Dictionary Learning for Reconstructing Measurements of Analog Wireless Sensor Nodes

Wireless Sensor Nodes communicating measurements to a base station is one of the scenarios in the emerging field of Machine-Type-Communication. Those systems rely on low complexity of the nodes, due to cost and energy consumption. The main idea of this paper is to employ a low complexity analog modulation scheme in the node, and combine it with state of the art digital signal processing in the base station. Specifically, we focus on Amplitude Modulation in a point to point scenario facing noise and hardware offsets. We show that under certain assumptions this transmission can be described by a linear model. Subsequently we utilize payload (measurement) signal structure, namely sparsity, to estimate the payload signals as well as the hardware offsets using a dictionary learning algorithm. Numerical simulations show, that for realistic noise assumptions the algorithms are able to reconstruct payload signals and estimate hardware offsets.