Towards a better sensor data accuracy via quality of monitoring and semantic clustering

• Published in: International journal of communication systems Vol. 32; no. 10
• Main Authors: Campos, Nidia G S, Rocha, Atslands R, Gomes, Danielo G
• Format: Journal Article
• Language: English
• Published: Chichester Wiley Subscription Services, Inc 10.07.2019
• Subjects: Accuracy; Clustering; distributed data fusion; Energy conservation; Local area networks; Monitoring; Nodes; Power consumption; QoM; semantic clustering; Semantics; Wireless sensor networks
• ISSN: 1074-5351; 1099-1131
• DOI: 10.1002/dac.3957

Summary Semantic clustering organizes wireless sensor nodes in clusters to detect a relevant event for an application. These nodes send only a message reporting it to the sink node while the other nodes decrease their resource utilization to save energy. In this work, we enhance semantic clustering with quality of monitoring (QoM) attributes of the event detection probability with variance reduction or dissimilarity measure. This approach allows each node to evaluate the accuracy of detection and the correlation level of the data gathered locally and by its neighbors. Experiments in a WLAN and FIT/IoT‐Lab show QoM attributes improve the accuracy, especially in an extensive network where dissimilarity measure can keep it high independently of the cluster size. QoM also increases the sampling of these events by up to 60.7% while it decreases the number of messages sent to the sink between 13.1% and 32.9% without affecting the cluster formation or power consumption in almost all experiments. Each node executes the algorithm of semantic clustering with QoM analysis at sensing interval. The node realizes if the detected event is relevant or not and if its neighborhood is also monitoring it with high accuracy. A relevant event triggers the cluster formation after the analysis of the event detection probability (EDP), the variance reduction (VR), or the dissimilarity measure. Our approach enhances the accuracy of detected events independently of cluster size and without damage wireless sensor network (WSN) power consumption.