A memristor-based compressive sensing architecture

Abstract

Memristors are considered as one promising candidate for future memory and computing fabrics. However, the design of memristor-based circuits is under a critical challenge of inevitable variations due to non-ideal fabrication processes and the resulted performance uncertainties. This kind of randomness can be utilized in many other applications, such as compressive sensing based data acquisition, which is conducted by a random sensing matrix. Existing compressive sensing systems are usually implemented in digital CMOS circuits, which suffer the problems of high hardware complexity and limited sampling speed. In this paper, we exploit the inherent variations in memristor devices to generate random sensing matrices for compressive sensing and achieve low cost and high performance operations. Simulation results demonstrate the advantages of the proposed memristor-based compressive sensing architecture.