RDF description Dr. Gorka Azkune

Researcher & Project Manager

Sep. 2013  -  Present
gorka.azkune [at] deusto.es

 +34 944 139 003

Learning for Dynamic and Personalised Knowledge-based Activity Models

Directed by: Diego López-de-Ipiña     Co-advisor:  Liming Luke Chen

 15 Jul 2015 - 09:30
 Universidad de Deusto
 Cum Laude by unanimity

 Viva panel

Javier Jaén Vocal
Cruz Enrique Borges Secretary
Basilio Sierra Chair

Human activity recognition is one of the key competences for human adaptive technologies. The idea of such technologies is to adapt their services to human users, so being able to recognise what human users are doing is an important step to adapt services suitably.

One of the most promising approaches for human activity recognition is the knowledge-driven approach, which has already shown very interesting features and advantages. Knowledge-driven approaches allow using expert domain knowledge to describe activities and environments, providing efficient recognition systems. However, there are also some drawbacks, such as the usage of generic and static activity models, i.e. activities are defined by their generic features - they do not include personal specificities - and once activities have been defined, they do not evolve according to what users do.

This dissertation presents an approach to using data-driven techniques to evolve knowledge-based activity models with a user’s behavioural data. The approach includes a novel clustering process where initial incomplete models developed through knowledge engineering are used to detect action clusters which describe activities and aggregate new actions. Based on those action clusters, a learning process is then designed to learn and model varying ways of performing activities in order to acquire complete and specialised activity models. The approach has been tested with real users’ inputs, noisy sensors and demanding activity sequences. Results have shown that the 100% of complete and specialised activity models are properly learnt at the expense of learning some false positive models.