Selected Readings in Vision and Graphics
edited by Luc Van Gool, Gábor Székely, Markus Gross, Bernt Schiele

Volume 41

Nicholas J. Kern


Multi-Sensor Context-Awareness for Wearable Computing.

First edition 2006, XII, 142 pages, € 64,00. ISBN 3-86628-056-4

In wearable computing sensor data is used to infer the user’s situation and adapt a device’s behavior accordingly. We investigate three principal issues of context aware wearable computer systems, namely the perception, modeling, and use of context information. The investigation is embedded in the context of two application scenarios.

We investigate two sources of context information: acceleration and audio sensors. Using body-worn acceleration sensors, we extract information about the user’s physical activity, such as sitting, walking, or writing on a white board. We use audio to find information about the social situation of the user. We therefore classify the auditory scene into restaurant, street, or lecture hall sounds.

The first application scenario is a meeting recorder that records not only audio and video of meetings, but also additional personalized annotations of the user’s context, allowing for associative retrieval within large recordings by the user. For example the detection of hand-shakes allows to find greeting situations, or a speaker identification scheme allows to tell a presentation from a discussion phase.

For the second application scenario, the estimation of the user’s interruptibility for automatic mediation of notifications, we developed a (context) model of human interruptibility, which distinguishes between the interruptibility of the user and that of his environment. We evaluate the model in a user study and propose an algorithm to estimate the interruptibility from sensor data. The algorithm allows to learn the interruptibility directly from sensor data, requiring only the annotations for the interruptibility. We show the feasibility of the approach using data sets of acceleration, audio, and location sensor data of a maximum length of two days. Using this data set we obtain a classification score of 90-97%.


Nicky Kern studied at the Swiss Federal Institute of Technology (ETH Zurich) in Zurich, Switzerland and Institut Eurécom in Sophia Antipolis, France. He obtained a M.Sc. (Diplom) in computer science from ETH Zurich in 2001. He joined the Perceptual Computing and Computer Vision Group as a research assistant to work on sensor-based context awareness for wearable computing. He moved to the University of Technology Darmstadt (TU Darmstadt) in Germany in 2004, where he finished his PhD in 2005. He mainly worked on context extraction from body-worn sensors, notification systems, and usability aspects of context-aware systems.

Keywords: Multi-Sensor, Context-Awareness, wearable Computing

Reihe " Selected Readings in Vision and Graphics " im Hartung-Gorre Verlag

Direkt bestellen bei / to order directly from: