Partial Matching Estimation Method of Walking Trajectories for Generating Indoor Pedestrian Networks

We propose a method that estimates the partial matching of a walking trajectory for integrate indoor walking trajectories and generate an indoor pedestrian network. Building structure information is necessary for the realization of indoor location information services (indoor LBSs). In indoor pedest...

Full description

Saved in:

Bibliographic Details
Published in:	2018 Eleventh International Conference on Mobile Computing and Ubiquitous Network (ICMU) pp. 1 - 6
Main Authors:	Sugimoto, Sou, Ito, Nobuyuki, Naito, Katsuhiro, Chujo, Naoya, Mizuno, Tadanori, Kaji, Katsuhiko
Format:	Conference Proceeding
Language:	English
Published:	IPSJ 01-10-2018
Subjects:	Buildings Estimation Legged locomotion Machine learning Sensors Trajectory Wireless fidelity
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	We propose a method that estimates the partial matching of a walking trajectory for integrate indoor walking trajectories and generate an indoor pedestrian network. Building structure information is necessary for the realization of indoor location information services (indoor LBSs). In indoor pedestrian networks and building structure information, since the shortest route calculation can be directly applied, the route from an arbitrary position to a destination can be obtained by a computer. Therefore, navigation and recommending a healthier route is possible by route recommendation. However, the generation of the indoor pedestrian network is costly, so it does not necessarily exist. Therefore, as a low-cost generation method, our final goal is to automatically generate an indoor pedestrian network structure by collecting the walking sensing data from many users and integrating the estimated 3-D walking trajectories. In the integration of such trajectories, such partial matching places as walking in the same passage are used as clues. To estimate a partially matched place, we use a stable zone (a stable walking zone) and linear walking. Since the stable walking zone detects the continuation of the state with only a slight change, the estimation accuracy becomes higher than the right and left turns. We estimate the correspondence among stable walking zones by machine learning using the walking time, the walking distance, the height, and Wi-Fi information. For machine learning, we use a support vector machine (SVM). From the results of an evaluation experiment with HASC-IPSC, an indoor walking sensing corpus, we obtained a result where the F value was 0.81.
DOI:	10.23919/ICMU.2018.8653585