High-speed Signal Reconstruction for Compressive Sensing Applications

Abstract

Compressive sensing (CS) is an emerging technique that has great significance to the design of resource-constrained embedded signal processing systems. However, signal reconstruction remains a challenging problem due to its high computational complexity, which limits the practical application of compressive sensing. In this paper, we propose an algorithmic transformation referred to as Matrix Inversion Bypass (MIB) to reduce the computational complexity of Orthogonal Matching Pursuit (OMP) based signal reconstruction. The proposed MIB transform naturally leads to a parallel architecture for dedicated high-speed hardware implementations. Furthermore, by applying the proposed MIB transform, the energy consumption of signal reconstruction can be reduced as well. This is vital to many embedded signal processing systems that are powered by batteries or renewable energy sources. Simulation results of a wireless video monitoring system demonstrate the advantages of the proposed technique over the conventional OMP-based technique in improving the speed, energy efficiency, and performance of signal reconstruction.