Compressive sensing and matrix completion in Wireless Sensor Networks

As the size of In Wireless Sensor Networks (WSNs) continues to grow, the amount of data, for processing and transmitting becomes enormous. This large number of data is one of the factors that is speeding up power dissipation. Once the wireless devices are spatially distributed across the physical monitored field, their batteries, designed to operate for a determined period of time, cannot be changed anymore. This challenging issue requires effective mechanisms to handle constraints such as energy consumption, delay and cost. The Compressive Sensing (CS) theory as well as the Matrix Completion (MC) theory are two recently developed techniques in signal processing, which in combination with the sparse and correlated nature of WSNs data, hold promising enhancement to the network lifetime. In this paper, we compare the performance of the MC and the CS. These techniques take advantage of the intra-sensor (temporal) and the inter-sensors (spatial) data structure correlation differently. The simulation analysis shows that the MC with a naive random sampling scheme outperforms two high complex CS based algorithms.