Main Article Content
Compressed sensing theory breaks the traditional sampling limit. It projects the high-dimensional signal into a low-dimensional space to get a small number of measured values through the observation matrix, and then uses the reconstruction algorithm to get the original signal with high probability. In order to solve the problem of unknown signal sparsity in practical applications, it is proposed an improved regularized adaptive matching pursuit algorithm based on LDPC measurement matrix. In the case of unknown sparsity, it is used the LDPC matrix with quasi-cyclic characteristic for observation in the improved algorithm, which sets adaptive threshold automatically to adjust the number of atoms of candidates, and passes back to eliminate error atoms. At the same time, the LDPC matrix corresponding to the new atomic number is updated to improve the accuracy of reconstruction. The experimental results show that the step size can gradually approach the value of sparsity, so as to reconstruct the original signal accurately under the premise of unknown sparsity and the same test conditions. Thus it can ensure the global optimization and reduce the reconstruction time. In addition, because the selected LDPC observation matrix is quasi-cyclic, the storage space of the observation matrix can be saved, which is beneficial to hardware implementation, and these provide a better implementation method for the practical application of compressed sensing theory.