A Comprehensive Study of Activity Recognition Using Accelerometers
Open accessPublished 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.
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.
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
Breiman, L. (2001) · Machine Learning · doi:10.1023/a:1010933404324 · Cited by 133,538
ImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155 · cited 1×
“Convolutional Neural Networks Convolutional Neural Networks (CNNs), also referred to as ConvNets, are a widely used extension of MLPs which have seen great success in computer vision tasks [80].”
BackgroundHochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.8.1735 · Cited by 102,318 · cited 13×
“Our experiments use LSTM since they have been shown to perform similarly to other methods (e.g., GRU) in many settings [87].”
MethodInfluentialAdam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946 · cited 1×
“All networks were trained with the Adam optimiser [88] and parameters are tuned to minimise categorical cross entropy.”
MethodLearning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation
Cho, K., van Merrienboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., & Bengio, Y. (2014) · Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.3115/v1/d14-1179 · Cited by 25,545 · cited 2×
“Our experiments use LSTM since they have been shown to perform similarly to other methods (e.g., GRU) in many settings [87].”
MethodPan, S. J., & Yang, Q. (2010) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2009.191 · Cited by 23,387 · cited 1×
“The introduction of such methods will reduce the risk of overfitting, but Bayesian models can naturally be adapted to hierarchical models which can naturally lead to transfer learning frameworks [96].”
BackgroundAdvances in neural information processing systems 7
Cios, K., & Shields, M. (1997) · Neurocomputing · doi:10.1016/s0925-2312(97)90018-7 · Cited by 22,059
Multilayer feedforward networks are universal approximators
Hornik, K., Stinchcombe, M., & White, H. (1989) · Neural Networks · doi:10.1016/0893-6080(89)90020-8 · Cited by 21,728 · cited 1×
“However, since these models can model arbitrarily complex functions [46] care must be taken to minimise the risk of overfitting.”
BackgroundQuinlan, J. R. (1986) · Machine Learning · doi:10.1007/bf00116251 · Cited by 21,439 · cited 2×
“We note that whilst research has often focused on which ML algorithm performs best for the given dataset, we will assume instead here that virtually any state-of-the-art ML algorithm (e.g., kernel Support Vector Machines (SVMs) [12], Decision Trees [13], Bayesian classifiers [14]) can be made to perform equivalently g…”
BackgroundConditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data
Lafferty, J., McCallum, A., & Pereira, F. C. N. (2001) · ScholarlyCommons (University of Pennsylvania) · Cited by 12,935
Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling
Chung, J., Gulcehre, C., Cho, K., & Bengio, Y. (2014) · arXiv · doi:10.48550/arxiv.1412.3555 · Cited by 11,008
Learning Deep Architectures for AI
Bengio, Y. (2009) · Foundations and Trends® in Machine Learning · doi:10.1561/2200000006 · Cited by 9,239 · cited 1×
“This is usually done by optimising an objective function that captures the appropriateness 160 of the features, such as by energy minimisation or so-called “deep learning” (see [31] for a review) 161 Building on this, [32] developed sparse-coding framework for activity recognition exploits unlabelled 162 sample data,…”
MethodMatching pursuits with time-frequency dictionaries
Mallat, S. G., & Zhang, Z. (1993) · IEEE Transactions on Signal Processing · doi:10.1109/78.258082 · Cited by 9,112 · cited 1×
“Gabor time-frequency atoms are scaled, translated and modulated Gaussian functions g(t) [72].”
BackgroundAn introduction to variable and feature selection
GuyonIsabelle, & ElisseeffAndré (2003) · Journal of Machine Learning Research · doi:10.5555/944919.944968 · Cited by 7,715
An Introduction to Variable and Feature Selection.
Guyon, I., & Elisseeff, A. (2003) · Cited by 6,576
Emergence of simple-cell receptive field properties by learning a sparse code for natural images
Olshausen, B. A., & Field, D. J. (1996) · Nature · doi:10.1038/381607a0 · Cited by 5,694 · cited 4×
“Convolutional Sparse Coding (CSC) explicitly models local interactions through the convolution operator [73], however the resulting optimisation problem is considerably more complex than traditional sparse coding.”
BackgroundInfluentialAn introduction to hidden Markov models
Rabiner, L., & Juang, B. (1986) · IEEE ASSP Magazine · doi:10.1109/massp.1986.1165342 · Cited by 4,775 · cited 1×
“Hidden Markov Models (HMMs) 95 [20] or Conditional Random Fields (CRFs) [21]) and unstructured models (e.”
