Expanding sensor networks to automate knowledge acquisition

The availability of accurate, low-cost sensors to scientists has resulted in widespread deployment in a variety of sporting and health environments. The sensor data output is often in a raw, proprietary or unstructured format. As a result, it is often difficult to query multiple sensors for complex properties or actions. In our research, we deploy a heterogeneous sensor network to detect the various biological and physiological properties in athletes during training activities. The goal for exercise physiologists is to quickly identify key intervals in exercise such as moments of stress or fatigue. This is not currently possible because of low level sensors and a lack of query language support. Thus, our motivation is to expand the sensor network with a contextual layer that enriches raw sensor data, so that it can be exploited by a high level query language. To achieve this, the domain expert specifies events in a tradiational event-condition-action format to deliver the required contextual enrichment