Fast, accurate event classification on resource-lean embedded sensors

  • Authors:
  • Hao Jiang;Jason O. Hallstrom

  • Affiliations:
  • School of Computing, Clemson University, Clemson, SC;School of Computing, Clemson University, Clemson, SC

  • Venue:
  • EWSN'11 Proceedings of the 8th European conference on Wireless sensor networks
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

In wireless sensing applications, it is often necessary to identify high-level events based on low-level sensor signals. Due to the limited computing and energy resources available on existing hardware platforms, achieving high precision classification of high-level events in-network is a challenge. In this paper, we present a new classification technique for identifying events of interest on resource-lean sensors. The approach introduces an innovative condensed kd-tree data structure to represent processed sensor data and uses a fast nearest neighbor search to determine the likelihood of class membership for incoming samples. The classifier consumes limited resources and provides high precision classification. To evaluate the approach, two case studies are considered, in the contexts of human movement and vehicle navigation, respectively. The classification accuracy is above 85% across the two case studies.