Talk: Mohammadreza Mahmudimanesh

Title: Compressed Sensing in Wireless Sensor Networks

Speaker: Mohammadreza Mahmudimanesh


Compressed Sensing (CS) is a novel sampling theory that allows to accurately recover a compressible signal from few random linear measurements. CS has applications in virtually all sensory systems where acquiring individual samples is expensive or infeasible. A Wireless Sensor Network (WSN) is a sensory system comprised of resource-limited sensor nodes. Transferring every single sample in WSN usually causes a data traffic that exceeds the network capacity and renders the WSN lifetime unacceptable.

This talk will review the fundamental theory of CS and its applications in WSN. Then, we present our Spatiotemporal CS model for WSN. According to the CS theory, sampling rate grows logarithmically with the size of the discrete compressible signal. Therefore, extending the dimension of the CWS signal to the temporal domain leads to a more efficient data acquisition technique. In particular, when the data collection can tolerate higher latencies, we can acquire compressive measurements over longer periods. The higher the level of delay-tolerance, the more significant is the reduction of sensor data traffic. Accordingly, our model allows a tunable sampling period in order to maximize the benefit from temporal correlations. We also shortly introduce our new work, integrating CS into Collection Tree Protocol.