Indexed by:会议论文

Date of Publication:2017-07-22

Included Journals:EI

Page Number:877-881

Abstract:To solve the traditional BBS problems, the present study investigates smartphone-based activities recognition and a system is designed in this paper. In this system, smartphones fixed with workers collect acceleration and angle data, and then run a classification model to label the activity as safe or dangerous, ending with a notification or a warning as the feedback to workers and managers. Taking an example of climbing activities in aerial work to validate its feasibility, an experimental study was carried out. Two types of classifiers were evaluated by five indexes and an accuracy of 99% was obtained with both of them, while support vector machine run with less time and less memory consumption. The results from preliminary studies had shown the potential of the proposed method.