity of target echo signal in thetime, frequency or time-frequency domain. Especially, with respect to the parameterestimation of random modulated signal based on compressed sensing, restrictedisometry property (RIP) validation, precise estimation of continuous parameter, fastsequential processing and time-frequency analysis are investigated, which can be furtherapplied to signal processing of the random modulated radar.The first chapter reveals the relationship between compressed sensing theory andrandom modulated radar, and then elaborates the history of random radar signal andcompressed sensing radar. The scientific and technical problems in random modulatedradar signal processing based on compressed sensing are reviewed and summarized.The second chapter mainly focuses on the RIP validation of random modulateddictionaries, which can be applied to signal parameter design of random pulse repetitioninterval (RPRI) radar and random frequency hopping (RFH) radar. The target echoesfrom RPRI radar and RFH radar are analyzed, based on which the parameter estimationmodel of random modulated signal is abstracted. Then the statistical Cramer-Rao lowerbounds (CRLB) of the random modulated signal parameters are deduced as well as theperformance of maximum likelihood estimation (MLE). After verifying the RIP ofequivalent random modulated dictionaries, parameter estimation of random modulatedsignal based on compressed sensing is detailed and some typical fast sparse recoveryalgorithms are compared, which provide some foundations for the subsequent signaldesign and processing.The third chapter mainly focuses on the precise estimation of continuous parameter,which can be applied to precise velocity measurement and moving target detection inRPRI radar. Firstly the grid mismatch effect in parameter estimation based ontraditional compressed sensing is analyzed, and the iterative grid optimization (IGO)algorithm is developed to solve this problem by referring to the mutual iterative principle. This algorithm could obtain optimal dictionaries as well as consecutiveparameter estimations, the mean square error of which can approach CRLB. Byapplying the IGO algorithm to moving target detection in RPRI radar, we can achievealiasing-free precise velocity measurement and improve the detection performance.Furthermore, for moving target detection in airborne cases with strong ground clutter,we present an adaptive suppression method by using prior knowledge of clutterboundary in Doppler spectrum. The prewhitening filter can be effectively obtained tocancel the mainlobe clutter and the clutter residue can be suppressed by the iterativereweighted l1minimization to enhance the target response, which finally offers a highperformance of output signal to clutter and noise ratio as well as target parameterestimation.The fourth chapter mainly focuses on fast sequential processing of sparse recovery,which can be applied to one-dimensional or two-dimensional target dynamic imaging inRFH radar. An adaptive algorithm for parameter estimation in practical cases withunknown sparsity is proposed based on the sequential compressed sensing (SCS), withwhich we can seek the optimal measurement number required by defining the properstopping rules. Under this framework, the range and cross-range compressions ofdecoupled inverse synthetic aperture radar (ISAR) imaging in RFH radar are researched.For the range compression, the high resolution range profile (HRRP) is generated by thefast sequential homotopy, and precise motion compensation is achieved by theminimum l1-norm criterion. For the cross-range compression, the sequential smoothed l0(SL0) algorithm in the matrix form is proposed. Both of the algorithms can largelyreduce the measuring data and coherent processing interval, and improve theperformance and efficiency of compressed sensing imaging.The fifth chapter mainly focuses on the time-frequency analysis of random sampledsignal, which can be applied to micro-motion target analysis and imaging in RPRI radar.An aliasing-free time-frequency analysis approach named short-time compressedsensing (STCS) is presented for random sampled signal, in which the widths of theparticular rectangle windows are determined adaptive to the data by defining a propersimilarity rule between two sequential spectra. In order to speed up the STCS procedure,the SL0algorithm is chosen for sparse recovery, where the pseudo-inverse ofdictionaries can be calculated iteratively. In practice, the STCS algorithm can be wellapplied to RPRI radar for micro-Doppler analysis and the aliasing-free spectrogram ofthe target echo signal can be obtained as well as the target micro-Doppler parameters.Furthermore, when the narrowband RPRI radar is used for fast rotational target imaging,the novel scale-down-dictionary CS (SDD-CS) processing schemes are proposed, wherethe modified generalized Radon transform (GRT) is applied on the time-frequencydomain to generate the low resolution images and then the dictionaries are scaled downby random undersampling as well as reserving the atoms corresponding to those strong scattering areas. This algorithm can achieve preferable images with no aliasing as wellas acceptable computational cost.The sixth chapter makes a summery of the research studies and main contributions inthis thesis. The perspectives of compressed sensing theory and random modulated radarare provided and some open problems are also presented.In conclusion, this thesis is of both theoretical and applicational significances. On theone hand, the proposed parameter estimation algorithms for random modulated signalenrich and extend the theory of semi-parametric spectral estimation. On the other hand,the proposed target detection and imaging techniques provides some fundamentalsupports for producing new radar frameworks or improving existed radars. All theresults in this thesis can establish good groundwork for successive researches onapplying compressed sensing to complex random modulated radars, which plays animportant role on the developing of radar signal processing technologies.