Improved Sensor Sampling Method for the Joint Dictionary Learning and Compressive Data Gathering in WSNs with the Aid of Information Theory

In the last decade, to reduce the costs of environmental monitoring, the data aggregation based on the joint dictionary learning and compressive sensing technique in wireless sensor networks has been considered. In this article, a deterministic and non-random sampling design for use in this data aggregation method is presented. This method is based on estimating the amount of mutual information of sensor data and is obtained by sampling all of them in a short part of the data collection round named the training phase. In the next and main stage of the data collection period, only the nodes that provide the most information about the non-sampled nodes are scheduled to sample. Simulation results for real signals show that when the number of sampling sensors comprises still about 25% of the total network nodes, average energy savings of more than 12% can be achieved over a reference sampling method.