چكيده به لاتين
Abstract:
The main goal of this thesis is to improve the target detection and parameter estimation using sparse signal models. Due to the high dimension of sparse signal models in pulse Doppler MIMO radars, proposing some solutions for decreasing the computational complexities of sparse processing is the other goal of this thesis. In so doing, the two dimensional (2D) sparse model is proposed for pulse Doppler MIMO radars in which the computational complexity is much less than that of the one dimensional sparse signal model. To solve the proposed 2D sparse signal model directly, four 2D algorithms including 2D-IAA, 2D-SLIM, 2D-SBL-LP, and 2D-TNIPM are proposed based on different criteria and norms. Accordingly, the behavior of each proposed algorithm is different from the others, and some ones have better performance in special situations such as low SNRs, high sparsity levels, the low number of measurements, and so on.
In addition, a method is proposed to cope with the destructive effect of jammers on the 2D proposed sparse recovery algorithms for MIMO radars. Also, to improve the target detection for sub-Nyquist sampling rates, a measurement matrix design is proposed for 2D compressed sensing based on minimization of mutual coherency of the columns of the sensing matrix.
To evaluate the computational complexity of different algorithms, the number of flops needed for execution of each algorithm and also the running time of different algorithms are compared with each other. Also, the target detection and parameter estimation of different algorithms are compared with each other. The simulation results show that the proposed 2D-SBL-LP algorithm compered to well-known BCS-LP not only has a better performance in sparse recovery, but also has a less computational cost. Furthermore, the proposed 2D-SLIM, 2D-IAA, and 2D-TNIPM algorithms compared to the 1D ones reduce the computational burden drastically while both related 1D and 2D algorithms achieve the same performance.
Keywords: Pulse Doppler MIMO radar, sparse learning via iterative minimization, Two-Dimensional sparse signal model, truncated Newton interior point method, sparse signal l1-norm based recovery, Sparse Bayesian learning, Bayesian compressive sensing.