BackgroundLearning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation
Cho, K., van Merrienboer, B., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., & Bengio, Y. (2014) · arXiv · doi:10.48550/arxiv.1406.1078 · Cited by 3,317 · cited 2×
“Our experiments use LSTM since they have been shown to perform similarly to other methods (e.g., GRU) in many settings [87].”
MethodActivity Recognition from User-Annotated Acceleration Data
Bao, L., & Intille, S. S. (2004) · Lecture Notes in Computer Science · doi:10.1007/978-3-540-24646-6_1 · Cited by 3,147 · cited 7×
“193 The results of [14], which considered locations 1-5 of the above, suggested that multiple 194 accelerometers aided in recognition, since conjunctions between acceleration feature values at different 195 sites were useful for discriminating many activities.”
BackgroundMethodResultInfluentialActivity recognition using cell phone accelerometers
Kwapisz, J. R., Weiss, G. M., & Moore, S. A. (2011) · ACM SIGKDD Explorations Newsletter · doi:10.1145/1964897.1964918 · Cited by 2,912
Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition
Ordóñez, F., & Roggen, D. (2016) · Sensors · doi:10.3390/s16010115 · Cited by 2,761
Zur Theorie der orthogonalen Funktionensysteme: Erste Mitteilung
Haar, A. (1910) · Mathematische Annalen · doi:10.1007/bf01456326 · Cited by 2,138 · cited 3×
“The Haar wavelet basis defines a sparse representation of piecewise regular signals, and has therefore received much attention from the image processing community.”
MethodInfluentialHierarchical Dirichlet Processes
Teh, Y. W., Jordan, M. I., Beal, M. J., & Blei, D. M. (2006) · Journal of the American Statistical Association · doi:10.1198/016214506000000302 · Cited by 2,120 · cited 2×
“166 Finally, an interesting approach using Bayesian non-parametric methods was taken by [33], in 167 which they employed an Hierarchical Dirichlet Process (HDP) model [34] (a form of TM) to infer 168 physical activity levels from the raw accelerometer data, and used the extracted mixture proportions 169 as features to…”
MethodA Critical Review of Recurrent Neural Networks for Sequence Learning
Lipton, Z. C., Berkowitz, J. A., & Elkan, C. (2015) · arXiv (Cornell University) · doi:10.48550/arxiv.1506.00019 · Cited by 1,996 · cited 1×
“LSTMs have been shown to successfully learn time dependencies and perform well in a variety of tasks [82].”
Background(2013) · RECERCAT (Consorci de Serveis Universitaris de Catalunya) · Cited by 1,959
(2000) · CrossRef Listing of Deleted DOIs · doi:10.1162/15324430152748236 · Cited by 1,819
Activity recognition from accelerometer data
Ravi, N., Dandekar, N., Mysore, P., & Littman, M. L. (2005) · Innovative Applications of Artificial Intelligence · Cited by 1,775
A tutorial on human activity recognition using body-worn inertial sensors
Bulling, A., Blanke, U., & Schiele, B. (2014) · ACM Computing Surveys · doi:10.1145/2499621 · Cited by 1,668 · cited 1×
“7) These issues are shared with many other settings where Machine Learning (ML) is applied to Digital Signal Processing (DSP), and as such this is a fairly mature research area [10,11].”
BackgroundA tutorial on Principal Components Analysis
Smith, L. I. (2002) · Otago University Research Archive (University of Otago) · Cited by 1,636
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 · cited 1×
“The baseline performance of the PAMAP2 dataset [94] is approximately 80–90%.”
BackgroundIEEE Transactions on Signal Processing
(2023) · IEEE Transactions on Signal Processing · doi:10.1109/tsp.78 · Cited by 1,455
IEEE Transactions on Biomedical Engineering
(2025) · IEEE Transactions on Biomedical Engineering · doi:10.1109/tbme.10 · Cited by 1,400
Karantonis, D. M., Narayanan, M. R., Mathie, M., Lovell, N. H., & Celler, B. G. (2006) · IEEE Transactions on Information Technology in Biomedicine · doi:10.1109/titb.2005.856864 · Cited by 1,236 · cited 1×
“[20] examined two activities and various postures.”
Bayesian classification with Gaussian processes
Williams, C. K. I., & Barber, D. (1998) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/34.735807 · Cited by 766 · cited 1×
“kernel Support 63 Vector Machines (SVMs) [11], Decision Trees [12], Bayesian classifiers [13]) can be made to perform 64 equivalently given the appropriate feature set.”
