Activity sensing in the wild: a field trial of ubifit garden
Proceedings of the SIGCHI Conference on Human Factors in Computing Systems
UbiGreen: investigating a mobile tool for tracking and supporting green transportation habits
Proceedings of the SIGCHI Conference on Human Factors in Computing Systems
Using mobile phones to determine transportation modes
ACM Transactions on Sensor Networks (TOSN)
The Jigsaw continuous sensing engine for mobile phone applications
Proceedings of the 8th ACM Conference on Embedded Networked Sensor Systems
MobiCom '11 Proceedings of the 17th annual international conference on Mobile computing and networking
Balancing energy, latency and accuracy for mobile sensor data classification
Proceedings of the 9th ACM Conference on Embedded Networked Sensor Systems
ACE: exploiting correlation for energy-efficient and continuous context sensing
Proceedings of the 10th international conference on Mobile systems, applications, and services
Proceedings of the 10th international conference on Mobile systems, applications, and services
Energy-Efficient Continuous Activity Recognition on Mobile Phones: An Activity-Adaptive Approach
ISWC '12 Proceedings of the 2012 16th Annual International Symposium on Wearable Computers (ISWC)
Hi-index | 0.00 |
Our vision is a smartphone service that accurately tracks users' physical activities and transportation modes using accelerometer-based activity recognition carried out on the phone. A key challenge that needs to be addressed to realize this vision is the high power consumption of keeping the smartphone awake in order to perform sensor sampling, feature extraction, and activity classification. In this paper, we propose a two-tier classifier for reducing the wake time percentage of the activity recognition system, which we define as the percentage of time the activity recognition system keeps the smartphone awake. We compare our two-tier classifier to (i) a conventional single-tier classifier, and (ii) a confidence-based multi-tier classifier designed to reduce wake time. We evaluate our approaches using activity-labeled smartphone accelerometer data traces from 2 subjects over a total of 60 hours performing 7 different physical activities. Given an accuracy lower bound of 91%, our two-tier approach achieves the lowest wake times and is able to reduce the wake time percentage of the best single-tier classifier by 93.0% and 70.1% respectively for the 2 subjects.