$\beta$ Link: A Pattern Recognition Approach for Identifying Superior Event-to-Sink Links in a Spatiotemporal Correlated Sensor Network
Lawrence Mwenda Muriira · 2022
In large-scale wireless sensor networks (WSNs) deployment, sensed data for dissemination are highly spatial-temporal correlated within close geographical proximity. The sink is not interested in individual's sensor data but accurate information of the physical phenomenon. Pattern recognition has been proven to be a successful technique in identifying WSN links with superior performance for data transmission among spatial-temporal correlated links. This paper proposes βLink a data-driven design framework, which exploits a pattern recognition technique to identify this superior performance route for data transmission (Beta link). The technique provides an insight into; a metric that measures the degree of temporal link correlation using a Two State Markov Chain approach, and pattern classification using Support Vector Machine that also models spatiotemporal link correlation. The scheme could reduce the number of links to be utilized for data dissemination to 21 %, a significant reduction to the overall network's transmission cost. Therefore, the results to be presented are encouraging as it clearly indicates the potential of Pattern Recognition approach in identifying Beta links.