Reliable activity detection for massive machine to machine communication via multiple measurement vector compressed sensing

Abstract

Compressed sensing based multiuser detection is a novel research field in massive machine to machine communication. Mainly focusing at decreasing signaling overhead, this approach implements sophisticated detection algorithms at the physical layer that jointly estimate activity and data. As a consequence, the reliability of the activity detection is crucial for the system performance as data is lost if users are erroneously classified as inactive. This paper introduces a novel approach to estimate node activity on a per frame basis by Multiple Measurement Vector Compressed Sensing approaches. This approach allows for reliable activity detection with complexity invariant of the length of the transmitted frame. Moreover, we are able to show that this approach works with only a few measurements available to the detector. In particular we demonstrate that reliable activity detection is possible if the number of observations is larger than the square root of the number of nodes in the system.