Activity and Location Recognition Using Wearable Sensors

  • Authors:
  • Seon-Woo Lee;Kenji Mase

  • Affiliations:
  • -;-

  • Venue:
  • IEEE Pervasive Computing
  • Year:
  • 2002

Quantified Score

Hi-index 0.00

Visualization

Abstract

The authors propose a new method for detecting and classifying a person's activity using dead-reckoning-based location recognition and a set of inexpensive, wearable sensorsýa biaxial accelerometer, an angular-velocity sensor, and a digital compass. Using the measured acceleration and angular-velocity data, the method can recognize activities such as sitting, standing, and walking. It can also classify walking behaviors into three subcategories: walking on level ground, ascending a stairway, or descending a stairway. Based on this activity recognition, the authors propose a method for detecting transitions between preselected locations, which uses the integration of incremental user motions over time with heading measurements and a simple nearest-neighborhood algorithm. The authors conducted experiments at five indoor locations.