Open positions

Dynamic machine learning of human activities

Static models of human activity are typically inefficient due to their required effort in development and inability to cope with changing concepts during deployment.

Estimating quality of context information in patient daily rhythms

This project targets to develop a framework for reliable estimation of sensor signal quality and recognition performance of patient states.

Minimally supervised, iterative, template-based activity learning with body-worn sensors

The aim of this project is the development, implementation, and evaluation of methods for minimally supervised, iterative 'self training' of context recognition systems based on small numbers of s

Wearable optical sensors for activity spotting

This project will investigate a novel concept for wearable cameras in activity recognition, specifically to use an optical sensors with a small field of view placed at strategic locations to refine