Smart Devices are Different: Assessing and MitigatingMobile Sensing Heterogeneities for Activity Recognition
The DOI matches its Crossref record; 48 of the 49 references it cites were found in a registry or an index.
Abstract
The widespread presence of motion sensors on users' personal mobile devices has spawned a growing research interest in human activity recognition (HAR). However, when deployed at a large-scale, e.g., on multiple devices, the performance of a HAR system is often significantly lower than in reported research results. This is due to variations in training and test device hardware and their operating system characteristics among others. In this paper, we systematically investigate sensor-, device- and workload-specific heterogeneities using 36 smartphones and smartwatches, consisting of 13 different device models from four manufacturers. Furthermore, we conduct experiments with nine users and investigate popular feature representation and classification techniques in HAR research. Our results indicate that on-device sensor and sensor handling heterogeneities impair HAR performances significantly. Moreover, the impairments vary significantly across devices and depends on the type of recognition technique used. We systematically evaluate the effect of mobile sensing heterogeneities on HAR and propose a novel clustering-based mitigation technique suitable for large-scale deployment of HAR, where heterogeneity of devices and their usage scenarios are intrinsic.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
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?
In the citing paper
“Stisen et al. [179] demonstrate the problem of sensing heterogeneities in HAR, showcasing how the same movements recorded at identical locations but with different sensors result in vastly different sensor readings.”
From Past, Present, and Future of Sensor-based Human Activity Recognition Using Wearables: A Surveying Tutorial on a Still Challenging Task (2025) · cites this paper as Stisen et al. [179]
In the citing paper
“This fragmentation stems from varying sensor characteristics across devices (e.g., watches, phones, wristbands), placement-dependent data variations, diverse sensing specifications [38,41,43], and inconsistent activity taxonomies [43].”
From Towards Customizable Foundation Models for Human Activity Recognition with Wearable Devices (2025) · cites this paper as [43]
In the citing paper
“The diversity in sensing conditions, including device types [43], placements [5], and user demographics [38] coupled with rapidly evolving sensing technologies and emerging activity types, makes it impractical to collect comprehensive datasets for every scenario.”
From Towards Customizable Foundation Models for Human Activity Recognition with Wearable Devices (2025) · cites this paper as [43]
In the citing paper
“In total, nine participants performed scripted activities for 5 minutes, ensuring an equal distribution of categories.”
From TS2ACT: Few-Shot Human Activity Sensing with Cross-Modal Co-Learning (2023) · cites this paper as [46]
In the citing paper
“The dataset covers a diverse set of devices, including 36 smart devices of 13 models from 4 manufacturers, with varying sampling rates to accommodate heterogeneous data.”
From TS2ACT: Few-Shot Human Activity Sensing with Cross-Modal Co-Learning (2023) · cites this paper as [46]
3 more claims about this paper
Create a free account to see every claim other papers make about this one.
Citations
Cites
Pattern Recognition and Machine Learning
(2006) · doi:10.1007/978-0-387-45528-0 · Cited by 9,824
Pattern Recognition and Machine Learning
(2007) · Kybernetes · doi:10.1108/03684920710743466 · Cited by 8,265
(2006) · Springer eBooks · Cited by 8,123
Clustering by Passing Messages Between Data Points
Frey, B. J., & Dueck, D. (2007) · Science · doi:10.1126/science.1136800 · Cited by 6,758 · cited 1×
“To this end, we chose an affinity propagation clustering algorithm [21].”
MethodNote on the Sampling Error of the Difference Between Correlated Proportions or Percentages
McNemar, Q. (1947) · Psychometrika · doi:10.1007/bf02295996 · Cited by 4,754 · cited 1×
“Lastly, we perform statistical testing to measure variations in recognition performances using the McNemar χ(2)-tests with Yates’ correction [35].”
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 1×
“Accelerometer-based HAR has been deployed in a large number of domains including smart homes [5, 32, 38], health care [31, 37], daily activity tracking [29], fitness tracking [16], fall detection of elderly people [18] and transportation mode detection [9, 24, 42].”
BackgroundActivity 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 · cited 2×
“selected activity classes (or a subset of them) have been investigated prominently in a large number of HAR publications [10, 11, 14, 27, 29, 41, 47].”