BackgroundThe Opportunity challenge: A benchmark database for on-body sensor-based activity recognition
Chavarriaga, R., Sagha, H., Calatroni, A., Digumarti, S. T., Tröster, G., Millán, J. D. R., & Roggen, D. (2013) · Pattern Recognition Letters · doi:10.1016/j.patrec.2012.12.014 · Cited by 742 · cited 1×
“the SPHERE challenge [45] and Opportunity [54]) capture data and annotations 676 in less controlled settings, but do not yet capture the aspects of activity required to be considered 677 naturalistic.”
BackgroundOptimization with Sparsity-Inducing Penalties
Bach, F., Jenatton, R., Mairal, J., & Obozinski, G. (2025) · Foundations and Trends® in Machine Learning · doi:10.1561/2200000015 · Cited by 730 · cited 1×
“) is the L1 penalty Ω(zi) = ∑ i=1 |zi| (see [68] for a review).”
Reverend Bayes on Inference Engines: A Distributed Hierarchical Approach
Pearl, J. (2022) · Probabilistic and Causal Inference · doi:10.1145/3501714.3501727 · Cited by 727 · cited 1×
“In general, the probability distribution over the n-th node is influenced by the neighbouring nodes with graphical models, and this influence is propagated over the structure using algorithms based on message passing [84].”
MethodProceedings of the Twenty-Fifth International Joint Conference on Artificial Intelligence
Chen, Y., Diethe, T., & Flach, P. A. (2016) · National Conference on Artificial Intelligence · Cited by 680
Self-citationTracking of sedentary behaviours of young people: A systematic review
Biddle, S. J. H., Pearson, N., Ross, G. M., & Braithwaite, R. (2010) · Preventive Medicine · doi:10.1016/j.ypmed.2010.07.018 · Cited by 616 · cited 1×
“estimating calorie consumption or 16 monitoring (non-)sedentary behaviour [1]), the latter is attempting to discern the actual activities 17 occurring.”
BackgroundActivity Recognition and Monitoring Using Multiple Sensors on Different Body Positions
Maurer, U., Smailagic, A., Siewiorek, D. P., & Deisher, M. (2006) · International Workshop on Wearable and Implantable Body Sensor Networks (BSN'06) · doi:10.1109/bsn.2006.6 · Cited by 601 · cited 2×
“In [42], the authors investigate whether the high frequency information in the signal is relevant to the classification problem, and if not what level of down-sampling can be applied without affecting”
BackgroundSEGMENTING TIME SERIES: A SURVEY AND NOVEL APPROACH
Keogh, E., Chu, S., Hart, D., & Pazzani, M. (2004) · Series in Machine Perception and Artificial Intelligence · doi:10.1142/9789812565402_0001 · Cited by 600 · cited 1×
“Segmentation 176 Explicit segmentation of the sensor data stream is in itself is a non-trivial problem, and approaches 177 can roughly be partitioned into methods that rely on a sliding window [36], and probabilistic methods 178 based on HMMs (e.”
MethodAn Introduction to Conditional Random Fields
Sutton, C., & McCallum, A. (2012) · Foundations and Trends® in Machine Learning · doi:10.1561/2200000013 · Cited by 564 · cited 1×
“486 Conditional Random Fields (CRFs) [21,74] constitute a structured classification model of the 487 distribution of ym conditional on xm.”
BackgroundUSC-HAD: a daily activity dataset for ubiquitous activity recognition using wearable sensors
Zhang, M., & Sawchuk, A. A. (2012) · Proceedings of the 2012 ACM Conference on Ubiquitous Computing · doi:10.1145/2370216.2370438 · Cited by 554
Activity recognition on streaming sensor data
Krishnan, N. C., & Cook, D. J. (2014) · Pervasive and Mobile Computing · doi:10.1016/j.pmcj.2012.07.003 · Cited by 488 · cited 1×
“Daily Living (ADL) recognition using other types of sensor, such as Red/Green/Blue-Depth (RGB-D) 33 sensors [6] or other environmental sensors [7], or the fusion thereof [8].”
MethodDetection of posture and motion by accelerometry: a validation study in ambulatory monitoring
Foerster, F., Smeja, M., & Fahrenberg, J. (1999) · Computers in Human Behavior · doi:10.1016/s0747-5632(99)00037-0 · Cited by 480 · cited 1×
“An early study [44] reported an 272 overall accuracy of 95.”
BackgroundDiscovery of activity patterns using topic models
Huynh, T., Fritz, M., & Schiele, B. (2008) · Proceedings of the 10th international conference on Ubiquitous computing · doi:10.1145/1409635.1409638 · Cited by 442 · cited 1×
“3 [54] 13 days raw N/A N/A 17 11 7, 10–19 2-axis (BodyMedia Senswear) U 1 Automatic N/K wild Labels given by sensor 4 [55] 7 days T 773,817 12 1 37 1, 9, 12, 13, 20–47 3-axis (Porcupine) W, L 2.”
