Mixed Signals Separation And Parameter Estimation Based On The Joint Distribution

Synthetic aperture radar(SAR) technology plays an important role in strategic reconnaissance and surveillance because of its advantages of all-time, all-weather and long range work. Combined ground moving targets indication(GMTI) and SAR technology, SAR-GMTI can obtain stationary and moving target information, and can form a detailed overall situation map of the battlefield. As SAR-GMTI has been widely used in both military surveillance and civilian exploration, it faces many new problems and challenges, such as complicated interference suppression, maneuvering target parameter estimation, multiple targets imaging, etc. In these challenges, the echoes are usually nonstationary, which cannot be well described via one-dimension(1-D) signal processing methods. In this dissertation, the joint distribution(time-frequency distribution and time-frequency rate distribution) of echoes are developed and successfully applied to suppress interferences, estimate target Doppler parameter and separate multiple targets echoes.The main contents of this dissertation are summarized as follows:1. The first part focuses on the interference suppression algorithm. Narrowband interference(NBI) and wideband interference(WBI) are common jamming signals against SAR, which can degrade the imaging quality severely. Especially WBI, it highly overlaps in the time or the frequency domain, and the interference suppression algorithms cannot be accomplished via 1-D signal processing methods. To solve this problem, an interference suppression algorithm based on Wigner distribution(WD) and sliding window masking technique is proposed. First, a fast signal synthesis algorithm from WD is developed using the inherent relationship between original signal and synthesized signal. Then, the smoothed pseudo Wigner distribution(SPWD) is utilized as a time-frequency mask to suppress the cross-terms. By combination of signal synthesis algorithm and time-frequency sliding window masking technique, the interference can be extracted and synthesized. Finally, the reconstructed interference is removed from the original echoes to suppress the interference. The proposed algorithm can not only suppresses the NBI and WBI, but also preserves the useful information as much as possible. The effectiveness of the proposed algorithm is verified by experiments on both simulation and measured data.2. In the second part, we focus on the Doppler parameters estimation of moving target with constant velocity. After range walk and range curve correction, the moving target can be characterized as a linear modulated signal(LFM). The centroid frequency and chirp rate(second-order phase) of LFM signal corresponds to radial and azimuth velocity, respectively. The motion parameters of moving targets can be obtained by estimating the parameters of LFM signal. Cubic phase function(CPF) based estimator is efficient in estimating the parameters for mono-component LFM signal. However, it suffers from cross-terms and spurious peaks when dealing with multi-component LFM signals. Aiming at this identifiability problem, a coherently integrated CPF(CICPF) algorithm is proposed. This algorithm exploits the property that the auto-terms in time-frequency rate plane are concentrated along straight lines parallel to the time axis. Then, a dechirping technique is utilized to coherently integrate the auto-terms. Finally, the centroid frequency and chirp rate can be estimated simultaneously by amplitude search in the frequency-chirp rate plane. This algorithm is a nonsearching method, and can suppress the cross-terms and spurious peaks efficiently. The simulation and measured data results demonstrate that the CICPF algorithm serves as a good candidate for LFM signals analysis.3. In the third part, the Doppler parameter estimation of maneuvering target is studied. For maneuvering target, the acceleration induces the third-order phase(quadratic chirp rate) in target echoes, which can be modeled as cubic phase signal(CPS). The third-order phase will result in the linear chirp rate migration in the time-chirp rate plane. To solve this problem, an improved axis rotation(IAR) based coherently integrated CPF(CICPF) is proposed. First, the linear chirp rate migration of CPS is corrected by IAR. After that, a dechirping technique with the rotation angle is utilized to enhance the auto-terms. Since the accumulation is coherent, IAR-CICPF algorithm has lower SNR threshold. Furthermore, it can suppress the cross-terms and spurious peaks effectively. Compared with the existing correlation algorithms, IAR-CICPF improves the estimation performance at low SNR, while sacrifices some computational load.4. In the fourth part, we focuses on multiple targets separated imaging method at high SNR. As echoes of targets at each range-frequency slice can be considered as multiple LFM signals, we propose a robust signal synthesis algorithm to separate multi-component signals. This algorithm consists of two stages. First, based on the property that the auto-terms of LFM signals pass through the origin of delay-Doppler plane, a masked ambiguity function algorithm using Radon and its inversed transforms is proposed to suppress the cross-terms. Then, a novel signal synthesis method from ambiguity function is presented based on the matrix rearrangement and eigenvalue decomposition. This algorithm can solve the problems in masked Wigner distribution based signal synthesis method. According to the simulation results, the proposed algorithm can separate multiple targets effectively in the range-frequency and azimuth-time domain at high SNR.