Wearable Sensor-Based Human Activity Recognition with Transformer Model
Open accessPublished version of a preprint
The DOI matches its Crossref record; 43 of the 43 references it cites were found in a registry or an index.
Abstract
Computing devices that can recognize various human activities or movements can be used to assist people in healthcare, sports, or human-robot interaction. Readily available data for this purpose can be obtained from the accelerometer and the gyroscope built into everyday smartphones. Effective classification of real-time activity data is, therefore, actively pursued using various machine learning methods. In this study, the transformer model, a deep learning neural network model developed primarily for the natural language processing and vision tasks, was adapted for a time-series analysis of motion signals. The self-attention mechanism inherent in the transformer, which expresses individual dependencies between signal values within a time series, can match the performance of state-of-the-art convolutional neural networks with long short-term memory. The performance of the proposed adapted transformer method was tested on the largest available public dataset of smartphone motion sensor data covering a wide range of activities, and obtained an average identification accuracy of 99.2% as compared with 89.67% achieved on the same data by a conventional machine learning method. The results suggest the expected future relevance of the transformer model for human activity recognition.
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.
Integrity signals
Registry
Cites
Statistics
Figures









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.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“This parallel design minimizes redundancy and improves efficiency relative to the sequential attention in [24] or the resource-intensive Transformer-based attention in [23].”
From Adaptive temporal attention mechanism and hybrid deep CNN model for wearable sensor-based human activity recognition (2025) · cites this paper as [23]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Previous studies, including [10, 18, 23, 24], have amalgamated convolutional neural networks (CNN), recurrent neural networks (RNN), and attention mechanisms;”
From Adaptive temporal attention mechanism and hybrid deep CNN model for wearable sensor-based human activity recognition (2025) · cites this paper as [23]
Experimental — accuracy not yet measuredIn the citing paper
“Moreover, replacing LSTM with BiGRU and using an optimized attention mechanism reduces computational cost, making CNNd-TAm more efficient than Transformer-based [23] and hierarchical models [19], and suitable for deployment on resource-constrained wearable devices without compromising accuracy.”
From Adaptive temporal attention mechanism and hybrid deep CNN model for wearable sensor-based human activity recognition (2025) · cites this paper as [23]
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
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404 · cited 6×
“The entry determining the position of the features within the series is also added to the network input, because the transformer does not know the order of the features, for example, within the processed sentence [34]:”
BackgroundMethodResultInfluentialExploiting Generative AI to Scale up Intelligent Tutoring Systems
Jakubův, J., Chvalovský, K., Goertzel, Z., Kaliszyk, C., Olšák, M., Piotrowski, B., Schulz, S., Suda, M., & Urban, J. (2023) · DROPS (Schloss Dagstuhl – Leibniz Center for Informatics) · doi:10.4230/lipics.itp.2023.19 · Cited by 76,658
Introducing a New Benchmarked Dataset for Activity Monitoring
Reiss, A., & Stricker, D. (2012) · 2012 16th International Symposium on Wearable Computers · doi:10.1109/iswc.2012.13 · Cited by 1,456
Doherty, A., Jackson, D., Hammerla, N., Plötz, T., Olivier, P., Granat, M. H., White, T., van Hees, V. T., Trenell, M. I., Owen, C. G., et al. (2017) · PLOS ONE · doi:10.1371/journal.pone.0169649 · Cited by 1,250 · cited 2×
“The dataset selected for testing is the largest currently available for smartphone motion sensor data, covering a wide range of activities [33,36–40].”
MethodMultiscale Convolutional Neural Networks for Fault Diagnosis of Wind Turbine Gearbox
Jiang, G., He, H., Yan, J., & Xie, P. (2019) · IEEE Transactions on Industrial Electronics · doi:10.1109/tie.2018.2844805 · Cited by 892 · cited 1×
“Several types of deep neural networks are typically used for time series classification of sensor signals, such as convolutional neural networks [14–16], fully convolutional neural networks [17], multiscale convolutional neural networks [18], time-LeNet [19], stacked denoising autoencoder [20], deep belief neural netw…”
MethodTransition-Aware Human Activity Recognition Using Smartphones
Reyes-Ortiz, J.-L., Oneto, L., Samà, A., Parra, X., & Anguita, D. (2016) · Neurocomputing · doi:10.1016/j.neucom.2015.07.085 · Cited by 740
Failure diagnosis using deep belief learning based health state classification
Tamilselvan, P., & Wang, P. (2013) · Reliability Engineering & System Safety · doi:10.1016/j.ress.2013.02.022 · Cited by 630 · cited 1×
“…fully convolutional neural networks [17], multiscale convolutional neural networks [18], time-LeNet [19], stacked denoising autoencoder [20], deep belief neural networks [21], Long Short-Term Memory (LSTM) deep recurrent neural network [22], echo state networks [23], or inception-ResNet [24].”
