Design of robust projection matrix using prior information for signal compression

This paper deals with designing optimal projection matrix for signal compression. With prior information taken into account, a novel measure is proposed, which allows us to design optimal robust projection matrix such that the sparse representation error is minimized and the clean signal is sensed as much as possible. An analytical solution is derived for the optimal projection matrix. Simulation and experiments, carried out using synthetic data and real image signals, show that the CS system with the proposed projection matrix outperforms the prevailing ones in terms of reducing the effect of sparse representation errors.