Annotating smart environment sensor data for activity learning

  • Authors:

  • S. Szewcyzk;K. Dwan;B. Minor;B. Swedlove;D. Cook

  • Affiliations:

  • Washington State University, Pullman, WA, USA;Washington State University, Pullman, WA, USA;Washington State University, Pullman, WA, USA;Washington State University, Pullman, WA, USA;(Correspd. Tel.: +1 509 335 4985/ Fax: +1 509 335 3818/ E-mail: cook@eecs.wsu.edu) Washington State University, Pullman, WA, USA

  • Venue:
  • Technology and Health Care - Smart Environments: Technology to Support Healthcare
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

The pervasive sensing technologies found in smart homes offer unprecedented opportunities for providing health monitoring and assistance to individuals experiencing difficulties living independently at home. In order to monitor the functional health of smart home residents, we need to design technologies that recognize and track the activities that people perform at home. Machine learning techniques can perform this task, but the software algorithms rely upon large amounts of sample data that is correctly labeled with the corresponding activity. Labeling, or annotating, sensor data with the corresponding activity can be time consuming, may require input from the smart home resident, and is often inaccurate. Therefore, in this paper we investigate four alternative mechanisms for annotating sensor data with a corresponding activity label. We evaluate the alternative methods along the dimensions of annotation time, resident burden, and accuracy using sensor data collected in a real smart apartment.