BackgroundDeep Learning in Human Activity Recognition with Wearable Sensors: A Review on Advances
Zhang, S., Li, Y., Zhang, S., Shahabi, F., Xia, S., Deng, Y., & Alshurafa, N. (2022) · Sensors · doi:10.3390/s22041476 · Cited by 625
mHealthDroid: A Novel Framework for Agile Development of Mobile Health Applications
Banos, O., Garcia, R., Holgado-Terriza, J. A., Damas, M., Pomares, H., Rojas, I., Saez, A., & Villalonga, C. (2014) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-13105-4_14 · Cited by 583 · cited 3×
“The The random forest method prediction accuracy on this dataset [36] was slightly lower than deep learning methods achieved on other datasets. random forest method prediction accuracy on this dataset [36] was slightly lower than deep learning methods achieved on other datasets.”
MethodJang, B., Kim, M., Harerimana, G., Kang, S.-U., & Kim, J. W. (2020) · Applied Sciences · doi:10.3390/app10175841 · Cited by 442 · cited 1×
“Before the transformer model was introduced, the cutting-edge systems for complex time series problems such as natural language processing were based on gated recurrent neural networks with a superimposed attention mechanism [42].”
MethodSmartphone and Smartwatch-Based Biometrics Using Activities of Daily Living
Weiss, G. M., Yoneda, K., & Hayajneh, T. (2019) · IEEE Access · doi:10.1109/access.2019.2940729 · Cited by 360
Yadav, S. K., Tiwari, K., Pandey, H. M., & Akbar, S. A. (2021) · Knowledge-Based Systems · doi:10.1016/j.knosys.2021.106970 · Cited by 311 · cited 1×
“Essentially, it aims at identifying human behavior based on data from sensors, available from personal devices such as smartphones, tablets, or smartwatches that can collect data from a wide sample of users and classify the signals using machine learning methods [1].”
BackgroundTime Series Classification With Multivariate Convolutional Neural Network
Liu, C.-L., Hsaio, W.-H., & Tu, Y.-C. (2019) · IEEE Transactions on Industrial Electronics · doi:10.1109/tie.2018.2864702 · Cited by 310
Imaging and fusing time series for wearable sensor-based human activity recognition
Qin, Z., Zhang, Y., Meng, S., Qin, Z., & Choo, K.-K. R. (2020) · Information Fusion · doi:10.1016/j.inffus.2019.06.014 · Cited by 309 · cited 3×
“Moreover, the model in this study is much smaller than ResNet used in [25].”
BackgroundMethodInfluentialiSPLInception: An Inception-ResNet Deep Learning Architecture for Human Activity Recognition
Ronald, M., Poulose, A., & Han, D. S. (2021) · IEEE Access · doi:10.1109/access.2021.3078184 · Cited by 217 · cited 1×
“Several types of deep neural networks are typically used for time series classification of sensor signals, such as convolutional neural networks [14–16], fully convolutional neural networks [17], multiscale convolutional neural networks [18], time-LeNet [19], stacked denoising autoencoder [20], deep belief neural netw…”
MethodWearable-Sensor-Based Detection and Prediction of Freezing of Gait in Parkinson’s Disease: A Review
Pardoel, S., Kofman, J., Nantel, J., & Lemaire, E. D. (2019) · Sensors · doi:10.3390/s19235141 · Cited by 201
Deep learning based human activity recognition (HAR) using wearable sensor data
Gupta, S. (2021) · International Journal of Information Management Data Insights · doi:10.1016/j.jjimei.2021.100046 · Cited by 197 · cited 1×
“Gupta [31] described the classification of simple as well as complex activities by widely used models such as InceptionTime or DeepConvLSTM.”
MethodGholamiangonabadi, D., Kiselov, N., & Grolinger, K. (2020) · IEEE Access · doi:10.1109/access.2020.3010715 · Cited by 196 · cited 1×
“The paper [7] focuses on the comparison of Feed Forward Neural Networks and Convolutional Neural Networks in terms of cross-validation on unseen subjects.”
