A Neural Network Model Based on Co-occurrence Matrix for Fall Prediction

Masoud Hemmatpour; Renato Ferrero; Bartolomeo Montrucchio; Maurizio Rebaudengo

Wireless Mobile Communication and Healthcare. 6th International Conference, MobiHealth 2016, Milan, Italy, November 14-16, 2016, Proceedings

Research Article

A Neural Network Model Based on Co-occurrence Matrix for Fall Prediction

Download

328 downloads

Cite: BibTeX Plain Text

@INPROCEEDINGS{10.1007/978-3-319-58877-3_32,
    author={Masoud Hemmatpour and Renato Ferrero and Bartolomeo Montrucchio and Maurizio Rebaudengo},
    title={A Neural Network Model Based on Co-occurrence Matrix for Fall Prediction},
    proceedings={Wireless Mobile Communication and Healthcare. 6th International Conference, MobiHealth 2016, Milan, Italy, November 14-16, 2016, Proceedings},
    proceedings_a={MOBIHEALTH},
    year={2017},
    month={6},
    keywords={},
    doi={10.1007/978-3-319-58877-3_32}
}

Masoud Hemmatpour
Renato Ferrero
Bartolomeo Montrucchio
Maurizio Rebaudengo
Year: 2017
A Neural Network Model Based on Co-occurrence Matrix for Fall Prediction
MOBIHEALTH
Springer
DOI: 10.1007/978-3-319-58877-3_32

Masoud Hemmatpour¹^,*, Renato Ferrero¹^,*, Bartolomeo Montrucchio¹^,*, Maurizio Rebaudengo¹^,*

1: Politecnico di Torino

*Contact email: masoud.hemmatpour@polito.it, renato.ferrero@polito.it, bartolomeo.montrucchio@polito.it, maurizio.rebaudengo@polito.it

Abstract

Fall avoidance systems reduce injuries due to unintentional falls, but most of them are fall detections that activate an alarm after the fall occurrence. Since predicting a fall is the most promising approach to avoid a fall injury, this study proposes a method based on new features and multilayer perception that outperforms state-of-the-art approaches. Since accelerometer and gyroscope embedded in a smartphone are recognized to be precise enough to be used in fall avoidance systems, they have been exploited in an experimental analysis in order to compare the proposal with state-of-the-art approaches. The results have shown that the proposed approach improves the accuracy from 83% to 90%.

Published: 2017-06-05
Appears in: SpringerLink

: http://dx.doi.org/10.1007/978-3-319-58877-3_32