Sparse Decomposition Algorithm Using Immune Matching Pursuit

The optimal sparse expression for a signal is critical to the application of compressed sensing theory. Since the standard matching pursuit uses greedy strategy to approximate the objective function, which is aim to search for the linear combination of the optimal atoms in a redundant dictionary. It is inevitable to cause enormous calculation quantity, to overcome this defect, a scheme based on immune matching pursuit algorithm is proposed in this paper which is suitable for speech signal sparse decomposition. In this matching pursuit algorithm, each matching process has been optimized by the immune algorithm with the characteristics of local fast convergence and global optimal. As a result, according to this algorithm, the speed of sparse decomposition and the quality of reconstructed signal all have been effectively improved, and the storage space requirement for redundant dictionary also has been degraded. Simulation results show that, the matching pursuit computational cost can be obviously reduced by using this algorithm, Moreover, the performance of this algorithm is stable, and it also has high signal decomposition accuracy. In a word, the efficiency of sparse decomposition can be significantly improved by combining the immune evolution mechanism and matching pursuit algorithm.