Proceedings of the 17th ACM conference on Information and knowledge management
Shanahan, J. G., Amer-Yahia, S., Manolescu, I., Yi, Z., Evans, D. A., Kolcz, A., Choi, K., & Chowdury, A. (2008) · doi:10.1145/1458082 · Cited by 410
Computer Vision and Pattern Recognition, 2008. CVPR 2008. IEEE Conference on
Ferrari, V., Marín-Jiménez, M. J., & Zisserman, A. (2008) · Edinburgh Research Explorer (University of Edinburgh) · Cited by 397
On-body localization of wearable devices: An investigation of position-aware activity recognition
Sztyler, T., & Stuckenschmidt, H. (2016) · 2016 IEEE International Conference on Pervasive Computing and Communications (PerCom) · doi:10.1109/percom.2016.7456521 · Cited by 369 · cited 1×
“As there are numerous placement locations on the body another questions arises: will activity recognition benefit from taking into account data from different on-body locations? In [43] a study was performed to determine if a model trained on the combined on-body locations performed better than a model that is aware o…”
BackgroundFeature Learning for Activity Recognition in Ubiquitous Computing
Plötz, T., Hammerla, N., & Olivier, P. (2011) · Newcastle University · doi:10.25405/ncl.34175319 · Cited by 337
Feature Selection and Activity Recognition System Using a Single Triaxial Accelerometer
Gupta, P., & Dallas, T. (2014) · IEEE Transactions on Biomedical Engineering · doi:10.1109/tbme.2014.2307069 · Cited by 335 · cited 1×
“For 136 example, [26] reports on applying Relief-F, a filter-based approach, to select accelerometer features for 137 activity recognition.”
MethodFast Convolutional Sparse Coding
Bristow, H., Eriksson, A., & Lucey, S. (2013) · 2013 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2013.57 · Cited by 329 · cited 2×
“Convolutional Sparse Coding (CSC) explicitly models local interactions through the convolution operator [73], however the resulting optimisation problem is considerably more complex than traditional sparse coding.”
BackgroundMethodHa, S., & Choi, S. (2016) · 2016 International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn.2016.7727224 · Cited by 307 · cited 1×
“More recently, deep architectures have also been proposed, which build on data from multiple body-worn sensors [44,45].”
BackgroundCreating and benchmarking a new dataset for physical activity monitoring
Reiss, A., & Stricker, D. (2012) · Proceedings of the 5th International Conference on PErvasive Technologies Related to Assistive Environments · doi:10.1145/2413097.2413148 · Cited by 295 · cited 1×
“8 [59] 9 h raw N/A N/A 42 1 10 3-axis (Porcupine) L 100 Polysomnography ±3 g lab Sleep study 9 [60] 1 day raw N/A N/A 8 1 10 3-axis (SleepTracker) L 100 Video N/K lab Sleep study 10 [61] 2 h raw N/A N/A 4 17 1, 4–6 3-axis U, L, C, W, B (12 total) 30 N/K lab 4 activities, 13 “gestures” 11 [62] 6 h raw N/A N/A 14 12 1–5…”
Activity Recognition from Accelerometer Data on a Mobile Phone
Brezmes, T., Gorricho, J.-L., & Cotrina, J. (2009) · Lecture Notes in Computer Science · doi:10.1007/978-3-642-02481-8_120 · Cited by 294
An HDP-HMM for systems with state persistence
Fox, E. B., Sudderth, E. B., Jordan, M. I., & Willsky, A. S. (2008) · Proceedings of the 25th international conference on Machine learning - ICML '08 · doi:10.1145/1390156.1390196 · Cited by 259
Bridging e-Health and the Internet of Things: The SPHERE Project
Zhu, N., Diethe, T., Camplani, M., Tao, L., Burrows, A., Twomey, N., Kaleshi, D., Mirmehdi, M., Flach, P., & Craddock, I. (2015) · IEEE Intelligent Systems · doi:10.1109/mis.2015.57 · Cited by 242 · cited 2×
“the SPHERE project [15]), which would provide a 86 specific driver for which activities are selected.”
BackgroundSelf-citationA Comprehensive Analysis on Wearable Acceleration Sensors in Human Activity Recognition
Janidarmian, M., Roshan Fekr, A., Radecka, K., & Zilic, Z. (2017) · Sensors · doi:10.3390/s17030529 · Cited by 217
Body-Worn Sensor Design: What Do Patients and Clinicians Want?
