Poster: Reducing power consumption of human activity sensing using compressed sensing

  • Authors:
  • Daito Akimura;Yoshihiro Kawahara;Tohru Asami

  • Affiliations:
  • The University of Tokyo, Hongo, Bunkyo-ku, Tokyo, Japan;The University of Tokyo, Hongo, Bunkyo-ku, Tokyo, Japan and Georgia Institute of Technology, Atlanta, Georgia;The University of Tokyo, Hongo, Bunkyo-ku, Tokyo, Japan

  • Venue:
  • Proceedings of the 9th ACM Conference on Embedded Networked Sensor Systems
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

Almost all recent mobile phones are equipped with multiple sensors, such as cameras, GPS, and accelerometers. By exploiting the sensing features, we capture many different events and share them over the mobile network. One of the most important challenges for such a participatory sensing system is the reduction of the battery consumption of the mobile device because the sensing task usually runs as a secondary task and should not disrupt the primary tasks of the mobile phone, such as phone calls. We overcome this problem by compressing the sensed data and sending a minimum amount of data over the wireless link. The compressed sensing (CS) technique is used for this compression with simple matrix operations at the mobile side, and the CPU-intensive reconstruction is performed on the resource-rich machine on the network side. In this paper, we validate this idea by implementing it on the iPhone/iPod platform. Since CS is a lossy compression technique, the reconstructed signal contains errors depending on the degree of sparseness of the original signal. We show that our system can reduce power consumption by approximately 16% compared with ZIP compression, and evaluate the reconstruction error using real sensing data of 86 test subjects.