Research On 1-Bit Compressed Sensing Reconstruction Algorithm And Unified Generalized Linear Model

Compressed sensing is an emerging theory of signal processing.Compressed sensing can realize the synchronization of signal compression and sampling,while 1-Bit compressive sensing quantizes the measured values,effectively reducing the complexity and improvement of communication equipment.The working efficiency of the signal processing system also increases the requirements of the reconstruction algorithm on the reconstruction end.This paper studies the reconstruction algorithm under the 1-Bit compressive sensing model.The specific research contents are as follows:Firstly,in order to improve the reconstruction performance and robustness of the pinball iterative hard threshold algorithm,a pinball iterative hard threshold algorithm based on Heavy-ball is proposed.Then the adaptive outlier pursuit method is introduced in the above algorithm.The adaptive outlier pursuit Heavy-ball pinball iterative hard threshold algorithm is proposed.The algorithm can accurately reconstruct the original signal when the symbol is inverted.Secondly,in order to accelerate the convergence of the generalized sparse Bayesian learning algorithm,the step of damping operation is added in the iterative process of the algorithm to improve the convergence speed of the generalized sparse Bayesian learning algorithm.For further improving generalized sparse Bayesian learning algorithm under the 1-Bit compressive sensing framework for the reconstruction effect of block sparse signal,the method of pattern coupled is introduced,and the generalized pattern coupled sparse Bayesian learning algorithm is proposed.The algorithm has better reconstruction effect on block sparse signal.Finally,aiming at the problem that the quantization threshold is fixed to zero for the approximate message passing algorithm under the unified generalized linear model of the 1-Bit compressed sensing framework,an adaptive threshold unified generalized approximate message passing algorithm is proposed.The algorithm replaces the original zero threshold with an adaptive threshold to improve the reconstruction performance.Then,based on the advantages of the algorithm,the nearest neighbor sparsity pattern learning method is added to learn the sparse structure of the original signal,and an adaptive threshold unified generalized nearest neighbor sparsity pattern learning approximate message passing algorithm is proposed.The algorithm can accurately reconstruct the block sparse signal.