Fluctuations in step counts derived from accelerometer data

  • Lloyd Gabriel T Rizada National Institute of Physics, University of the Philippines Diliman
  • Damian N Dailisan National Institute of Physics, University of the Philippines Diliman
  • May T Lim National Institute of Physics, University of the Philippines Diliman
Keywords: fitness tracking, signal processing, algorithms

Abstract

Fitness tracking apps in smartphones measure activities such as step count. However, most app algorithms are hidden from the user, i.e. a blackbox, and what exactly is being measured is not known to the user. This paper implements a step counting algorithm disclosed by Pebble Technology Corporation. We also assessed how the calculated number of steps changed with the wearable (accelerometer) placement. Four mount points were considered: hand, arm, pocket, and ankle; while two activities: walking and running were analyzed. Step counts by the Pebble algorithm deviated from actual values by less than 20% the actual values. Running has generally better step counting estimates than walking (except on the ankle). An ankle placement yielded the most accurate walking measurement, while a hand placement had the closest measurement for running.

Published
2017-06-07
Section
Complex Systems and Data Analytics