Database Reference
In-Depth Information
Merged Ontology (SUMO)[8]. Sensor ontologies are used to establish a termi-
nology for sensors, their properties, capabilities and services. OntoSensor has
a number of advantages, including self-descriptive metadata embedded in the
descriptions, which can be used in various sensor discovery and reasoning ap-
plications. OntoSensor illustrates a semantic approach to sensor description and
provides an extensive knowledge model. However, this approach lacks a distinc-
tive data description model to facilitate interoperable data representation for
sensor observation and measurement data. Additionally, it does not facilitate
the specification or inclusion of context by the end user.
In [1], the authors describe a semantic model for heterogeneous sensor data
representation. A sensor data ontology is created based on the Sensor web En-
ablement (SWE)[7] and SensorML data component models. Semantic relation-
ships and operational constraints are deployed in a uniform structure to describe
the sensor data. The ontology based model allows machines to process and in-
terpret the emerging semantics to create intelligent sensor network applications.
However, this work is in an early stage of development, with many of its aims
and goals yet to be implemented, whereas we have a working prototype system
which facilitates interaction with domain experts and full query interface.
In [12], the authors represent context with varying granularity with a tuple con-
sisting of an RDF triple defining the relationship, a lifespan and a conditional con-
fidence value. This project aims to reduce uncertainty in context integration. The
method used to achieve this is combining multiple sources of information and us-
ing a Bayesian approach to calculate conditional confidence values. This is useful
for the target ubiquitous computing environment but is not suitable for an ever-
changing set of events to be detected using multiple sensors in multiple locations.
In the core target domain of analysing sensor data corresponding to cyclists,
there are a number of tools available which allow a limited analysis for sensor data.
The most successful commercial application for analysing power meter data in the
cycling domain is TrainingPeaks WKO+ [9]. An open source application, Golden
Cheetah [4] can also be used to analyse cycling sensor data. Querying in WKO+ is
limited to identifying the minimum/maximum/average data value for each stream
for a lap-by-lap or specific time period defined by the user. Apart from the wattage
analysis, no additional variables such as speed or current position can be applied
as a filter. Querying is not supported by Golden Cheetah. In addition, Neither of
these applications can support user defined events or context.
6 Conclusions
Sensor technology is used in many application areas now as a means of automated
data generation and collection. However, the low level nature of these devices and
the often complex query requirements of end users and specialists, means that a
considerable gap exists between the information generation and end user queries.
In this research, our goal was to minimse or even close that gap by allowing users
to specify events that would lead to contextual enrichement of the data sources.
Our system begins with an automatic process of basic enrichment which we refer
