Hard Thresholding Pursuit with Partially Known Support for Compressed Sensing

Abstract

Compressed sensing has attracted lots of interest in recent years. Recent works in modified compressed sensing exploited the prior information about the signal to reduce the number of measurements. In this paper, we propose a hard thresholding pursuit algorithm with partially known support (HTP-PKS), which incorporates the prior support information into the recovery process. Theoretical analysis shows that by using prior information of partially known support, the HTP-PKS algorithm presents stable and robust recovery performance under a relaxed restricted isometry property (RIP) condition. To illustrate, simulation experiments are given.