BackgroundA survey of mobile phone sensing
Lane, N., Miluzzo, E., Lu, H., Peebles, D., Choudhury, T., & Campbell, A. (2010) · IEEE Communications Magazine · doi:10.1109/mcom.2010.5560598 · Cited by 2,292 · cited 4×
“Motion sensors can be further paired with sensors, e.g., GPS, GSM, WiFi, and barometer, especially for recognizing tasks beyond basic HAR.”
BackgroundInfluentialA 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 8×
“A typical ARC for HAR is composed of five stages: (i) data collection, (ii) pre-processing, (iii) segmentation, (iv) feature extraction and (v) classification.”
BackgroundInfluentialUsing mobile phones to determine transportation modes
Reddy, S., Mun, M., Burke, J., Estrin, D., Hansen, M., & Srivastava, M. (2010) · ACM Transactions on Sensor Networks · doi:10.1145/1689239.1689243 · Cited by 852 · cited 1×
“Accelerometer-based HAR has been deployed in a large number of domains including smart homes [5, 32, 38], health care [31, 37], daily activity tracking [29], fitness tracking [16], fall detection of elderly people [18] and transportation mode detection [9, 24, 42].”
Background
Research works printed but not linked: 3
- 10.5555/1771592.1771620 · doi:10.5555/1771592.1771620
- 10.5555/2283516.2283683 · doi:10.5555/2283516.2283683
- 10.5555/1929661.1929712 · doi:10.5555/1929661.1929712
Cited by
Items 301–400 of 800
POND: Multi-Source Time Series Domain Adaptation with Information-Aware Prompt Tuning
Wang, J., Bai, G., Cheng, W. D., Chen, Z., Zhao, L., & Chen, H. (2024) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3637528.3671721 · Cited by 24
Wu, C., Wang, H., Zhang, X., Fang, Z., & Bu, J. (2024) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3664647.3680733 · Cited by 27
Using Smartphone Accelerometer for Human Physical Activity and Context Recognition in-the-Wild
Ehatisham-ul-Haq, M., Azam, M. A., Asim, Y., Amin, Y., Naeem, U., & Khalid, A. (2020) · Procedia Computer Science · doi:10.1016/j.procs.2020.10.007 · Cited by 16
Yu, H. Z., Yang, Q., Wang, G., & Xie, Y. (2020) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2020.3046114 · Cited by 16
“[9] quantified three of heterogeneity based on experiences, analyzed different interpolation methods and discussed how to mitigate the effects of different sampling frequencies.”
BackgroundAttention-based hierarchical denoised deep clustering network
Dong, Y., Wang, Z., Du, J., Fang, W., & Li, L. (2022) · World Wide Web · doi:10.1007/s11280-022-01007-4 · Cited by 16
YONO: Modeling Multiple Heterogeneous Neural Networks on Microcontrollers
Kwon, Y. D., Chauhan, J., & Mascolo, C. (2022) · ACM/IEEE International Conference on Information Processing in Sensor Networks (IPSN) · doi:10.1109/ipsn54338.2022.00030 · Cited by 20
Frequency domain data encoding in apache IoTDB
Wang, H., & Song, S. (2022) · Proceedings of the VLDB Endowment · doi:10.14778/3565816.3565829 · Cited by 16
“TEMP 171,012 Air temperatures of a wind farm PV 44,642,859 Voltage of a PV inverter [14] POWER 2,049,280 Household global active power [1] GAS 4,178,504 Readings of chemical sensor [28] HHAR 13,062,475 Smartphone accelerometer samples [44] GPS 263,718 GPS trajectory of seabirds [12] ECG 2,415,755 Electrocardiogram (EC…”
MethodStatistical Database of Human Motion Recognition Using Wearable IoT—A Review
Foroughi-Asl, E., Ebadollahi, S., Vahidnia, R., & Jalali, A. (2023) · IEEE Sensors Journal · doi:10.1109/jsen.2023.3282171 · Cited by 14
“[78] have recorded the HHAR dataset for detecting six different user activities: biking, sitting, standing, walking, stair up, and stair down.”
MethodFusion of Deep Neural Networks for Activity Recognition: A Regular Vine Copula Based Approach
Zhang, S., Geng, B., Varshney, P. K., & Rangaswamy, M. (2019) · International Conference on Information Fusion (FUSION) · doi:10.23919/fusion43075.2019.9011282 · Cited by 19
“[32], contains the sensor readings from two modalities: smartphone and smart watch.”
