.png){kind=icon}
.png){kind=icon}
Return
Indoor Semantic-Rich Link-Node Model Construction Using Crowdsourced Trajectories From Smartphones
.png){kind=icon}
PRE
AI
S
M
DOI：10.1109/JSEN.2019.2933746.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Link-node models have recently emerged as promising indoor positioning techniques for indoor location-based Internet of Things (IoT) applications. However, the link-node models reported in the literature are commonly extracted from indoor maps that are both coverage-limited and costly. Furthermore, the extraction process is also time-consuming and error-prone. In this work, a novel method is developed to automatically construct semantic-rich indoor link-node models from the crowdsourced trajectory data collected by smartphone sensors without requiring maps and additional devices. More specifically, the pedestrian trajectories are first obtained using inertial sensors built in smartphones. After that, indoor link-node models are constructed by exploiting the pedestrian activity information derived from human activity recognition (HAR) and structural nodes (e.g. doors and elevators) extracted from the trajectory. Furthermore, a trajectory similarity measure in terms of both semantics and geometry is developed to identify similar segments from the crowdsourced trajectories. Capitalizing on the similarity measure, a short-range trajectory clustering method is proposed to improve the accuracy of the indoor link-node model. In addition to indoor positioning, the resulting model can provide structural information of an indoor environment as well as semantic information about pedestrians and the environment, which is particularly useful for advanced IoT applications and services. Extensive field measurements demonstrate that the resulting indoor link-node models are of about one-meter accuracy in most experiments.
Keywords:
Crowdsourcing
link-node model
indoor trajectory
indoor positioning
activity recognition
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
4.5
Papers:
2.1W
Citations:
7.3W
Organization
Cited Papers