Bergmann, J. H. M., & McGregor, A. H. (2011) · Annals of Biomedical Engineering · doi:10.1007/s10439-011-0339-9 · Cited by 216 · cited 4×
“Further to this, the positioning of sensors also plays 69 an important role, and it is likely that this will be a limiting factor for many applications, since 70 the positioning of sensors is often largely driven by user acceptance rather than optimality of ADL 71 recognition performance [9].”
BackgroundMethodInfluentialRecognizing Human Activities User-independently on Smartphones Based on Accelerometer Data
Siirtola, P., & Röning, J. (2012) · International Journal of Interactive Multimedia and Artificial Intelligence · doi:10.9781/ijimai.2012.155 · Cited by 206
Personalization and user verification in wearable systems using biometric walking patterns
Casale, P., Pujol, O., & Radeva, P. (2012) · Personal and Ubiquitous Computing · doi:10.1007/s00779-011-0415-z · Cited by 172
Optimising sampling rates for accelerometer-based human activity recognition
Khan, A., Hammerla, N., Mellor, S., & Plötz, T. (2016) · Pattern Recognition Letters · doi:10.1016/j.patrec.2016.01.001 · Cited by 171 · cited 1×
“[43] performed a comprehensive study 252 on optimising the sampling frequency of accelerometers in the context of human activity recognition.”
BackgroundADL recognition based on the combination of RFID and accelerometer sensing
Stikic, M., Huynh, T., Van Laerhoven, K., & Schiele, B. (2008) · 2008 Second International Conference on Pervasive Computing Technologies for Healthcare · doi:10.1109/pcthealth.2008.4571084 · Cited by 163 · cited 1×
“5 Self ±3 g wild 5 [56] 20 min raw N/A N/A 12 10 33, 48–56 3-axis (Porcupine) L 100 Video ±3 g lab 6 [57] 2 h raw N/A N/A 1 3 1–3 3-axis (Porcupine) L 100 N/K ±3 g lab Includes strap loosening 7 [58] 14 days raw N/A N/A 17 11 7, 10–19 2-axis (BodyMedia Senswear) U 1 Self N/K wild”
Gaussian mixture based HMM for human daily activity recognition using 3D skeleton features
Piyathilaka, L., & Kodagoda, S. (2013) · IEEE Conference on Industrial Electronics and Applications (ICIEA) · doi:10.1109/iciea.2013.6566433 · Cited by 160 · cited 1×
“Daily Living (ADL) recognition using other types of sensor, such as Red/Green/Blue-Depth (RGB-D) 33 sensors [6] or other environmental sensors [7], or the fusion thereof [8].”
MethodHammerla, N. Y., Kirkham, R., Andras, P., & Ploetz, T. (2013) · Proceedings of the 2013 International Symposium on Wearable Computers · doi:10.1145/2493988.2494353 · Cited by 138 · cited 7×
“Three classes of feature are extracted: 545 statistical [4], dictionary [63] and ECDF [27] features are extracted.”
BackgroundInfluentialDeep Activity Recognition Models with Triaxial Accelerometers
Alsheikh, M. A., Selim, A., Niyato, D., Doyle, L., Lin, S., & Tan, H.-P. (2015) · arXiv · doi:10.48550/arxiv.1511.04664 · Cited by 126 · cited 3×
“Recently, [33,34] examined the possibility of learning features automatically.”
BackgroundDensity, Overcompleteness, and Localization of Frames. I. Theory
Balan, R. V., Casazza, P. G., Heil, C., & Landau, Z. A. (2006) · Journal of Fourier Analysis and Applications · doi:10.1007/s00041-006-6022-0 · Cited by 123
Using unlabeled data in a sparse-coding framework for human activity recognition
Bhattacharya, S., Nurmi, P., Hammerla, N., & Plötz, T. (2014) · Pervasive and Mobile Computing · doi:10.1016/j.pmcj.2014.05.006 · Cited by 121
Feature extraction from smartphone inertial signals for human activity segmentation
San-Segundo, R., Montero, J. M., Barra-Chicote, R., Fernández-Martínez, F., & Pardo, J. M. (2015) · Signal Processing · doi:10.1016/j.sigpro.2015.09.029 · Cited by 106
Semi-Markov conditional random fields for accelerometer-based activity recognition
Vinh, L. T., Lee, S., Le, H. X., Ngo, H. Q., Kim, H. I., Han, M., & Lee, Y.-K. (2010) · Applied Intelligence · doi:10.1007/s10489-010-0216-5 · Cited by 106 · cited 1×
“An approach based on semi-Markov CRFs that allows for overlapping 101 activities was introduced by [23], whose results indicated that the proposed approach worked well 102 even for complicated (higher-level) activities such eating and driving a car.”
