A spatiotemporal compressive data gathering method combining block-wise compressed sensing with logical mapping

Abstract

To fully exploit the spatiotemporal correlation of big sensory data towards transmission reduction, we propose a spatiotemporal compressive data gathering method combining block-wise compressed sensing (BCS) with logical mapping (LM). The method explores data correlation according to both the locations and contents of sensors and performs the dual compression of spatial and temporal domains. Specifically, temporal compression is achieved by random sampling at each sensor, as the temporal correlation of data is neglected by traditional BCS-based methods. Additionally, the spatial correlations caused by the contents of sensors and clusters have not been wholly exploited by the existing compressive data gathering methods yet, thus at edge devices, the temporal measurements of sensors are roughly sorted by the intra-cluster LM before the BCS-based spatial compression is performed; while at sink, the inter-cluster LM is used to coarsely order the data of clusters with joint reconstruction done afterward. The simulation results based on real-world data show that LM can effectively improve the compressibility of data. Our method not only guarantees the reconstruction quality but greatly reduces the transmissions and energy consumption of the sensor network.