Fall detection algorithm based on accelerometer and gyroscope sensor data using Recurrent Neural Networks

In our daily life activity, sometimes there is a chance of getting fall unintentionally. Unintentional falls are dangerous to health and may cause a serious problem, especially for elderly people whose have a higher probability of getting fall. In this paper, we develop an algorithm to distinguish f...

Full description

Saved in:
Bibliographic Details
Published in:IOP conference series. Earth and environmental science Vol. 258; no. 1; pp. 12035 - 12043
Main Authors: Wayan Wiprayoga Wisesa, I, Mahardika, Genggam
Format: Journal Article
Language:English
Published: Bristol IOP Publishing 10.05.2019
Subjects:
Accelerometers
Algorithms
Datasets
Fall detection
Gyroscopes
Neural networks
Older people
Recurrent neural networks
Sensors
Wearable technology
ISSN:1755-1307, 1755-1315, 1755-1315
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In our daily life activity, sometimes there is a chance of getting fall unintentionally. Unintentional falls are dangerous to health and may cause a serious problem, especially for elderly people whose have a higher probability of getting fall. In this paper, we develop an algorithm to distinguish falls from other activity daily living (ADL) based on accelerometer and gyroscope sensor data embedded on a wearable device. Several fall detection algorithms exist, with the majority are using rule-based algorithm. We take advantage of recurrent neural networks (RNN) as a tool for analyzing sequence time series data from sensors. The experiment was conducted using publicly available dataset UMA FALL ADL from Universidad de Málaga. The dataset consists of several recorded sensor-tag data, consisting of accelerometer, gyroscope and magnetometer sensor, representing the daily activity of several subjects including falls. Based on our experiment, we found that our algorithm yields a good result distinguishing fall from ADL.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1755-1307
1755-1315
1755-1315
DOI:10.1088/1755-1315/258/1/012035