A quantitative method for revealing and comparing places in the home

  • Authors:

  • Ryan Aipperspach;Tye Rattenbury;Allison Woodruff;John Canny

  • Affiliations:

  • Berkeley Institute of Design, Computer Science Division, University of California, Berkeley;Berkeley Institute of Design, Computer Science Division, University of California, Berkeley;Intel Research, Berkeley;Berkeley Institute of Design, Computer Science Division, University of California, Berkeley

  • Venue:
  • UbiComp'06 Proceedings of the 8th international conference on Ubiquitous Computing
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

Increasing availability of sensor-based location traces for individuals, combined with the goal of better understanding user context, has resulted in a recent emphasis on algorithms for automatically extracting users' significant places from location data. Place-finding can be characterized by two sub-problems, (1) finding significant locations, and (2) assigning semantic labels to those locations (the problem of “moving from location to place”) [8]. Existing algorithms focus on the first sub-problem and on finding city-level locations. We use a principled approach in adapting Gaussian Mixture Models (GMMs) to provide a first solution for finding significant places within the home, based on the first set of long-term, precise location data collected from several homes. We also present a novel metric for quantifying the similarity between places, which has the potential to assign semantic labels to places by comparing them to a library of known places. We discuss several implications of these new techniques for the design of Ubicomp systems.