BackgroundKU-HAR: An open dataset for heterogeneous human activity recognition
Sikder, N., & Nahid, A.-A. (2021) · Pattern Recognition Letters · doi:10.1016/j.patrec.2021.02.024 · Cited by 162 · cited 17×
“The KU-HAR dataset used in the current manuscript was published by Sikder and Nahid [33].”
BackgroundMethodInfluentialTrifan, A., Oliveira, M., & Oliveira, J. L. (2019) · JMIR mHealth and uHealth · doi:10.2196/12649 · Cited by 159 · cited 1×
“The technology of detecting human activities using mobile devices has great potential in medicine where it is possible to monitor patients with various diagnoses [2–5] and control compliance with treatment procedures or to use it as prevention against performing prohibited activities [6–9].”
BackgroundHuman Activity Recognition Based on Residual Network and BiLSTM
Li, Y., & Wang, L. (2022) · Sensors · doi:10.3390/s22020635 · Cited by 156 · cited 1×
“Li et al. [29] joined 2D convolution layers followed by a BiLSTM layer.”
MethodLocomotion Activity Recognition Using Stacked Denoising Autoencoders
Gu, F., Khoshelham, K., Valaee, S., Shang, J., & Zhang, R. (2018) · IEEE Internet of Things Journal · doi:10.1109/jiot.2018.2823084 · Cited by 124 · cited 1×
“Several types of deep neural networks are typically used for time series classification of sensor signals, such as convolutional neural networks [14–16], fully convolutional neural networks [17], multiscale convolutional neural networks [18], time-LeNet [19], stacked denoising autoencoder [20], deep belief neural netw…”
MethodWearable Sensor-Based Human Activity Recognition Using Hybrid Deep Learning Techniques
Wang, H., Zhao, J., Li, J., Tian, L., Tu, P., Cao, T., An, Y., Wang, K., & Li, S. (2020) · Security and Communication Networks · doi:10.1155/2020/2132138 · Cited by 124 · cited 2×
“Although the dataset used in [26] contained several times more examples than the dataset used here, it focused only on basic activities and transition activities, not on more complex movements such as pick, jump, push-up, or sit-up used in the present study.”
BackgroundMethodAnagnostis, A., Benos, L. T., Tsaopoulos, D. E., Tagarakis, A. C., Tsolakis, N., & Bochtis, D. (2021) · Applied Sciences · doi:10.3390/app11052188 · Cited by 115 · cited 1×
“In addition to health monitoring and rehabilitation, this technology can be used in gaming [10], human–robot interaction and robotics [11,12], and sports [13].”
BackgroundZebin, T., Sperrin, M., Peek, N., & Casson, A. J. (2018) · 2018 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) · doi:10.1109/embc.2018.8513115 · Cited by 107 · cited 2×
“Several types of deep neural networks are typically used for time series classification of sensor signals, such as convolutional neural networks [14–16], fully convolutional neural networks [17], multiscale convolutional neural networks [18], time-LeNet [19], stacked denoising autoencoder [20], deep belief neural netw…”
MethodDeep Neural Networks for Sensor-Based Human Activity Recognition Using Selective Kernel Convolution
Gao, W., Zhang, L., Huang, W., Min, F., He, J., & Song, A. (2021) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2021.3102735 · Cited by 105 · cited 1×
“The paper [4] introduced a new approach - attention for learning multiscale features among multiple kernels of 1D convolution layers in HAR issues.”
BackgroundImpact of EEG Parameters Detecting Dementia Diseases: A Systematic Review
Sánchez, L.-M., Rodríguez‐Reséndiz, J., Ramírez, G. N. A., Gomar, M. L. G., & Robles-Ocampo, J. B. (2021) · IEEE Access · doi:10.1109/access.2021.3083519 · Cited by 84
A Natural Walking Monitor for Pulmonary Patients Using Mobile Phones
Juen, J., Cheng, Q., & Schatz, B. (2015) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2015.2427511 · Cited by 83 · cited 1×
“The technology of detecting human activities using mobile devices has great potential in medicine where it is possible to monitor patients with various diagnoses [2–5] and control compliance with treatment procedures or to use it as prevention against performing prohibited activities [6–9].”
BackgroundEcheverri, C. J. O., Salazar-Colores, S., Rodríguez‐Reséndiz, J., & Gómez-Loenzo, R. A. (2019) · Sensors · doi:10.3390/s19204541 · Cited by 78 · cited 1×
“Similar to recurrent or convolutional networks [41], transformers efficiently handle time series classification and employ correlations between features within time steps.”