MethodSPHERE: A Sensor Platform for Healthcare in a Residential Environment
Woznowski, P., Burrows, A., Diethe, T., Fafoutis, X., Hall, J., Hannuna, S., Camplani, M., Twomey, N., Kozłowski, M., Tan, B., et al. (2016) · Springer eBooks · doi:10.1007/978-3-319-44924-1_14 · Cited by 98 · cited 1×
“However, the SPHERE project is endeavouring to capture and release these datasets [16,17].”
BackgroundSelf-citationA multi-modal sensor infrastructure for healthcare in a residential environment
Woznowski, P., Fafoutis, X., Song, T., Hannuna, S., Camplani, M., Tao, L., Paiement, A., Mellios, E., Haghighi, M., Zhu, N., et al. (2015) · IEEE International Conference on Communications workshops · doi:10.1109/iccw.2015.7247190 · Cited by 80
Self-citationUsing rhythm awareness in long-term activity recognition
Laerhoven, K. V., Kilian, D., & Schiele, B. (2008) · Proceedings · doi:10.1109/iswc.2008.4911586 · Cited by 74 · cited 1×
“5 Self ±3 g wild 5 [56] 20 min raw N/A N/A 12 10 33, 48–56 3-axis (Porcupine) L 100 Video ±3 g lab 6 [57] 2 h raw N/A N/A 1 3 1–3 3-axis (Porcupine) L 100 N/K ±3 g lab Includes strap loosening 7 [58] 14 days raw N/A N/A 17 11 7, 10–19 2-axis (BodyMedia Senswear) U 1 Self N/K wild”
Extending the battery lifetime of wearable sensors with embedded machine learning
Fafoutis, X., Marchegiani, L., Elsts, A., Pope, J., Piechocki, R., & Craddock, I. (2018) · 2018 IEEE 4th World Forum on Internet of Things (WF-IoT) · doi:10.1109/wf-iot.2018.8355116 · Cited by 72 · cited 1×
“More recent works also demonstrate that simple classification tasks can be effectively 250 conducted at very low sampling frequency and resolution, increasing the battery lifetime of wearable 251 sensors by more than an order of magnitude [42].”
BackgroundSelf-citationCombining wearable and environmental sensing into an unobtrusive tool for long-term sleep studies
Borazio, M., & Laerhoven, K. V. (2012) · International Health Informatics Symposium · doi:10.1145/2110363.2110375 · Cited by 70
Conditional random fields for activity recognition in smart environments
Nazerfard, E., Das, B., Holder, L. B., & Cook, D. J. (2010) · Proceedings of the 1st ACM International Health Informatics Symposium · doi:10.1145/1882992.1883032 · Cited by 65 · cited 1×
“CRFs have been successfully employed for activity recognition in a smart-home environment [24], which although using environmental sensors rather than body-worn sensors would appear to have the same temporal characteristics.”
MethodTowards Benchmarked Sleep Detection with Wrist-Worn Sensing Units
Borazio, M., Berlin, E., Kucukyildiz, N., Scholl, P., & Van Laerhoven, K. (2014) · 2014 IEEE International Conference on Healthcare Informatics · doi:10.1109/ichi.2014.24 · Cited by 50
Designing Wearable Sensing Platforms for Healthcare in a Residential Environment
Fafoutis, X., Vafeas, A., Janko, B., Sherratt, R. S., Pope, J., Elsts, A., Mellios, E., Hilton, G., Oikonomou, G., Piechocki, R., et al. (2017) · EAI Endorsed Transactions on Pervasive Health and Technology · doi:10.4108/eai.7-9-2017.153063 · Cited by 46 · cited 2×
“For example, the SPW-2 wearable 228 sensor [41] employs the ADXL362 accelerometer and the LSM6DS0 gyroscope; ADXL362 consumes 229 approximately 8μW at 50 Hz while LSM6DS0 consumes approximately 2.”
BackgroundMethodSelf-citationThe SPHERE Challenge: Activity Recognition with Multimodal Sensor DataPreprint
Twomey, N., Diethe, T., Kull, M., Song, H., Camplani, M., Hannuna, S., Fafoutis, X., Zhu, N., Woznowski, P., Flach, P. A., et al. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1603.00797 · Cited by 46 · cited 2×
“276 A recent activity recognition challenge [45] introduced a new semi-naturalistic dataset with several 277 interesting features.”
BackgroundSelf-citationLearning a two-stage SVM/CRF sequence classifier
Hoefel, G., & Elkan, C. (2008) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/1458082.1458120 · Cited by 45 · cited 2×
“(17) In this work, we incorporate the methodology of [85] for our analysis of CRFs where unigram potentials of the CRF derive from the class-membership probability estimates of a base classifier.”
