Cross-position activity recognition with stratified transfer learning

Human activity recognition (HAR) aims to recognize the activities of daily living by utilizing the sensors attached to different body parts. HAR relies on the machine learning models trained using sufficient activity data. However, when the labels from a certain body position (i.e. target domain) ar...

Full description

Saved in:
Bibliographic Details
Main Authors: Chen, Yiqiang, Wang, Jindong, Huang, Meiyu, Yu, Han
Other Authors: School of Computer Science and Engineering
Format: Article
Language:English
Published: 2020
Subjects:
Online Access:https://hdl.handle.net/10356/143186
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
id sg-ntu-dr.10356-143186
record_format dspace
spelling sg-ntu-dr.10356-1431862021-02-04T07:11:07Z Cross-position activity recognition with stratified transfer learning Chen, Yiqiang Wang, Jindong Huang, Meiyu Yu, Han School of Computer Science and Engineering Engineering::Computer science and engineering Activity Recognition Transfer Learning Human activity recognition (HAR) aims to recognize the activities of daily living by utilizing the sensors attached to different body parts. HAR relies on the machine learning models trained using sufficient activity data. However, when the labels from a certain body position (i.e. target domain) are missing, how to leverage the data from other positions (i.e. source domain) to help recognize the activities of this position? This problem can be divided into two steps. Firstly, when there are several source domains available, it is often difficult to select the most similar source domain to the target domain. Secondly, with the selected source domain, we need to perform accurate knowledge transfer between domains in order to recognize the activities on the target domain. Existing methods only learn the global distance between domains while ignoring the local property. In this paper, we propose a Stratified Transfer Learning (STL) framework to perform both source domain selection and activity transfer. STL is based on our proposed Stratified distance to capture the local property of domains. STL consists of two components: 1) Stratified Domain Selection (STL-SDS), which can select the most similar source domain to the target domain; and 2) Stratified Activity Transfer (STL-SAT), which is able to perform accurate knowledge transfer. Extensive experiments on three public activity recognition datasets demonstrate the superiority of STL. Accepted version 2020-08-11T09:27:00Z 2020-08-11T09:27:00Z 2019 Journal Article Chen, Y., Wang, J., Huang, M., & Yu, H. (2019). Cross-position activity recognition with stratified transfer learning. Pervasive and Mobile Computing, 57, 1-13. doi:10.1016/j.pmcj.2019.04.004 1574-1192 https://hdl.handle.net/10356/143186 10.1016/j.pmcj.2019.04.004 2-s2.0-85064479331 57 1 13 en Pervasive and Mobile Computing © 2019 Elsevier B.V. All rights reserved. This paper was published in Pervasive and Mobile Computing and is made available with permission of Elsevier B.V. application/pdf
institution Nanyang Technological University
building NTU Library
continent Asia
country Singapore
Singapore
content_provider NTU Library
collection DR-NTU
language English
topic Engineering::Computer science and engineering
Activity Recognition
Transfer Learning
spellingShingle Engineering::Computer science and engineering
Activity Recognition
Transfer Learning
Chen, Yiqiang
Wang, Jindong
Huang, Meiyu
Yu, Han
Cross-position activity recognition with stratified transfer learning
description Human activity recognition (HAR) aims to recognize the activities of daily living by utilizing the sensors attached to different body parts. HAR relies on the machine learning models trained using sufficient activity data. However, when the labels from a certain body position (i.e. target domain) are missing, how to leverage the data from other positions (i.e. source domain) to help recognize the activities of this position? This problem can be divided into two steps. Firstly, when there are several source domains available, it is often difficult to select the most similar source domain to the target domain. Secondly, with the selected source domain, we need to perform accurate knowledge transfer between domains in order to recognize the activities on the target domain. Existing methods only learn the global distance between domains while ignoring the local property. In this paper, we propose a Stratified Transfer Learning (STL) framework to perform both source domain selection and activity transfer. STL is based on our proposed Stratified distance to capture the local property of domains. STL consists of two components: 1) Stratified Domain Selection (STL-SDS), which can select the most similar source domain to the target domain; and 2) Stratified Activity Transfer (STL-SAT), which is able to perform accurate knowledge transfer. Extensive experiments on three public activity recognition datasets demonstrate the superiority of STL.
author2 School of Computer Science and Engineering
author_facet School of Computer Science and Engineering
Chen, Yiqiang
Wang, Jindong
Huang, Meiyu
Yu, Han
format Article
author Chen, Yiqiang
Wang, Jindong
Huang, Meiyu
Yu, Han
author_sort Chen, Yiqiang
title Cross-position activity recognition with stratified transfer learning
title_short Cross-position activity recognition with stratified transfer learning
title_full Cross-position activity recognition with stratified transfer learning
title_fullStr Cross-position activity recognition with stratified transfer learning
title_full_unstemmed Cross-position activity recognition with stratified transfer learning
title_sort cross-position activity recognition with stratified transfer learning
publishDate 2020
url https://hdl.handle.net/10356/143186
_version_ 1692012961226293248