An Ontology-Based Framework for Semantic Data Preprocessing Aimed at Human Activity Recognition

Over the last few years, complex systems which collect data from a considerable number of sources are increasing. However, it is not always possible to have a clear overall view of the information ontained within data, due to both their granularity and to their wide amount. Since an analysis procedure able to take into account the semantics of records is often needed, ontologies are becoming widely used to describe the domain and to enrich the acquired data with its significance. In this paper, we propose an ontology-based methodology aiming to perform semantic queries on a data repository, whose records originate from a network of heterogeneous sources. The main goal of such queries is the pattern matching process, i.e., recognition of specific temporal sequences in fine-grained data. In our framework, benefits deriving from the implementation of a domain ontology are exploited in different levels of abstraction. Thereafter, reasoning techniques represent a preprocessing method to prepare data for the final temporal analysis. Our proposed approach will be applied to the ongoing AALISABETH, an Ambient Assisted Living project aimed to discover and manage the behaviour of monitored users.