Vision-based toddler physical activity recognition using deep learning

Human activity recognition (HAR) is a system for understanding human movements and behaviour. It has been applied in many fields such as video surveillance, behaviour analysis, and human-computer interaction. The state-of-the-art studies on HAR generally focus their attention on public dataset which...

Full description

Bibliographic Details
Main Authors: Norasyikin, Fadilah, Mohd Zamri, Ibrahim, Rosdiyana, Samad
Format: Conference or Workshop Item
Language:English
English
Published: Institution of Engineering and Technology 2022
Subjects:
Online Access:http://umpir.ump.edu.my/id/eprint/42025/
http://umpir.ump.edu.my/id/eprint/42025/1/Vision-based%20toddler%20physical%20activity%20recognition.pdf
http://umpir.ump.edu.my/id/eprint/42025/2/Vision-based%20toddler%20physical%20activity%20recognition%20using%20deep%20learning_ABS.pdf
_version_ 1848826498604072960
author Norasyikin, Fadilah
Mohd Zamri, Ibrahim
Rosdiyana, Samad
author_facet Norasyikin, Fadilah
Mohd Zamri, Ibrahim
Rosdiyana, Samad
author_sort Norasyikin, Fadilah
building UMP Institutional Repository
collection Online Access
description Human activity recognition (HAR) is a system for understanding human movements and behaviour. It has been applied in many fields such as video surveillance, behaviour analysis, and human-computer interaction. The state-of-the-art studies on HAR generally focus their attention on public dataset which mostly consist of adults as their subjects. Research on HAR for children especially toddlers is important to facilitate their surveillance by monitoring their activities automatically. Since toddlers possess different anatomical proportions than adults, their unusual movements can be a challenge to infer. In this paper, a vision-based deep learning HAR system for toddlers was developed based on skeleton features. Videos of toddlers' activities in a day-care were obtained through different public sources. 2D skeleton data were then extracted from every frame of these videos using a pre-trained deep learning network. These skeleton data were trained on LSTM and fully connected network to infer the toddler's activities. Results showed that this proposed framework managed to achieve 75% accuracies for three toddlers' activities which are jumping, sitting, and standing.
first_indexed 2025-11-15T03:45:47Z
format Conference or Workshop Item
id ump-42025
institution Universiti Malaysia Pahang
institution_category Local University
language English
English
last_indexed 2025-11-15T03:45:47Z
publishDate 2022
publisher Institution of Engineering and Technology
recordtype eprints
repository_type Digital Repository
spelling ump-420252024-09-30T04:38:49Z http://umpir.ump.edu.my/id/eprint/42025/ Vision-based toddler physical activity recognition using deep learning Norasyikin, Fadilah Mohd Zamri, Ibrahim Rosdiyana, Samad T Technology (General) TA Engineering (General). Civil engineering (General) TK Electrical engineering. Electronics Nuclear engineering Human activity recognition (HAR) is a system for understanding human movements and behaviour. It has been applied in many fields such as video surveillance, behaviour analysis, and human-computer interaction. The state-of-the-art studies on HAR generally focus their attention on public dataset which mostly consist of adults as their subjects. Research on HAR for children especially toddlers is important to facilitate their surveillance by monitoring their activities automatically. Since toddlers possess different anatomical proportions than adults, their unusual movements can be a challenge to infer. In this paper, a vision-based deep learning HAR system for toddlers was developed based on skeleton features. Videos of toddlers' activities in a day-care were obtained through different public sources. 2D skeleton data were then extracted from every frame of these videos using a pre-trained deep learning network. These skeleton data were trained on LSTM and fully connected network to infer the toddler's activities. Results showed that this proposed framework managed to achieve 75% accuracies for three toddlers' activities which are jumping, sitting, and standing. Institution of Engineering and Technology 2022 Conference or Workshop Item PeerReviewed pdf en http://umpir.ump.edu.my/id/eprint/42025/1/Vision-based%20toddler%20physical%20activity%20recognition.pdf pdf en http://umpir.ump.edu.my/id/eprint/42025/2/Vision-based%20toddler%20physical%20activity%20recognition%20using%20deep%20learning_ABS.pdf Norasyikin, Fadilah and Mohd Zamri, Ibrahim and Rosdiyana, Samad (2022) Vision-based toddler physical activity recognition using deep learning. In: IET Conference Proceedings. 2022 Engineering Technology International Conference, ETIC 2022 , 7 - 8 September 2022 , Kuantan, Virtual. pp. 377-383., 2022 (22). ISSN 2732-4494 ISBN 978-183953782-0 (Published) https://doi.org/10.1049/icp.2022.2647
spellingShingle T Technology (General)
TA Engineering (General). Civil engineering (General)
TK Electrical engineering. Electronics Nuclear engineering
Norasyikin, Fadilah
Mohd Zamri, Ibrahim
Rosdiyana, Samad
Vision-based toddler physical activity recognition using deep learning
title Vision-based toddler physical activity recognition using deep learning
title_full Vision-based toddler physical activity recognition using deep learning
title_fullStr Vision-based toddler physical activity recognition using deep learning
title_full_unstemmed Vision-based toddler physical activity recognition using deep learning
title_short Vision-based toddler physical activity recognition using deep learning
title_sort vision-based toddler physical activity recognition using deep learning
topic T Technology (General)
TA Engineering (General). Civil engineering (General)
TK Electrical engineering. Electronics Nuclear engineering
url http://umpir.ump.edu.my/id/eprint/42025/
http://umpir.ump.edu.my/id/eprint/42025/
http://umpir.ump.edu.my/id/eprint/42025/1/Vision-based%20toddler%20physical%20activity%20recognition.pdf
http://umpir.ump.edu.my/id/eprint/42025/2/Vision-based%20toddler%20physical%20activity%20recognition%20using%20deep%20learning_ABS.pdf