BackgroundTime Series Classification Using Time Warping Invariant Echo State Networks
Tanisaro, P., & Heidemann, G. (2016) · 2016 15th IEEE International Conference on Machine Learning and Applications (ICMLA) · doi:10.1109/icmla.2016.0149 · Cited by 75 · cited 1×
“Several types of deep neural networks are typically used for time series classification of sensor signals, such as convolutional neural networks [14–16], fully convolutional neural networks [17], multiscale convolutional neural networks [18], time-LeNet [19], stacked denoising autoencoder [20], deep belief neural netw…”
MethodSiT: Self-supervised vIsion Transformer
Atito, S., Awais, M., & Kittler, J. (2021) · arXiv · doi:10.48550/arxiv.2104.03602 · Cited by 75 · cited 1×
“Vision transformers were taught in a similar way, where randomly selected parts of the image were replaced by noise, and the transformer model aimed to fill these places identically to the original image [47].”
BackgroundMobile Devices For The Real Time Detection Of Specific Human Motion Disorders
Irrera, F., Lorenzi, P., Rao, R., Romano, G., & Kita, A. (2016) · IEEE Sensors Journal · doi:10.1109/jsen.2016.2530944 · Cited by 65
Deep Temporal Convolution Network for Time Series Classification
Koh, B. H. D., Lim, C. L. P., Rahimi, H., Woo, W. L., & Gao, B. (2021) · Sensors · doi:10.3390/s21020603 · Cited by 63 · cited 1×
“Several types of deep neural networks are typically used for time series classification of sensor signals, such as convolutional neural networks [14–16], fully convolutional neural networks [17], multiscale convolutional neural networks [18], time-LeNet [19], stacked denoising autoencoder [20], deep…”
Rosafalco, L., Manzoni, A., Mariani, S., & Corigliano, A. (2020) · Advanced Modeling and Simulation in Engineering Sciences · doi:10.1186/s40323-020-00174-1 · Cited by 60 · cited 1×
“…typically used for time series classification of sensor signals, such as convolutional neural networks [14–16], fully convolutional neural networks [17], multiscale convolutional neural networks [18], time-LeNet [19], stacked denoising autoencoder [20], deep belief neural networks [21], Long…”
MethodNew Sensor Data Structuring for Deeper Feature Extraction in Human Activity Recognition
Alemayoh, T. T., Lee, J. H., & Okamoto, S. (2021) · Sensors · doi:10.3390/s21082814 · Cited by 59 · cited 2×
“Several types of deep neural networks are typically used for time series classification of sensor signals, such as convolutional neural networks [14–16], fully convolutional neural networks [17], multiscale convolutional neural networks [18], time-LeNet [19], stacked denoising autoencoder [20], deep…”
BackgroundGupta, R., & Rajan, S. (2020) · Procedia Computer Science · doi:10.1016/j.procs.2020.04.165 · Cited by 32 · cited 1×
“…such as convolutional neural networks [14–16], fully convolutional neural networks [17], multiscale convolutional neural networks [18], time-LeNet [19], stacked denoising autoencoder [20], deep belief neural networks [21], Long Short-Term Memory (LSTM) deep recurrent neural network [22], echo…”
BackgroundActivity classification with smart phones for sports activities
Taylor, K., Abdulla, U. A., Helmer, R. J. N., Lee, J., & Blanchonette, I. (2011) · Procedia Engineering · doi:10.1016/j.proeng.2011.05.109 · Cited by 29 · cited 1×
“In addition to health monitoring and rehabilitation, this technology can be used in gaming [10], human–robot interaction and robotics [11,12], and sports [13].”
BackgroundLearning Bounded Context-Free-Grammar via LSTM and the Transformer: Difference and the Explanations
Shi, H., Gao, S., Tian, Y., Chen, X., & Zhao, J. (2022) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v36i8.20801 · Cited by 26
Sánchez, L.-M., Rodríguez‐Reséndiz, J., Salazar-Colores, S., Ramírez, G. N. A., & Perez-Soto, G. I. (2020) · Applied Sciences · doi:10.3390/app10031098 · Cited by 23
Accelerometer-based Hand Gesture Recognition for Human-Robot Interaction
Anderez, D. O., Santos, L. P. D., Lotfi, A., & Yahaya, S. W. (2019) · IEEE Symposium Series on Computational Intelligence (SSCI) · doi:10.1109/ssci44817.2019.9003136 · Cited by 23 · cited 1×
“In addition to health monitoring and rehabilitation, this technology can be used in gaming [10], human–robot interaction and robotics [11,12], and sports [13].”