Comparing self-supervised learning techniques for wearable human activity recognition
Ek, S., Presotto, R., Civitarese, G., Portet, F., Lalanda, P., & Bettini, C. (2025) · CCF Transactions on Pervasive Computing and Interaction · doi:10.1007/s42486-024-00182-9 · Cited by 15
“The Heterogeneity Human Activity Recognition (HHAR) dataset (Stisen et al. 2015) showcases a diverse range of smartphone and smartwatch models, adding to our variability.”
Liu, X., Cheng, D., Fu, Z., Zhang, L., Wu, H., & Song, A. (2026) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2026.3664424 · Cited by 17
“We validate the effectiveness of our proposed prompt tuning method on three HAR benchmark datasets: HHAR [56], which primarily includes 6 daily activities: ‘Biking’, ‘Sitting’, ‘Stand-ing’, ‘Walking’, ‘Upstairs’, and ‘Downstairs’; PAMAP2 [57], which collects data on 12 daily activities (‘Lying’,…”
ContrastSense: Domain-invariant Contrastive Learning for In-the-Wild Wearable Sensing
Dai, G., Xu, H., Yoon, H., Li, M., Tan, R., & Lee, S.-J. (2024) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3699744 · Cited by 27
Ubiquitous smart learning system for smart cities
Elhoseny, H., Elhoseny, M., Abdelrazek, S. M., Riad, A. M., & Hassanien, A. E. (2017) · International Conference on Intelligent Computing and Information Systems (ICICIS) · doi:10.1109/intelcis.2017.8260058 · Cited by 16
IoT System Based on parameter optimization of Deep Learning using Genetic Algorithm
Slim, S. O., Elfattah, M. M. A., Atia, A., & Mostafa, M.-S. (2021) · International journal of intelligent engineering and systems · doi:10.22266/ijies2021.0430.20 · Cited by 16
Abdel‐Basset, M., Chang, V. I. C., Hawash, H., Chakrabortty, R. K., & Ryan, M. J. (2021) · Annals of Operations Research · doi:10.1007/s10479-021-04164-3 · Cited by 18
Multilevel Privacy Preservation Scheme Based on Compressed Sensing
Liang, J., Xiao, D., Huang, H. M., & Li, M. (2022) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2022.3209153 · Cited by 15
“The test is conducted on Human Activity Recognition (HAR) dataset [40], which contains 7035 records of 561-dimensional from 30 participants aged 19 to 48, and includes six activities: 1) Walking, 2) Walking Upstairs, 3) Walking Downstairs, 4) Sitting, 5) Standing, and 6) Laying.”
MethodSharing leaky-integrate-and-fire neurons for memory-efficient spiking neural networks
Kim, Y., Li, Y., Moitra, A., Yin, R., & Panda, P. (2023) · Frontiers in Neuroscience · doi:10.3389/fnins.2023.1230002 · Cited by 25
“• HHAR [60] is collected from nine subjects and encompasses six daily activities: biking, sitting, standing, walking, stair ascent, and stair descent.”
Contrastive self-representation learning for data clustering
Zhao, W., Gao, Q., Mei, S., & Yang, M. (2023) · Neural Networks · doi:10.1016/j.neunet.2023.08.050 · Cited by 16
Li, X., Liu, S., Zhou, Z., Guo, B., Xu, Y., & Yu, Z. (2024) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3643560 · Cited by 23
“EchoPFL focus on clustering-based PFL algorithms arises from the high clusterability observed in the data distributions across many mobile applications [6, 44, 51].”
BackgroundA CNN Based Transfer Learning Model for Automatic Activity Recognition from Accelerometer Sensors
Chikhaoui, B., Gouineau, F., & Sotir, M. (2018) · Lecture notes in computer science · doi:10.1007/978-3-319-96133-0_23 · Cited by 20
“Characteristics Heterogeneity HAR [27] RealWorld (HAR) [28] MobiAct [29]”
BackgroundHappyFeet: Challenges in Building an Automated Dance Recognition and Assessment Tool
Faridee, A. Z. M., Ramamurthy, S. R., & Roy, N. (2019) · GetMobile Mobile Computing and Communications · doi:10.1145/3308755.3308759 · Cited by 18
Carvalho, L. I., & Sofia, R. C. (2020) · IoT · doi:10.3390/iot1020025 · Cited by 18
From Shared Subspaces to Shared Landmarks: A Robust Multi-Source Classification Approach
Erfani, S., Baktashmotlagh, M., Moshtaghi, M., Nguyen, V., Leckie, C. A., Bailey, J., & Ramamohanarao, K. (2017) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v31i1.10870 · Cited by 19
“, domains) with different types, placements, orientations, and sampling frequencies (Stisen et al. 2015).”
