Exploiting Patterns of Object Use for Human Activity Recognition

5G Networks And Advancements In Computing, Electronics, And Electrical Engineering

also raises privacy concerns. The process of recognising activities via sensor-based activity recognition can be data driven, knowledge driven or a combination of both techniques. Data-driven techniques employ machine learning and statistical methods to discover the data patterns for activities to be inferred. Although data driven techniques have been used by the authors in (Ordonez et al., 2013; Kasteren et al., 2010; Chen, Nugent, & Wang, 2011) demonstrating its strengths in the recognition of activities, inferences in most cases for data driven approaches are hidden and or latent. Recognised activities are however needed to be expressed in an understandable format for the end user. Data driven techniques, in addition, are unable to integrate context aware features to enhance activity recognition.

Conversely, ontology based and knowledge driven techniques model activities as concepts. These concepts are associated to the everyday knowledge of object usage in the home environment through a knowledge engineering process. The modelling process involves associating low level sensor data to the relevant activity to build a knowledge base of activities in relation to sensors and object. The recognition of activities is achieved by logical inference and or inclusion of subsumption reasoning. Unlike data driven techniques, knowledge driven techniques are expressively clear and activity recognition results are in the format easily understood by the end user (Chen, Nugent, Hoey, Cook, & Yu, 2012). Modelling activity ontology concepts for Knowledge driven techniques largely depend on everyday knowledge of activities and object use to build and construct activity ontologies. Knowledge of object use are generic and mostly by assumptions regular everyday knowledge of the objects are used for routine activities or even wikiknowhow¹ (Okeyo, Chen, Wang, & Sterritt, 2012a).