BackgroundHuman computer interaction using gestures for mobile devices and serious games: A review
Spanogianopoulos, S., Sirlantzis, K., Mentzelopoulos, M., & Protopsaltis, A. (2014) · International Conference on Interactive Mobile Communication, Technologies and Learning · doi:10.1109/imctl.2014.7011154 · Cited by 18 · cited 1×
“In addition to health monitoring and rehabilitation, this technology can be used in gaming [10], human–robot interaction and robotics [11,12], and sports [13].”
BackgroundSpeed controller-based fuzzy logic for a biosignal-feedbacked cycloergometer
Ensastiga, S. A. L., Rodríguez‐Reséndiz, J., & Estévez-Bén, A. A. (2021) · Computer Methods in Biomechanics & Biomedical Engineering · doi:10.1080/10255842.2021.1977799 · Cited by 6 · cited 1×
“The technology of detecting human activities using mobile devices has great potential in medicine where it is possible to monitor patients with various diagnoses [2–5] and control compliance with treatment procedures or to use it as prevention against performing prohibited activities [6–9].”
BackgroundImplementation of a Socket for Hip Disarticulation based on Ergonomic Analysis
González, A. K., Bolivar, S. G., & Rodríguez‐Reséndiz, J. (2018) · IEEE-EMBS Conference on Biomedical Engineering and Sciences · doi:10.1109/iecbes.2018.8626706 · Cited by 1
Research works printed but not linked: 2
- Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., and Gelly, S. (2020). An image is worth 16x16 words: Transformers for image recognition at scale. arXiv.
- Bao, H., Dong, L., and Wei, F. (2021). Beit: Bert pre-training of image transformers. arXiv.
Cited by
Machine Learning Methods for Cancer Classification Using Gene Expression Data: A Review
Alharbi, F., & Vakanski, A. (2023) · Bioengineering · doi:10.3390/bioengineering10020173 · Cited by 241
The major human AP endonuclease (Ape1) is involved in the nucleotide incision repair pathway
Gros, L. (2004) · Nucleic Acids Research · doi:10.1093/nar/gkh165 · Cited by 209
A review of machine learning-based human activity recognition for diverse applications
Kulsoom, F., Narejo, S., Mehmood, Z., Chaudhry, H. N., Butt, A., & Bashir, A. K. (2022) · Neural Computing and Applications · doi:10.1007/s00521-022-07665-9 · Cited by 168
“Some research worked used attention models to extract feature-based spatiotemporal context [116, 130, 114].”
Method2D mapping of texture and lattice parameters of dental enamel
Al‐Jawad, M., Steuwer, A., Kilcoyne, S. H., Shore, R. C., Cywiński, R., & Wood, D. J. (2007) · Biomaterials · doi:10.1016/j.biomaterials.2007.02.019 · Cited by 119
Metadata
- Authors
- Dirgová Luptáková, I., Kubovčík, M., & Pospíchal, J.
- Affiliations
- Department of Applied Informatics, Faculty of Natural Sciences, University of Ss. Cyril and Methodius, J. Herdu 2, 917 01 Trnava, Slovakia
- University of Ss. Cyril and Methodius in Trnava (SK)
- Venue
- Sensors
- Publisher
- MDPI AG
- Published
- 2022-03-01
- Volume
- 22
- Issue
- 5
- Pages
- 1911
- ISSN
- 1424-8220 (ISSN-L)
- PMC
- PMC8914677
- OpenAlex
- W4214843096
- Cited by
- 270
- References
- 42
- Citations by year
- 2026: 60 · 2025: 54 · 2024: 64 · 2023: 47 · 2022: 16
- FWCI
- 14.80
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Sensors (publishedVersion, cc-by)
- Preprints.org (acceptedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- Slovak Research and Development Agency (APVV-17-0116)
- Cultural and Educational Grant Agency MŠVVaŠ SR (KEGA 012UCM-4/2021)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Context-Aware Activity Recognition Systems
- Physical Sciences › Computer Science › Signal Processing › Time Series Analysis and Forecasting
- Physical Sciences › Engineering › Biomedical Engineering › Gait Recognition and Analysis
- Keywords
- human activity recognition, wearable sensors, smartphone motion sensors, accelerometer, gyroscope, transformer models, self-attention, time series classification, motion signals
- MeSH terms
- Machine Learning, Wearable Electronic Devices, Algorithms, Humans, Human Activities, Neural Networks, Computer
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
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.