BackgroundHaresamudram, H., Essa, I., & Plötz, T. (2024) · Sensors · doi:10.3390/s24041238 · Cited by 14
“Further, the reduction in performance of SAX-REPEAT relative to SAX on HHAR, MHEALTH, and PAMAP2, indicates that modifying SAX to apply to multi-variate data is challenging.”
BackgroundInfluentialWei, Q., Huang, J., Gao, Y. C., & Dong, W. (2025) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3749509 · Cited by 21
Extremely Fast Hoeffding Adaptive Tree
Manapragada, C., Salehi, M., & Webb, G. I. (2022) · Proceedings · doi:10.1109/icdm54844.2022.00042 · Cited by 20
“1 aws—price-discretized [19] 27410309 7 (4, 3) 10 2 chess [20] 28056 6(3, 3) 18 3 covtype [20], [21] 581012 54 (10, 44) 7 4 cpe [22], [23] 1455525 72 (22, 50) 10 5 fonts [20], [24] 745000 411 (410, 1) 153 6 hhar [20], [25] 43930257 9 (6, 3) 6 7 kdd [20] 4000000 42 (34, 8) 23 8 localization [20], [26] 164860 8 (4, 4) 1…”
Hoelzemann, A., & Laerhoven, K. V. (2024) · Frontiers in Computer Science · doi:10.3389/fcomp.2024.1379788 · Cited by 16
“According to [10], this modification decreases the runtime up to 48 % compared to a two-layered DeepConvLSTM while significantly increasing the overall classification performance on 4/5 publicly available datasets: [17], [18], [1], [19], [20].”
MethodHuman Activity Recognition: A Dynamic Inductive Bias Selection Perspective
Hamidi, M., & Osmani, A. (2021) · Sensors · doi:10.3390/s21217278 · Cited by 18
“This makes it difficult both to integrate new deployments and their constant evolution [11,72,73].”
BackgroundNeural generative model for clustering by separating particularity and commonality
Wang, W., Bao, J., & Guo, S. (2021) · Information Sciences · doi:10.1016/j.ins.2021.12.037 · Cited by 14
Two-Stage Alignments Framework for Unsupervised Domain Adaptation on Time Series Data
Xiang, X., Liu, Y., Fang, G., Liu, J., & Zhao, M. (2023) · IEEE Signal Processing Letters · doi:10.1109/lsp.2023.3264621 · Cited by 14
“Dataset: HAR [30] collects accelerometer, gyroscope and estimated body acceleration data from 30 subjects via smart-phone and smartwatch and simulates 6 activities.”
MethodMäder, A. R., Meegahapola, L., & Gática-Pérez, D. (2024) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642444 · Cited by 19
“This underscores the notion that different cultural and social practices in various countries impact the performance of social context inference models, as revealed in the analysis presented in Table 3.”
BackgroundMobile Computing, Applications, and Services
Sigg, S., Nurmi, P., & Salim, F. D. (2015) · Lecture notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering · doi:10.1007/978-3-319-29003-4 · Cited by 14
Survey on the application of deep learning in the Internet of Things
Shadroo, S., Rahmani, A. M., & Rezaee, A. (2022) · Telecommunication Systems · doi:10.1007/s11235-021-00870-2 · Cited by 14
“HHAR [76] The Heterogeneity Human Activity Recognition 43,930,257 Heterogeneous [36, 37] UserID [36] User identification Not specified Not specified [36, 37] NUS-WIDE [77] Web Image Dataset 20,000 Heterogeneous [34] Animal-20 Subset of animal 12,000 animal Heterogeneous [34] As shown in Fig.”
