Design methodology for context-aware wearable sensor systems

  • Authors:
  • Urs Anliker;Holger Junker;Paul Lukowicz;Gerhard Tröster

  • Affiliations:
  • Institut für Elektronik, ETH Zürich, Zürich, Schweiz;Institut für Elektronik, ETH Zürich, Zürich, Schweiz;Institut für Elektronik, ETH Zürich, Zürich, Schweiz;Institut für Elektronik, ETH Zürich, Zürich, Schweiz

  • Venue:
  • PERVASIVE'05 Proceedings of the Third international conference on Pervasive Computing
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

Many research projects dealing with context-aware wearable systems encounter similar issues: where to put the sensors, which features to use and how to organize the system. These issues constitute a multi-objective optimization problem largely determined by recognition accuracy, user comfort and power consumption. To date, this optimization problem is mostly addressed in an ad hoc manner based on experience as well as trial and error approaches. In this paper, we seek to formalize this optimization problem and pave the way towards an automated system design process. We first present a formal description of the optimization criteria and system requirements. We then outline a methodology for the automatic derivation of Pareto optimal systems from such a description. As initial verification, we apply our methodology to a simple standard recognition task using a set of hardware components, body locations and features typically used in wearable systems. We show that our methodology produces good results and that a simple example can provide information that an experienced system designer would have problems extracting by other means.