D5.3 Behavioural Profile Learning

This document presents the current work carried out with respect to behavioural profile learning for the purpose of supporting the patientcustomised services targeted within Dem@Care. The objective of behavioural profile learning is to dynamically discover person-specific behaviour patterns that can be utilised to improve the recognition of the performed activities and to allow for PwD-tailored behaviour interpretation and assessment. These behaviour patterns may encapsulate a variety of aspects, including the manner in which daily activities are performed, idiosyncratic and habitual knowledge, as well as recurrent routines. In this report, we present the first efforts towards the aforementioned directions. Two approaches for discovering, modelling and recognising ADL are proposed, a supervised one using egocentric video data as input, and an unsupervised one using as input data from a fixed camera. The goal is to allow for the discovery of behaviour patterns through machine learning. In addition, an ontology-based pattern-oriented approach is presented for capturing in a formal manner higher-level behavioural aspects that encapsulate richer semantics.
Download the report

© 2011 Dem@Care | Centre for Research and Technology Hellas :: Information Technologies Institute