BackgroundInfluentialAdversarially regularized joint structured clustering network
Yang, Y., Ju, F., Sun, Y., Gao, J., & Yin, B. (2022) · Information Sciences · doi:10.1016/j.ins.2022.09.066 · Cited by 15
Dataset of inertial measurements of smartphones and smartwatches for human activity recognition
Matey-Sanz, M., Casteleyn, S., & Granell, C. (2023) · Data in Brief · doi:10.1016/j.dib.2023.109809 · Cited by 14
“Even though the sampling rate used in the data collection applications was set to 100 Hz, Android applications are not always able to apply the requested sampling rate [2] .”
BackgroundHydra: Hybrid-model federated learning for human activity recognition on heterogeneous devices
Wang, P., Ouyang, T., Wu, Q., Huang, Q., Gong, J., & Chen, X. (2023) · Journal of Systems Architecture · doi:10.1016/j.sysarc.2023.103052 · Cited by 16
MASA: Multimodal Federated Learning Through Modality-Aware and Secure Aggregation
Guo, J., Fu, Y., Zhai, Z., Li, X., Deng, Y., Yue, S., Chen, L., Pan, H., & Ren, J. (2025) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2025.3548954 · Cited by 18
Automatic Ontology-Based Model Evolution for Learning Changes in Dynamic Environments
Jabla, R., Khemaja, M., Buendía, F., & Faïz, S. (2021) · Applied Sciences · doi:10.3390/app112210770 · Cited by 13
“Evolved ontology-based model 3 26 38 HHAR [32] Ordinary apartment 16 43,930,250”
MethodActive Learning and Machine Teaching for Online Learning: A Study of Attention and Labelling Cost
Tegen, A., Davidsson, P., & Persson, J. (2021) · International Conference on Machine Learning and Applications · doi:10.1109/icmla52953.2021.00197 · Cited by 14
Federated Self-Supervised Learning in Heterogeneous Settings: Limits of a Baseline Approach on HAR
Ek, S., Rombourg, R., Portet, F., & Lalanda, P. (2022) · IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops) · doi:10.1109/percomworkshops53856.2022.9767369 · Cited by 16
“We selected the following datasets: the UCI dataset [12] – the standard benchmark dataset in the HAR community – which, in our experiments, represents five homogeneous clients, the Heterogeneity Human Activity Recognition (HHAR) dataset [13] which comprises 51 heterogeneous clients, the REALWORLD dataset [14] – a larg…”
MethodAn eager splitting strategy for online decision trees in ensembles
Manapragada, C., Gomes, H. M., Salehi, M., Bifet, A., & Webb, G. I. (2022) · Data Mining and Knowledge Discovery · doi:10.1007/s10618-021-00816-x · Cited by 13
“1 airlines [2] 539383 8 2 2 aws—price-discretized [53] 27410309 7 10 3 chess [18] 28056 6 18 4 covtype [9, 18] 581012 54 7 5 cpe [6, 8] 1455525 73 10 6 fonts [33, 18] 745000 411 153 7 hhar [49, 18] 43930257 9 6 8 kdd [18] 4000000 42 23 9 localization [30, 18] 164860 8 11 10 miniboone [42, 18] 130065 50 2 11 nbaiot [35…”
Domain Adaptation with Representation Learning and Nonlinear Relation for Time Series
Hussein, A., & Hajj, H. M. (2022) · ACM Transactions on Internet of Things · doi:10.1145/3502905 · Cited by 15
“Using the HAR dataset, the group of users {a, g, c, b, h} with {Samsung S+} were selected as source domain and {e, d, f} with {Nexus, S3, S3mini} were picked as the target domain.”
BackgroundMethodInfluentialBrard, R., Bellanger, L., Chevreuil, L., Doistau, F., Drouin, P., & Stamm, A. (2022) · Sensors · doi:10.3390/s22093555 · Cited by 14
“To choose the best classifier, we rigorously compared the most common supervised classification algorithms used in the HAR literature [22,23], namely, decision trees [24–26], support vector machine [10,19,24–28], k-nearest neighbors [24,25,27,28] and logistic regression [28,29].”
