Using Deep Learning and Smartphone for Automatic Detection of Fall and Daily Activities

Xiaodan Wu, Lingyu Cheng, Chao Hsien Chu, Jungyoon Kim

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Scopus citations

Abstract

The rapid growth of elderly population makes the health of the elderly one of the major social concerns. The elderly is often facing with several physical and mental healthcare related problems, among those, instance of fall and injuries ranked at the top. If people fall unexpectedly and without timely assistance, it is easy to cause irreparable harm. Therefore, how to automatically detect fall and alert for care/attention using advanced assisted technologies is a hot area of research. In this paper, we examine six machine learning-based methods and propose and carefully configure two novel deep learning-based architectures for fall detection. We compare the relative performance of these methods using an open source dataset, MobiAct, which was collected with four simulated fall types and nine daily living activities using smartphones. Our experimental results show that the proposed long short-term memory (LSTM) deep learning model is quite effective for the fall detection classification; its accuracy reaches 98.83%, the specificity is 99.38%, the sensitivity is 90.57% and the F1 score is 90.33%. These results are better than existing machine learning methods in all types of fall and most of daily activities.

Original languageEnglish (US)
Title of host publicationSmart Health - International Conference, ICSH 2019, Proceedings
EditorsHsinchun Chen, Daniel Zeng, Xiangbin Yan, Chunxiao Xing
PublisherSpringer
Pages61-74
Number of pages14
ISBN (Print)9783030344818
DOIs
StatePublished - 2019
Event7th International Conference for Smart Health, ICSH 2019 - Shenzhen, China
Duration: Jul 1 2019Jul 2 2019

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11924 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference7th International Conference for Smart Health, ICSH 2019
CountryChina
CityShenzhen
Period7/1/197/2/19

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint Dive into the research topics of 'Using Deep Learning and Smartphone for Automatic Detection of Fall and Daily Activities'. Together they form a unique fingerprint.

Cite this