| Compressed Sensing(CS)is a new signal processing theory proposed recently years,which synchronizes the process of data collection and compression.It breaks the sampling rate limit of the Shannon sampling theorem,which saves a lot of data storage,transmission and computing resources,reduces the hardware performance requirements.In the field of modern information processing,Compressed Sensing has a prominent advantage,gets extensive attention and has broad application prospects.Reconstruction algorithm as an important part of compression sensing,which determines the performance of compression sensing.As to sparse signals,if the reconstruction algorithm can not use less observations to recover the signals effectively,the CS theory can not show its superiority to the traditional sampling method.The better reconstruction algorithm improves the research significance of CS theory,and improves its practical application value.Therefore,in this paper,the following work is done for the compressive sensing and reconstruction algorithm:1.Firstly,we introduce the framework of compressive sensing,sparse representation,design of observation matrix and reconstruction algorithm.2.The algorithms of OMP,ROMP,StOMP,CoSaMP and SAMP are compared respectively.The performance of the algorithms is analyzed,and the corresponding defects are pointed out and the improvement ideas are put forward.3.An improved compressed sampling matching pursuit algorithm based on dual threshold is proposed.The algorithm reduces the reconfiguration time of the CoSaMP algorithm and improves the accuracy of the final atomic set by using the double threshold strategy.4.A modified sparse adaptive matching pursuit algorithm(MSAMP)is proposed.The algorithm estimates the sparsity firstly,then improves the accuracy of candidate atoms by soft threshold,and optimizes the iteration stopping condition finally.Compared with the original algorithm,the reconstruction time and the reconstruction quality are both improved. |
PDF Full Text Request