New Bernoulli And Gaussian Sensing Matrices For Cluster Structured Sparse Signals

Cluster structured compressive sensing (CSCS) is a new methodology to reconstruct cluster structured sparse (CSS) signals from a small number of measurements. As in conventional compressive sensing, the useful information is captured from the original signal using a sensing matrix that impacts the accuracy of signal recovery. In this paper, we propose two new sensing matrices for cluster structured compressive sensing. Our idea is to improve the orthogonality property of Bernoulli and Gaussian sensing matrices. Simulation results show that the proposed matrices outperform the existing ones.