Exploiting prior information for greedy compressed sensing based detection in machine-type communications

Abstract

In machine-type communications (MTC), a large number of devices are connected to an access point (AP), but only a few devices are active at a time. This sparse activity of devices makes compressed sensing (CS) technique a possible solution for joint activity detection and channel estimation problem in MTC. In this paper, we improve the performance of greedy CS based detection in MTC, by exploiting the prior probability of each device being active. We propose new improved greedy algorithms that minimize the probability of incorrect selection of nonzero indices using a correction function. Simulation results demonstrate the performance improvement of CS recovery with the improved greedy algorithms. In addition, we investigate the empirical performance of the improved algorithms when the prior information is inaccurate, which is natural in practice. With inaccurate prior information, we demonstrate that the performance of CS based joint activity detection and channel estimation employing the improved orthogonal matching pursuit (OMP) is superior to that of OMP with partially known support (OMP-PKS) in which the AP knows 30% or less active devices in advance.