Advanced

Group affiliation detection in a challenging environment

Jonsson, Håkan LU and Nugues, Pierre LU (2018) 141. p.507-512
Abstract
Social interaction sensing and indoor positioning using are widely researched. However, many use cases only need to determine proximity, and not the exact location. In this paper, we describe two methods to determine which meeting each user is participating in using proximity data collected from a challenging real-world office.
We show that the RSSI threshold approach to detecting proximity is not feasible due to the optimal RSSI range being very small. Instead, we achieve an F-score of 82% with a simple method, k-nearest neighbor classification, using data from the whole population. This method does not need any historical data or training, calibration to an environment, nor find a specific RSSI threshold. Finally, we present result... (More)
Social interaction sensing and indoor positioning using are widely researched. However, many use cases only need to determine proximity, and not the exact location. In this paper, we describe two methods to determine which meeting each user is participating in using proximity data collected from a challenging real-world office.
We show that the RSSI threshold approach to detecting proximity is not feasible due to the optimal RSSI range being very small. Instead, we achieve an F-score of 82% with a simple method, k-nearest neighbor classification, using data from the whole population. This method does not need any historical data or training, calibration to an environment, nor find a specific RSSI threshold. Finally, we present result from a user study with a prototype meeting application that identifies meeting participants, and advice on consequences of the above result for UI design. (Less)
Please use this url to cite or link to this publication:
author
organization
publishing date
type
Chapter in Book/Report/Conference proceeding
publication status
published
subject
host publication
The 5th International Symposium on Emerging Information, Communication and Networks
volume
141
pages
507 - 512
publisher
Elsevier Limited
external identifiers
  • scopus:85058321043
DOI
10.1016/j.procs.2018.10.134
project
Embedded Applications Software Engineering
language
English
LU publication?
yes
id
71597384-67f1-4e9c-a7e9-4f44ef37031d
date added to LUP
2018-09-13 15:35:07
date last changed
2019-03-08 03:05:29
@inproceedings{71597384-67f1-4e9c-a7e9-4f44ef37031d,
  abstract     = {Social interaction sensing and indoor positioning using are widely researched. However, many use cases only need to determine proximity, and not the exact location. In this paper, we describe two methods to determine which meeting each user is participating in using proximity data collected from a challenging real-world office.<br/>We show that the RSSI threshold approach to detecting proximity is not feasible due to the optimal RSSI range being very small. Instead, we achieve an F-score of 82% with a simple method, k-nearest neighbor classification, using data from the whole population. This method does not need any historical data or training, calibration to an environment, nor find a specific RSSI threshold. Finally, we present result from a user study with a prototype meeting application that identifies meeting participants, and advice on consequences of the above result for UI design.},
  author       = {Jonsson, Håkan and Nugues, Pierre},
  booktitle    = {The 5th International Symposium on Emerging Information, Communication and Networks},
  language     = {eng},
  pages        = {507--512},
  publisher    = {Elsevier Limited},
  title        = {Group affiliation detection in a challenging environment},
  url          = {http://dx.doi.org/10.1016/j.procs.2018.10.134},
  doi          = {10.1016/j.procs.2018.10.134},
  volume       = {141},
  year         = {2018},
}