*4.2. Context Model of Activity*

An eating activity is a complex activity which consists of many low-level contexts, such as the spatial and temporal background, movement of the wrist, and temperature. Table 5 shows the web ontology language (OWL) representation of the proposed context model based on the activity theory and the "Five W's", for systemic analysis on an eating activity. Four subclasses represent the components of the Five W's, except 'Why", as this context is considered difficult to measure with the limited sensor environment. A subject property consists of goal-directed processes (actions) and the unconsciously appearing status of the body (body temperature, posture, and so on; operations). Nine properties describe the low-level context of the eating activity. Each intermediate node is linked to leaf nodes, namely, sensors, which are considered as related. Although the movement of the user is the main feature to recognize activities, used for most intermediate nodes, environmental features could

also contribute, especially when the movement patterns are diverse. The proposed context model has three other subclasses (object, spatial, and temporal properties) to consider those environmental factors. A temporal property uses the system time for judging one property, whether the current time is appropriate for eating. A spatial property has four properties, such as whether the user is indoors or outdoors, changes of space, and whether the intensity of illumination of the space is appropriate for eating.


**Table 5.** OWL representation of the context model for eating activity recognition.
