Robust Extraction of Temporal Correlation from Noisy Data

Wireless sensor networks (WSNs) play an important role in several application fields. They are composed by a dense deployment of small and low-power devices (called sensors) that cooperate together in order to perform some special tasks. These networks are often characterized by the spatial and temporal correlation of the sensor observations. Several works are investigating these correlations in order to design efficient communication protocols able to improve the performance of the networks. On these basis, it is necessary to: 1) extract the correlation function and 2) ensure that it is always valid, even if data are impaired by noise. The proposed approach is inspired to [1], where the authors have developed a new technique to robustly extract the statistical characteristics of process variations from experimental measurements in order to analyze and optimize the design of integrated systems. Our aim here is to investigate the correlation issue in the WSNs context where the data sensed by the nodes are usually correlated in the space and temporal domains. Motivated by this consideration, we propose a robust method to extract temporal correlation function parameters from noisy data gathered by a WSN and use some simulation results to validate our technique. Moreover, we analyze the reliability of our method with respect to the amount of collected measurement data and show that, on the basis of the obtained results, the temporal extraction process is efficient and robust in real cases.