WLAN RSS-Based Fingerprinting for Indoor Localization: A Machine Learning Inspired Bag-of-Features Approach

Location-based services have permeated Smart academic institutions, enhancing the quality of higher education. Position information of people and objects can predict different potential requirements and provide relevant services to meet those needs. Indoor positioning system (IPS) research has attai...

Full description

Saved in:
Bibliographic Details
Published in:Sensors (Basel, Switzerland) Switzerland), 2022-07, Vol.22 (14), p.5236
Main Authors: Altaf Khattak, Sohaib Bin, Fawad, Nasralla, Moustafa M., Esmail, Maged Abdullah, Mostafa, Hala, Jia, Min
Format: Article
Language:English
Subjects:
Accuracy
Algorithms
Classification
Clustering
College campuses
Fingerprinting
Fingerprints
Global positioning systems
GPS
higher education
Indoor environments
indoor positioning system
learning environment
Localization
Location based services
Machine learning
Methods
Neural networks
Propagation
Radio transmission
WLAN fingerprinting
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Location-based services have permeated Smart academic institutions, enhancing the quality of higher education. Position information of people and objects can predict different potential requirements and provide relevant services to meet those needs. Indoor positioning system (IPS) research has attained robust location-based services in complex indoor structures. Unforeseeable propagation loss in complex indoor environments results in poor localization accuracy of the system. Various IPSs have been developed based on fingerprinting to precisely locate an object even in the presence of indoor artifacts such as multipath and unpredictable radio propagation losses. However, such methods are deleteriously affected by the vulnerability of fingerprint matching frameworks. In this paper, we propose a novel machine learning framework consisting of Bag-of-Features and followed by a k-nearest neighbor classifier to categorize the final features into their respective geographical coordinate data. BoF calculates the vocabulary set using k-mean clustering, where the frequency of the vocabulary in the raw fingerprint data represents the robust final features that improve localization accuracy. Experimental results from simulation-based indoor scenarios and real-time experiments demonstrate that the proposed framework outperforms previously developed models.
ISSN:1424-8220
1424-8220
DOI:10.3390/s22145236