MethodTransformation-invariant representation and NMF
Eggert, J. P., Wersing, H., & Körner, E. (2005) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn.2004.1381038 · Cited by 41 · cited 2×
“In this work the authors used a shift invariant form of Non-negative Matrix Factorisation (NMF) [75], which is closely related to CSC, except that the signals are required to be non-negative.”
MethodLearning Features for Activity Recognition with Shift-Invariant Sparse Coding
Vollmer, C., Groß, H., & Eggert, J. P. (2013) · Lecture notes in computer science · doi:10.1007/978-3-642-40728-4_46 · Cited by 38
Unsupervised learning of sensor topologies for improving activity recognition in smart environments
Twomey, N., Diethe, T., Craddock, I. J., & Flach, P. A. (2016) · Neurocomputing · doi:10.1016/j.neucom.2016.12.049 · Cited by 36 · cited 1×
“However, un- and semi-supervised techniques [77,78] and others involving active and 681 transfer learning [79–81] can be utilised in these settings.”
BackgroundSelf-citationExperiences and Lessons Learned From Making IoT Sensing Platforms for Large-Scale Deployments
Fafoutis, X., Elsts, A., Piechocki, R. J., & Craddock, I. J. (2017) · IEEE Access · doi:10.1109/access.2017.2787418 · Cited by 31 · cited 1×
“Whilst energy consumption is not a challenge when data is 224 collected in controlled environments, it constitutes a major challenge when data is collected in natural 225 environments, particularly when the duration of the experiment exceeds the battery lifetime of the 226 sensor, as it can lead to loss of blocks of r…”
BackgroundSelf-citationActive transfer learning for activity recognition
Diethe, T., Twomey, N., & Flach, P. A. (2016) · Bristol Research (University of Bristol) · Cited by 28
Self-citationClustering-based activity classification with a wrist-worn accelerometer using basic features
Siirtola, P., Laurinen, P., Haapalainen, E., Röning, J., & Kinnunen, H. O. (2009) · IEEE Symposium on Computational Intelligence and Data Mining (CIDM) · doi:10.1109/cidm.2009.4938635 · Cited by 28 · cited 1×
“This is similar in nature to an earlier study by [35], who used an Expectation 172 Maximisation (EM)-based clustering algorithm to generate features for their classifier which they used 173 to recognise 9 sporting activities, and reported a ≈ 5% improvement over a standard classification 174 approach.”
MethodOn-Board Feature Extraction from Acceleration Data for Activity Recognition
Elsts, A., McConville, R., Fafoutis, X., Twomey, N., Piechocki, R. J., Santos‐Rodríguez, R., & Craddock, I. J. (2018) · Explore Bristol Research · doi:10.5555/3234847.3234868 · Cited by 27
Self-citationComparison of accelerometer-based energy expenditure estimation algorithms
Twomey, N., Faul, S., & Marnane, W. P. (2010) · International Conference on Pervasive Computing Technologies for Healthcare · doi:10.4108/icst.pervasivehealth2010.8879 · Cited by 24
Self-citationProbabilistic Sensor Fusion for Ambient Assisted LivingPreprint
Diethe, T., Twomey, N., Kull, M., Flach, P. A., & Craddock, I. J. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1702.01209 · Cited by 21 · cited 1×
“Note also that although outside the scope of this study, there is recent research in the area of Activities of Daily Living (ADL) recognition using other types of sensor, such as Red/Green/Blue-Depth (RGB-D) sensors [6] or other environmental sensors [7], or the fusion thereof [8].”
BackgroundSelf-citationEnabling Efficient Time Series Analysis for Wearable Activity Data
Laerhoven, K. V., Berlin, E., & Schiele, B. (2009) · International Conference on Machine Learning and Applications · doi:10.1109/icmla.2009.112 · Cited by 20 · cited 1×
“5 Self ±3 g wild 5 [56] 20 min raw N/A N/A 12 10 33, 48–56 3-axis (Porcupine) L 100 Video ±3 g lab 6 [57] 2 h raw N/A N/A 1 3 1–3 3-axis (Porcupine) L 100 N/K ±3 g lab Includes strap loosening 7 [58] 14 days raw N/A N/A 17 11 7, 10–19 2-axis (BodyMedia Senswear) U 1 Self N/K wild”
Bayesian Active Transfer Learning in Smart Homes
Diethe, T., Twomey, N., & Flach, P. A. (2015) · doi:10.13140/rg.2.1.3212.0487 · Cited by 14
Self-citationUsing time use with mobile sensor data: a road to practical mobile activity recognition?
