Research on matching pursuit sparse decomposition based on iterative residual and extension algorithm

摘要

Matching pursuit (MP) is usually used for the signal sparse decomposition. The criteria for selecting optimal core function in typical MP decomposition is that the primitive has the largest inner product with the core function. However, this criteria can lead to large error on reconstructed signal. Contrary to this problem, a criteria is proposed, which has less error of reconstructed signal than typical MP decomposition. In order to solve this problem, an extension algorithm is proposed. Results show that this algorithm is more effective than iterative residual algorithm in weakening endpoint effect.