Falling Angel – A Wrist Worn Fall Detection System Using K-NN Algorithm

Hamidur Rahman; Johan Sandberg; Lennart Eriksson; Mohammad Heidari; Jan Arwald; Peter Eriksson; Shahina Begum; Maria Lindén; Mobyen Ahmed

Internet of Things Technologies for HealthCare. Third International Conference, HealthyIoT 2016, Västerås, Sweden, October 18-19, 2016, Revised Selected Papers

Research Article

Falling Angel – A Wrist Worn Fall Detection System Using K-NN Algorithm

Download

231 downloads

Cite: BibTeX Plain Text

@INPROCEEDINGS{10.1007/978-3-319-51234-1_25,
    author={Hamidur Rahman and Johan Sandberg and Lennart Eriksson and Mohammad Heidari and Jan Arwald and Peter Eriksson and Shahina Begum and Maria Lind\^{e}n and Mobyen Ahmed},
    title={Falling Angel -- A Wrist Worn Fall Detection System Using K-NN Algorithm},
    proceedings={Internet of Things Technologies for HealthCare. Third International Conference, HealthyIoT 2016, V\aa{}ster\ae{}s, Sweden, October 18-19, 2016, Revised Selected Papers},
    proceedings_a={HEALTHYIOT},
    year={2017},
    month={1},
    keywords={Fall detection Angel device k-Nearest Neighbor},
    doi={10.1007/978-3-319-51234-1_25}
}

Hamidur Rahman
Johan Sandberg
Lennart Eriksson
Mohammad Heidari
Jan Arwald
Peter Eriksson
Shahina Begum
Maria Lindén
Mobyen Ahmed
Year: 2017
Falling Angel – A Wrist Worn Fall Detection System Using K-NN Algorithm
HEALTHYIOT
Springer
DOI: 10.1007/978-3-319-51234-1_25

Hamidur Rahman¹^,*, Johan Sandberg¹^,*, Lennart Eriksson¹^,*, Mohammad Heidari¹^,*, Jan Arwald²^,*, Peter Eriksson²^,*, Shahina Begum¹^,*, Maria Lindén¹^,*, Mobyen Ahmed¹^,*

1: Mälardalen University
2: Exformation AB

*Contact email: hamidur.rahman@mdh.se, johan.sandberg@mdh.se, lennart.eriksson@mdh.se, mohammad.heidari@mdh.se, jan.arwald@exformation.com, peter.eriksson@exformation.com, shahina.begum@mdh.se, maria.linden@mdh.se, mobyen.ahmed@mdh.se

Abstract

A wrist worn fall detection system has been developed where the accelerometer data from an angel sensor is analyzed by a two-layered algorithm in an android phone. Here, the first layer uses a threshold to find potential falls and if the thresholds are met, then in the second layer a machine learning i.e., k-Nearest Neighbor (k-NN) algorithm analyses the data to differentiate it from Activities of Daily Living (ADL) in order to filter out false positives. The final result of this project using the k-NN algorithm provides a classification sensitivity of 96.4%. Here, the acquired sensitivity is 88.1% for the fall detection and the specificity for ADL is 98.1%.

Keywords: Fall detection Angel device k-Nearest Neighbor

Published: 2017-01-20
Appears in: SpringerLink

: http://dx.doi.org/10.1007/978-3-319-51234-1_25