CoChat

A Comprehensive Study of Activity Recognition Using Accelerometers

Open access

Published version of a preprint

The DOI matches its Crossref record; 98 of the 99 references it cites were found in a registry or an index, and 1 are not checked yet.

Abstract

This paper serves as a survey and empirical evaluation of the state-of-the-art in activity recognition methods using accelerometers. The paper is particularly focused on long-term activity recognition in real-world settings. In these environments, data collection is not a trivial matter; thus, there are performance trade-offs between prediction accuracy, which is not the sole system objective, and keeping the maintenance overhead at minimum levels. We examine research that has focused on the selection of activities, the features that are extracted from the accelerometer data, the segmentation of the time-series data, the locations of accelerometers, the selection and configuration trade-offs, the test/retest reliability, and the generalisation performance. Furthermore, we study these questions from an experimental platform and show, somewhat surprisingly, that many disparate experimental configurations yield comparable predictive performance on testing data. Our understanding of these results is that the experimental setup directly and indirectly defines a pathway for context to be delivered to the classifier, and that, in some settings, certain configurations are more optimal than alternatives. We conclude by identifying how the main results of this work can be used in practice, specifically in experimental configurations in challenging experimental conditions.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.3390/informatics5020027

Cites

98 of 99 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 This figure illustrates the main topics that are covered by this paper and lists the relevant sections and tables where each concept is discussed and introduced.
Figure 1 This figure illustrates the main topics that are covered by this paper and lists the relevant sections and tables where each concept is discussed and introduced.CC-BY · From the PDF
Image 2
CC-BY · From the PDF
Figure 3 The raw data ( top ) at 50 Hz, and resampled 5 Hz ( bottom ). Notice that the high-frequency aspects of the accelerometer data are removed with lower sampling frequencies. Samples are marked with × symbols. The x , y and z axes are depicted in blue, orange and green respectively.
Figure 3 The raw data ( top ) at 50 Hz, and resampled 5 Hz ( bottom ). Notice that the high-frequency aspects of the accelerometer data are removed with lower sampling frequencies. Samples are marked with × symbols. The x , y and z axes are depicted in blue, orange and green respectively.CC-BY · From the PDF
Image 4
CC-BY · From the PDF
Figure 5 Calibrated accelerometer readings ( upper ) that were derived from raw (uncalibrated) accelerometer values ( lower ). The intervals shaded in red were used to perform calibration.
Figure 5 Calibrated accelerometer readings ( upper ) that were derived from raw (uncalibrated) accelerometer values ( lower ). The intervals shaded in red were used to perform calibration.CC-BY · From the PDF
Figure 6 Visualisation of the classification performance on the HAR dataset for stat features ( left ), dict features ( middle ) and ecdf features ( right ) over varying window lengths (rows) and sampling rates (columns). Darker red colours indicate better performance.
Figure 6 Visualisation of the classification performance on the HAR dataset for stat features ( left ), dict features ( middle ) and ecdf features ( right ) over varying window lengths (rows) and sampling rates (columns). Darker red colours indicate better performance.CC-BY · From the PDF
Figure 7 Classification performance obtained by Logistic Regression-Conditional Random Field (LR-CRF) on the HAR dataset. ( a ) LR-CRF classification performance over the three feature categories considered. ( b ) Difference between LR-CRF and LR classification performance. Red indicates LR-CRF outperforms the basic LR model.
Figure 7 Classification performance obtained by Logistic Regression-Conditional Random Field (LR-CRF) on the HAR dataset. ( a ) LR-CRF classification performance over the three feature categories considered. ( b ) Difference between LR-CRF and LR classification performance. Red indicates LR-CRF outperforms the basic LR model.CC-BY · From the PDF
Figure 8 Improvement in classification accuracy obtained by incorporating structure on the classification task with Conditional Random Fields (CRFs).
Figure 8 Improvement in classification accuracy obtained by incorporating structure on the classification task with Conditional Random Fields (CRFs).CC-BY · From the PDF
Figure 9 Box plots over changes to performance in accuracy when using Conditional Random Field (CRF) models to capture sequential dynamics. Results on HAR are shown on the top row, Results on PAMAP2 in the middle row, and USCHAD on the bottom row. The first column presents the results of Logistic Regression (LR) CRFs, the middle column on Multi-layer Perceptron (MLP) CRFs and the final column on Random Forest CRFs.
Figure 9 Box plots over changes to performance in accuracy when using Conditional Random Field (CRF) models to capture sequential dynamics. Results on HAR are shown on the top row, Results on PAMAP2 in the middle row, and USCHAD on the bottom row. The first column presents the results of Logistic Regression (LR) CRFs, the middle column on Multi-layer Perceptron (MLP) CRFs and the final column on Random Forest CRFs.CC-BY · From the PDF
Figure 10 Mean aggregated change in performance when aggregated over datasets ( a ) and features ( b ). ( a ) Mean improvement of Conditional Random Fields (CRFs) aggregated over datasets. ( b ) Mean improvement of Conditional Random Fields (CRFs) aggregated over features.
Figure 10 Mean aggregated change in performance when aggregated over datasets ( a ) and features ( b ). ( a ) Mean improvement of Conditional Random Fields (CRFs) aggregated over datasets. ( b ) Mean improvement of Conditional Random Fields (CRFs) aggregated over features.CC-BY · From the PDF
Figure 11 Contingency tables of activities recognised on the HAR dataset with Linear Regression (LR) ( a ) and LR with Conditional Random Fields (CRFs) ( b ) with a window length of 3 s, and a sampling rate of 20 Hz. Rows indicate the ground truth and columns indicate predictions. ( a ) Contingency table for LR. ( b ) Contingency table for LR + CRF.
Figure 11 Contingency tables of activities recognised on the HAR dataset with Linear Regression (LR) ( a ) and LR with Conditional Random Fields (CRFs) ( b ) with a window length of 3 s, and a sampling rate of 20 Hz. Rows indicate the ground truth and columns indicate predictions. ( a ) Contingency table for LR. ( b ) Contingency table for LR + CRF.CC-BY · From the PDF
Figure A1. Classification performance on the HAR dataset with all iid feature configurations.
Figure A1. Classification performance on the HAR dataset with all iid feature configurations.CC-BY · From the PDF
Figure A2. Classification performance on the PAMAP2 dataset with all iid feature configurations.
Figure A2. Classification performance on the PAMAP2 dataset with all iid feature configurations.CC-BY · From the PDF
Figure A3. Classification performance on the USCHAD dataset with all iid feature configurations.
Figure A3. Classification performance on the USCHAD dataset with all iid feature configurations.CC-BY · From the PDF