Borazio, M., & Laerhoven, K. V. (2013) · International Conference on Mobile and Ubiquitous Multimedia (MUM) · doi:10.1145/2541831.2541850 · Cited by 12
A Bayesian Nonparametric Framework for Activity Recognition Using Accelerometer Data
Nguyen, T., Gupta, S., Venkatesh, S., & Phung, D. (2014) · 2014 22nd International Conference on Pattern Recognition · doi:10.1109/icpr.2014.352 · Cited by 11 · cited 1×
“166 Finally, an interesting approach using Bayesian non-parametric methods was taken by [33], in 167 which they employed an Hierarchical Dirichlet Process (HDP) model [34] (a form of TM) to infer 168 physical activity levels from the raw accelerometer data, and used the extracted mixture proportions 169 as features to…”
MethodBayesian Active Learning with Evidence-Based Instance Selection
Twomey, N., Diethe, T., & Flach, P. A. (2015) · Bristol Research (University of Bristol) · Cited by 7
Self-citationADL™: a topic model for discovery of activities of daily living in a smart home
Chen, Y., Diethe, T., & Flach, P. A. (2016) · International Joint Conference on Artificial Intelligence · Cited by 7
Self-citationOn the need for structure modelling in sequence prediction
Twomey, N., Diethe, T., & Flach, P. A. (2016) · Machine Learning · doi:10.1007/s10994-016-5571-y · Cited by 5 · cited 4×
“It has been 94 shown [19] that under certain conditions structured models (e.”
BackgroundMethodInfluentialSelf-citationBDL.NET: Bayesian dictionary learning in Infer.NET
Diethe, T., Twomey, N., & Flach, P. A. (2016) · Machine learning for signal processing ... · doi:10.1109/mlsp.2016.7738851 · Cited by 3
Self-citationSparse Machine Learning Methods with Applications in Multivariate Signal Processing
Diethe, T. (2010) · UCL Discovery (University College London) · Cited by 2
Self-citationEfficient approximate representations of computationally expensive features
Santos‐Rodríguez, R., & Twomey, N. (2018) · Bristol Research (University of Bristol) · Cited by 1
Self-citation
Research works printed but not linked: 3
- Shawe-Taylor, J., and Cristianini, N. (2000). Support Vector Machines, Cambridge University Press.
- Goodfellow, I., Bengio, Y., Courville, A., and Bengio, Y. (2016). Deep Learning, MIT Press.
- Tipping, 2001, Sparse Bayesian Learning and the Relevance Vector Machine, J. Mach. Learn. Res., 1, 211
Cited by
Measurement of physical activity in clinical practice using accelerometers
Arvidsson, D., Fridolfsson, J., & Börjesson, M. (2019) · Journal of Internal Medicine · doi:10.1111/joim.12908 · Cited by 239
Self-supervised learning for human activity recognition using 700,000 person-days of wearable data
Yuan, H., Chan, S., Creagh, A. P., Tong, C., Acquah, A., Clifton, D. A., & Doherty, A. (2024) · npj Digital Medicine · doi:10.1038/s41746-024-01062-3 · Cited by 166
“Current human activity recognition (HAR) models typically rely on manual feature engineering (Twomey et al., 2018; Haresamudram et al., 2019) partly due to the very limited size of existing labelled datasets.”
A Comprehensive Study of Activity Recognition Using Accelerometers
Twomey, N., Diethe, T., Fafoutis, X., Elsts, A., McConville, R., Flach, P., & Craddock, I. (2018) · doi:10.20944/preprints201803.0147.v1 · Cited by 143
Self-citationHARTH: A Human Activity Recognition Dataset for Machine Learning
Logacjov, A., Bach, K., Kongsvold, A., Bårdstu, H. B., & Mork, P. J. (2021) · Sensors · doi:10.3390/s21237853 · Cited by 184
“Human activity recognition (HAR) is a field of study that focuses on recognizing specific human physical activities and postures based on sensor data [7,8].”
BackgroundAn IoT-Based Design Using Accelerometers in Animal Behavior Recognition Systems
Tran, D.-N., Nguyen, T. N., Khanh, P. C. P., & Tran, D. (2021) · IEEE Sensors Journal · doi:10.1109/jsen.2021.3051194 · Cited by 142
“Incredibly we got high per-530 formance when classifying feeding (0.914 accuracy, 0.884 sensitivity, and 0.956 PPV) and standing (0.88 accuracy, 0.928 sensitivity, and 0.842 PPV), which was often a limit [4], [5], [22], [24]–[26], [45]–[47].”
BackgroundInfluential
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
- gold
- dx.doi.org
- Licence
- 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
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.