Adaptive-PCA: An event-based data aggregation using principal component analysis for WSNs

Abstract

© 2015 IEEE. Dimensionality reduction techniques are convenient for data aggregation to reduce battery energy consumption in sensor nodes. Normally, principal component analysis (PCA), a dimensionality reduction technique, has been used for data aggregation in WSNs. However, PCA yields to data errors when the sensing data are not related. The PCA processing time is also an issue in an urgent situation that the sensing data are required to be transmitted to the base station instantly. This paper proposes a novel data aggregation mechanism for WSNs, called Adaptive-PCA. In Adaptive-PCA, PCA is performed dynamically based on the sensing data. In a normal situation, PCA is performed for data aggregation to reduce the number of transmitted packets. On the other hand, in an urgent situation, sensing data change dramatically, PCA is not performed; the sensing data are transmitted to the base station instantly. Adaptive-PCA consists of two schemes which are 1) event checker and 2) PCA data accuracy checker. These two schemes drive each sensor node whether perform PCA or instantly transmit the sensing data. The simulation results show that Adaptive-PCA adjusts the number of transmitted packets to the environmental changes. Using Adaptive-PCA, the total battery energy consumption is less than that of a traditional WSN. Also, the data accuracy of Adaptive-PCA is higher than that of Non-adaptive-PCA.