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 99 references

Research works printed but not linked: 3

  1. Shawe-Taylor, J., and Cristianini, N. (2000). Support Vector Machines, Cambridge University Press.
  2. Goodfellow, I., Bengio, Y., Courville, A., and Bengio, Y. (2016). Deep Learning, MIT Press.
  3. Tipping, 2001, Sparse Bayesian Learning and the Relevance Vector Machine, J. Mach. Learn. Res., 1, 211

Cited by

See all 143

Metadata

Authors
Twomey, N., Diethe, T., Fafoutis, X., Elsts, A., McConville, R., Flach, P., & Craddock, I.
Affiliations
University of Bristol (GB)
Corresponding author
  • Twomey, Niall
  • Diethe, Tom
Venue
Informatics
Publisher
MDPI AG
Published
2018-05-30
Volume
5
Issue
2
Pages
27
ISSN
2227-9709 (ISSN-L)
OpenAlex
W2790414609
Cited by
143
References
95
Citations by year
2026: 1 · 2025: 9 · 2024: 20 · 2023: 10 · 2022: 19 · 2021: 30 · 2020: 24 · 2019: 26 · 2018: 4
FWCI
10.50
Type
journal article
Language
en
Open access
Locations
  • Informatics (publishedVersion, cc-by)
  • Preprints.org (acceptedVersion, cc-by)
  • Explore Bristol Research (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • Bristol Research (University of Bristol) (submittedVersion, cc-by)
Funders
  • Horizon 2020 Framework Programme
  • Medical Research Council (MC_PC_16029)
  • Engineering and Physical Sciences Research Council (EP/K031910/1, H2020)
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Context-Aware Activity Recognition Systems
  • Health Sciences › Medicine › Public Health, Environmental and Occupational Health › Physical Activity and Health
  • Social Sciences › Transportation › Human Mobility and Location-Based Analysis
Keywords
activity recognition, accelerometer, wearable sensors, accelerometer placement, time series segmentation, feature extraction, activity selection, test-retest reliability, generalization, field deployment

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: A Comprehensive Study of Activity Recognition Using Accelerometers. https://app.cochat.ai/papers/10.3390/informatics5020027
The paper
Twomey, N., Diethe, T., Fafoutis, X., Elsts, A., McConville, R., Flach, P., & Craddock, I. (2018). A Comprehensive Study of Activity Recognition Using Accelerometers. Informatics. https://doi.org/10.3390/informatics5020027

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.