Publication Type

Conference Proceeding Article

Version

publishedVersion

Publication Date

11-2017

Abstract

We propose a new activity sensing method, CapSense, which detects activities of daily living (ADL) by sampling the voltage of the kinetic energy harvesting (KEH) capacitor at an ultra low sampling rate. Unlike conventional sensors that generate only instantaneous motion information of the subject, KEH capacitors accumulate and store human generated energy over time. Given that humans produce kinetic energy at distinct rates for different ADL, the KEH capacitor can be sampled only once in a while to observe the energy generation rate and identify the current activity. Thus, with CapSense, it is possible to avoid collecting time series motion data at high frequency, which promises significant power saving for the sensing device. We prototype a shoe-mounted KEH-powered wearable device and conduct experiments with 10 subjects for detecting 5 different activities. Our results show that compared to the existing time-series-based activity recognition, CapSense reduces samplinginduced power consumption by 99% and the overall system power, after considering wireless transmissions, by 75%. CapSense recognizes activities with up to 90%.

Keywords

Energy-effciency, Activity Recognition, Wearable Device

Discipline

Artificial Intelligence and Robotics | Software Engineering

Research Areas

Intelligent Systems and Optimization

Publication

Proceedings of the 14th EAI International Conference on Mobile and Ubiquitous Systems: Computing, Networking and Services: MobiQuitous 2017, Melbourne, November 7-10

First Page

106

Last Page

115

ISBN

9781450353687

Identifier

10.1145/3144457.3144459

Publisher

ACM

City or Country

Melbourne, Australia

Download

Share

COinS