A study of deterministic sensors placement for sparse events detection in WSN based on Compressed Sensing

The problem of events detection in Wireless Sensor Networks (WSN) is investigated from the perspective of Compressed Sensing (CS). In WSN, when the events to detect are rare, the number of sensors with innovative and useful measurements is far lower than that of deployed sensors. In this way, to reduce the data processing step burden, the number of sensors measurements used in the events detection should be chosen lower than the number of monitored grids of the supervised area. Therefore, the sensors subset choice should be optimized in the sense of event detection performance. For this objective, CS theory is well adapted and recent works have addressed this problem. In particular, the well known Matching Pursuit (MP) algorithm is formulated in the context of events detection and counting through a Greedy version algorithm denoted GMP. In the existing work, the active sensors subset selection is processed randomly. In this paper, two contributions are proposed. First, we introduce new deterministic sensors selection schemes based on either the observation or the channel matrix correlation and energies measures. Second, we consider a two stages version of GMP (2S-GMP) through which we show that separating the detection and counting steps allows not only to reduce the computational burden but also to improve the performance.