Most of ambient intelligence studies have tried to employ inductive methods (e.g., data mining) to discover useful information and patterns from data streams on sensor networks. However, since the spaces have been sharing their information with each other, it is difficult for such inductive methods to conduct the discovery process from the sensor streams intermixed from the heterogeneous sensor networks. In this paper, we propose an ontology-based middleware system to improve sustainability of context-aware service in the interconnected smart spaces. Two main challenges of this work are i) sensor data preprocessing (i.e., session identification) and ii) information fusion (i.e., information integration). The ontology in each sensor space can provide and describe semantics of data measured by each sensor. By aligning these ontologies from the sensor spaces, the semantics of sensor data captured inside can be compared. Thus, we can find out not only relationships between sensor streams but also temporal dynamics of a data stream. To evaluate the proposed method, we have collected sensor streams from in our building during 30 days. By using two well-known data mining methods (i.e., co-occurrence pattern and sequential pattern), the results from raw sensor streams and ones from sensor streams with preprocessing were compared with respect to two measurements recall and precision.

ontology, semantic sensor networks;, preprocessing, stream mining