An Ontology-based Framework for Querying Heterogeneous Sensor Data

Abstract

Advancement in the Sensing Technology has resulted in the wide spectrum of modality and heterogeneity in data captured by heterogeneous sensors. This randomized temporal data has immense potential to generate information that can drive the process of gaining insightful domain knowledge to achieve efficient decision making. Due to this existing heterogeneity, there is a lack of interoperability of various sensor devices giving rise to complexities in the process of sensor data fusion and aggregation. Hence, there is a need for managing sensor data dynamics in a way that facilitates efficient aggregation, searching, analyzing, reuse and exchange of sensed data. To achieve all this, the aggregation needs to be performed at concept/entity level in meaningful ways to present information in an interpretable format that generates extensive knowledge and value. Semantic approaches - Ontologies - play an important role in solving these issues. In order to achieve the task of gaining knowledge through intelligent analysis of sensed data, Semantic Web technologies are implemented to achieve interoperability of sensing devices and systems. Ontologies are defined as “well-founded mechanism for the representation and exchange of structured information”. These technologies can aid in the management, storage, fusion of sensor devices and measurement data, so as to facilitate efficient information retrieval and mining from the Ontology-Based Knowledge Systems. This thesis presents research work undertaken to perform efficient heterogeneous sensor data analysis with the development of Ontological Knowledge models and fetching information through SPARQL-based query firing. Two ontological knowledge models are proposed for activity recognition. The first one targets heterogeneous sensors generating homogeneous data and the other one is a probabilistic ontological model for targeting and fusing multimodal data from heterogeneous sensors for efficient activity recognition. Another ontological model takes in input as digital heterogeneous data from ECG-based sensors to classify heartbeat data into arrhythmia classes. It solves the issue of interoperability and heterogeneity in the input sensor data. An interval type-2 fuzzy-based ontological model is developed to target the uncertainty and heterogeneity in the heterogeneous water sensors data. The proposed model ensures effective handling of ambiguous and uncertain sensor observations, a shortcoming of elementary/classical knowledge-driven models. For all the proposed ontological knowledge models, SPARQL-based querying service is deployed to draw insightful knowledge in the from of inferences. The resultant frameworks solve the issue of sensor heterogeneity and provides solutions that promote reusability, interoperability and exchange of domain knowledge.