Exploiting Patterns of Object Use for Human Activity Recognition

Abstract

The accurate recognition of activities of daily living (ADL) is fundamental in the support and provision of assistance for the elderly and cognitively impaired. Current ontology-based techniques and knowledge-driven model object concepts form assumptions and everyday knowledge of objects used for activities. Activities modelled from assumptions and everyday knowledge can lead to incorrect recognition results of routine activities and possible failure to detect abnormal activity trends. A significant step to the accurate recognition of activities of daily living is the discovery of the object usage for specific routine activities. This chapter presents an approach that discovers object usage for routine activities using latent Dirichlet allocation (LDA) topic modelling. The object usage discovery augments an activity ontology that enables recognition of simple activities of daily living in the home environment. The proposed approach is evaluated and validated using the Kasteren and Ordonez datasets.