Database Reference
In-Depth Information
and automatically apply data transformations based on this information. The
role of the Sensor Enablement processor is to generate machine queryable XML
files.
Contextual Enrichment.
Following Sensor Enablement, basic queries can be
performed using XQuery. However, many of the queries listed in Table 1 cannot
be expressed at this point as the data lacks the necessary semantics. To reduce
this semantic gap, the system uses Participant and Activity Profiles to under-
stand the deployment domain (or activity), the sensors, participants, and key
events. Essentially, these are metadata constructs used to describe any object
or person measured by the sensor. In the following section, we describe how the
specification of event detection definitions can mine for the information necessary
to end user queries.
Integration.
The querying needs of the coaches and sports scientists are exten-
sive, ranging from basic sensor analysis and comparison to anthropometric based
analysis of participants with multiple sensors deployed while engaged in some
activity. Some information such as sensor data is extensive and specific to a cer-
tain time span, or geographic location whereas much physiological or biological
data rarely changes. In general, a single sensor cannot meet information needs
and multiple sources of evidence must be integrated to provide both results and
highs levels of accuracy. For the current experiments used in this paper, sensors
were manually synchronised and this processor was not used. However, analysis
of sporting events is often chaotic and synchronisation of inexpensive sensors
devices cannot be guaranteed. Thus, integration is an important part of current
work.
Query Interface.
As all the sensor data is converted to XML during sensor
enrichment, queries can be expressed in XQuery or XPath. As neither of these
languages are intuitive to non-computing users, a view based system is currently
employed [5] which also offers optimisation features for high volume datasets.
However, Contextual Enrichment is an important enabler for query processing
as will be shown in the next section.
3 Context Profiles and Event Mining
Context Profiles provide genericity to the system and thus, facilitate hetero-
geneity. Individuals will have different physiological characteristics, activities
will have different timings, layouts and formats, and sensors will come and go,
bringing new information and heterogenous structures. The activity in which
participants are being measured provides the widest range of heterogeneities.
For this reason, it receives a more detailed discussion here, including how it can
be used to extract new knoweldge from the sensor database, that can later be ex-
ploited by the query processor. All profiles and function descriptions (discussed
later) are stored in the System Repository.
