Ontology-Based High-Level Context Inference for Human Behavior Identification

Abstract

Recent years have witnessed a huge progress in the automatic identification of individual primitives of human behavior, such as activities or locations. However, the complex nature of human behavior demands more abstract contextual information for its analysis. This work presents an ontology-based method that combines low-level primitives of behavior, namely activity, locations and emotions, unprecedented to date, to intelligently derive more meaningful high-level context information. The paper contributes with a new open ontology describing both low-level and high-level context information, as well as their relationships. Furthermore, a framework building on the developed ontology and reasoning models is presented and evaluated. The proposed method proves to be robust while identifying high-level contexts even in the event of erroneously-detected low-level contexts. Despite reasonable inference times being obtained for a relevant set of users and instances, additional work is required to scale to long-term scenarios with a large number of users.

Keywords: activities; context inference; context recognition; emotions; human behavior identification; locations; ontological reasoning; ontologies.

Figure 1

Graphical representation of the combination of low-level contexts that compose the high-level contexts modeled in the Mining Minds Context Ontology.

Figure 2

Mining minds context ontology: the class Context, its subclasses and the relations among them.

Figure 3

Mining minds context ontology: definition of the ten subclasses of HighLevelContext. (a) OfficeWork; (b) Sleeping; (c) HouseWork; (d) Commuting; (e) Amusement; (f) Gardening; (g) Exercising; (h) HavingMeal; (i) Inactivity; (j) NoHLC.

Figure 4

Exemplary scenario representing low-level contexts and high-level contexts.

Figure 5

Representation of the instances of low-level context for the exemplary scenario by using the Mining Minds Context Ontology in Protégé. (a) llc_358_office is a member of the class Office; (b) llc_359_boredom is a member of the class Boredom; and (c) llc_360_sitting is a member of the class Sitting.

Figure 6

Representation of the instances of unclassified high-level context for the exemplary scenario by using the Mining Minds Context Ontology in Protégé. (a) hlc_70; (b) hlc_71; (c) hlc_72; and (d) hlc_73 are composed of some of the low-level contexts llc_358_office (member of the class Office), llc_359_boredom (member of the class Boredom) and llc_360_sitting (member of the class Sitting).

Figure 7

Representation of the instances of classified high-level context for the exemplary scenario by using the Mining Minds Context Ontology in Protégé. (a) hlc_72; and (b) hlc_73, which are both inferred to be members of the class OfficeWork, are composed of some of the low-level contexts llc_358_office (member of the class Office), llc_359_boredom (member of the class Boredom) and llc_360_sitting (member of the class Sitting).

Figure 8

Mining Minds High-Level Context Architecture.

Figure 9

Processing time invested by each of the HLCA components in the context identification. The number of instances indicates the amount of previously processed high-level contexts when the recognition process is triggered.

Figure 10

Size of the Context Storage depending on the number of persisted instances of high-level context. It must be noted that the storage of each high-level context instance has associated the storage of the low-level context instance which triggered its creation. Thus, for example, 250,000 instances in the X-axis represent 250,000 high-level contexts plus 250,000 low-level contexts stored on disc.