Research On Adaptive And Robust Algorithm For CS-Based Wideband Radar Reconnaissance Receiver

With the rapid development of radar technology,large frequency coverage,large instantaneous bandwidth,large dynamic range and high sensitivity multi-signal processing capability have become an inevitable trend in the development of radar reconnaissance.Limited by the inherent analog bandwidth limitations,jitter of sampling pulses,and aperture effects,high-speed sampling by ADC has become a major bottleneck in the performance improvement of wideband digital receivers.In order to break through this bottleneck,compressed sensing(CS)has been applied to the next generation radar reconnaissance receiver architecture.CS can greatly reduce the sampling rate,cleverly avoiding the high-speed sampling bottleneck,ensuring large receiving bandwidth while taking into account large dynamic range,and reducing the rate of signal processing.At the same time,the CS framework can be deeply integrated with functions such as parameter estimation,signal detection and recognition,which lays a foundation for cognitive and intelligentization in the new generation of radar reconnaissance receivers.It has great practical significance and development potential.However,many studies on CS-based wideband radar reconnaissance receivers show two common problems:the one is the low robustness of the CS algorithm,which is caused by the sensitivity to the noise of the CS framework;The other is the inefficient adaptive reception of the unknown signals,which is caused by the specific application requirements of radar reconnaissance.In this thesis,the robustness and adaptive problems in the CS algorithm are studied and optimized in three aspects:the signal reconstruction,the compression observation and the sparse expression.The key research and contributions of this thesis are as follows:First,in order to make full use of the structural information of radar signal,a block sparsity model was introduced.In order to achieve complete adaptive reconstruction in the case of unknown block distribution and block sparsity,the binary tree search and supervision mechanism is proposed by referring to the binary search tree method in the computer field.Based on this mechanism,two complete adaptive block greedy algorithms,BTSM-B2OMP and BTSM-AB2MP are proposed.The latter is added a backtracking mechanism and therefore has stronger error correction capability and robustness than the former;The simulation verifies the adaptive reconstruction ability of the proposed algorithms in the ideal case and the CS receiver,respectively;The computational complexity of both algorithms is given,while the optimal reconstruction characteristics of the binary tree search and supervision mechanism are proposed.Second,in order to solve the problem of SNR degradation of the reconstructed signal caused by noise,a channelized CS model is proposed.Based on this model,the C-OMP algorithm is proposed,and the specific steps and the theoretical description are given.In order to obtain a considerable improvement in SNR,the algorithm firstly reduces the noise in the non-signal channels by channel screening,and then further suppresses the noise component in the signal channels by using the proposed residual attenuation slope discriminating mechanism;The theoretical analysis of reconstruction conditions,channel number,stop threshold,reconstruction sensitivity,computational complexity,and output SNR improvement are given in detail through mathematical derivation,which proves the performance advantages of C-OMP.Finally,the robustness,reconstruction conditions and advantages in the CS receiver of the proposed algorithm are verified by simulation.Thirdly,in order to solve the problem of high computational complexity and inconvenient parameter setting in C-OMP,an improved CD-gOMP algorithm is proposed combined with channelized CS model,sparsity pre-estimation,and gOMP.CD-gOMP performs two-step iteration in C-OMP with the generalized matching mode,reduceing the interference of the noisy channel through channel screening first,and then using the sparsity pre-estimation method and the backtracking mechanism to approximate the original support set in signal channels.The detailed steps and theoretical description are given;Also,the reconstruction conditions,computational complexity,theoretical SNR improvement of the algorithm are theoretically analyzed,respectively;Moreover,the mathematical derivation of the reconstruction error in the noise background prove the denoising performance of the CD-gOMP;Finally,the simulation verifies the advantages of the proposed scheme in adaptive reconstruction and in robustness under noise background.Fourthly,in order to comprehensively improve the robustness and adaptability of CS radar reconnaissance receiving,an adaptive robust cognitive analog-to-information converter(AIC)is proposed based on coupled observation-dictionary optimization.First,the compressed observation is robustly optimized based on the RIP rule,which improves the effectiveness of the observation under noise.Then,in order to solve the problem of the adaptive sparse expression,the dictionary learning method is introduced.The compressed samples are fully utilized,also the learning process is optimized for robustness,adaptability,and online form;By combining the compression observation optimization and the CS online robust dictionary learning,an observation-dictionary coupling learning algorithm(C2ORL)and a C2ORL-based adaptive robust cognitive AIC model are proposed.Finally,the effectiveness of the optimization is proved,also the performance of adaptive robust reception under noise background is verified by simulation experiments.