An investigation into smartphone based weakly supervised activity recognition systems

Research output: Contribution to journalArticle

Abstract

With smart-devices becoming increasingly more commonplace, methods of capturing an individual’s activities are becoming feasible. This is more generally performed through questionnaires or within unnatural environments bringing drawbacks in accuracy or requiring impractical conditions. This paper presents a simpler method of data collection which reduces the complications of typical activity data collection by collecting labels directly from a user. Instead of capturing activity beginning and end times, user requests are made at time intervals and labels are populated to feature vectors. These methods can provide a simpler method of data collection and could provide a solution to the annotation problem within activity recognition
LanguageEnglish
Article number56
Pages45-56
Number of pages12
JournalPervasive and Mobile Computing
Volume56
Early online date29 Mar 2019
DOIs
Publication statusPublished - 1 May 2019

Fingerprint

Activity Recognition
Smartphones
Labels
Complications
Feature Vector
Questionnaire
Annotation
Interval

Keywords

  • Gaussian means
  • Activity recognition
  • Multiple-instance learning
  • Random forest
  • Support vector machines

Cite this

@article{71bb456aef9d4d368a0abe83cd1d08f5,
title = "An investigation into smartphone based weakly supervised activity recognition systems",
abstract = "With smart-devices becoming increasingly more commonplace, methods of capturing an individual’s activities are becoming feasible. This is more generally performed through questionnaires or within unnatural environments bringing drawbacks in accuracy or requiring impractical conditions. This paper presents a simpler method of data collection which reduces the complications of typical activity data collection by collecting labels directly from a user. Instead of capturing activity beginning and end times, user requests are made at time intervals and labels are populated to feature vectors. These methods can provide a simpler method of data collection and could provide a solution to the annotation problem within activity recognition",
keywords = "Gaussian means, Activity recognition, Multiple-instance learning, Random forest, Support vector machines",
author = "William Duffy and Kevin Curran and Daniel Kelly and Tom Lunney",
year = "2019",
month = "5",
day = "1",
doi = "10.1016/j.pmcj.2019.03.005",
language = "English",
volume = "56",
pages = "45--56",
journal = "Pervasive and Mobile Computing",
issn = "1574-1192",
publisher = "Elsevier",

}

TY - JOUR

T1 - An investigation into smartphone based weakly supervised activity recognition systems

AU - Duffy, William

AU - Curran, Kevin

AU - Kelly, Daniel

AU - Lunney, Tom

PY - 2019/5/1

Y1 - 2019/5/1

N2 - With smart-devices becoming increasingly more commonplace, methods of capturing an individual’s activities are becoming feasible. This is more generally performed through questionnaires or within unnatural environments bringing drawbacks in accuracy or requiring impractical conditions. This paper presents a simpler method of data collection which reduces the complications of typical activity data collection by collecting labels directly from a user. Instead of capturing activity beginning and end times, user requests are made at time intervals and labels are populated to feature vectors. These methods can provide a simpler method of data collection and could provide a solution to the annotation problem within activity recognition

AB - With smart-devices becoming increasingly more commonplace, methods of capturing an individual’s activities are becoming feasible. This is more generally performed through questionnaires or within unnatural environments bringing drawbacks in accuracy or requiring impractical conditions. This paper presents a simpler method of data collection which reduces the complications of typical activity data collection by collecting labels directly from a user. Instead of capturing activity beginning and end times, user requests are made at time intervals and labels are populated to feature vectors. These methods can provide a simpler method of data collection and could provide a solution to the annotation problem within activity recognition

KW - Gaussian means

KW - Activity recognition

KW - Multiple-instance learning

KW - Random forest

KW - Support vector machines

UR - http://www.scopus.com/inward/record.url?scp=85063689626&partnerID=8YFLogxK

U2 - 10.1016/j.pmcj.2019.03.005

DO - 10.1016/j.pmcj.2019.03.005

M3 - Article

VL - 56

SP - 45

EP - 56

JO - Pervasive and Mobile Computing

T2 - Pervasive and Mobile Computing

JF - Pervasive and Mobile Computing

SN - 1574-1192

M1 - 56

ER -