MethodFlexible and Self-Adaptive Sense-and-Compress for Sub-MicroWatt Always-on Sensory Recording
Roose, J. D., Xin, H., Andraud, M., Harpe, P. J. A., & Verhelst, M. (2018) · European Solid-State Circuits Conference (ESSCIRC) · doi:10.1109/esscirc.2018.8494270 · Cited by 13
“To prove the systems self-adaptivity and automation capabilities, we apply our system on Human-Activity-Recognition (HAR) accelerometer data [8], constructing the dataset with 2”
MethodKumar, P., & Chauhan, S. (2021) · CCF Transactions on Pervasive Computing and Interaction · doi:10.1007/s42486-021-00063-5 · Cited by 13
RetractedDeep Convolutional Bidirectional LSTM for Complex Activity Recognition with Missing Data
Saha, S. S., Sandha, S. S., & Srivastava, M. (2021) · Smart Innovation, Systems and Technologies · doi:10.1007/978-981-15-8269-1_4 · Cited by 12
An Efficient CNN-LSTM Approach for Smartphone Sensor-Based Human Activity Recognition System
Choudhury, N. A., & Soni, B. (2022) · International Conference on Computational Intelligence and Networks (CINE) · doi:10.1109/cine56307.2022.10037495 · Cited by 12
Enokibori, Y. (2024) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3699733 · Cited by 17
In Shift and In Variance: Assessing the Robustness of HAR Deep Learning Models Against Variability
Khaked, A. A., Oishi, N., Roggen, D., & Lago, P. (2025) · Sensors · doi:10.3390/s25020430 · Cited by 14
“Orientation, position, and device variability have been identified as a problem in HAR [29, 31], but they have either not been studied in isolation or not in the context of DL models.”
MASTER: A Multi-modal Foundation Model for Human Activity Recognition
Zhu, G., Zhao, D. Y., Li, C., Zhao, M., Zhang, Z., Quan, H., & Ma, H. (2025) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3749511 · Cited by 16
Adapting to Unknown Conditions in Learning-Based Mobile Sensing
Gong, T., Kim, Y., Choi, R., Shin, J., & Lee, S.-J. (2021) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2021.3061130 · Cited by 14
“These dependencies, as they make sensor readings dissimilar, significantly degrade the performance of mobile sensing [13], [14], [15].”
BackgroundMethodInfluentialRobust Graph Convolutional Clustering With Adaptive Graph Learning
Zhao, J., Sun, Y., Guo, J., Gao, J., & Yin, B. (2022) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn55064.2022.9892395 · Cited by 16
“Database Description : To verify the effectiveness and robustness of the proposed RGCC method, we conduct the experiments on the six common databases, including USPS [33], HHAR [34], Reuters [35], ACM [36], DBLP [36], and CITE 1 databases.”
MethodAl-Saedi, A. A., & Boeva, V. (2023) · Communications in computer and information science · doi:10.1007/978-3-031-34204-2_41 · Cited by 14
Ensemble of deep learning techniques to human activity recognition using smart phone signals
Imanzadeh, S., Tanha, J., & Jalili, M. (2024) · Multimedia Tools and Applications · doi:10.1007/s11042-024-18935-0 · Cited by 14
Towards Indoor Transportation Mode Detection Using Mobile Sensing
Prentow, T. S., Blunck, H., Kjærgaard, M. B., & Stisen, A. (2015) · Lecture notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering · doi:10.1007/978-3-319-29003-4_15 · Cited by 12
Self-citationACTIVITY RECOGNITION ON SMART DEVICES: Dealing with diversity in the wild
Blunck, H., Bhattacharya, S., Stisen, A., Prentow, T. S., Kjærgaard, M. B., Dey, A. K., Jensen, M. M., & Sonne, T. (2016) · GetMobile Mobile Computing and Communications · doi:10.1145/2972413.2972425 · Cited by 16
“In this article, we give an overview over recent results [6] into how the differences among user devices, including device manufacturers, models, OS types, and among their current usage states, e.g., CPU load conditions, have similarly drastic negative effects on recognition accuracy.”
BackgroundSelf-citationDione: Profiling spark applications exploiting graph similarity
Zacheilas, N., Maroulis, S., & Kalogeraki, V. (2017) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata.2017.8257950 · Cited by 12
“K-Means, BK-Means Clustering Phones Gyroscope data [12] 16 Als Explicit, Implicit Recommendation Movie Lens 20M dataset [13] 0.”
MethodDinarević, E. C., Husić, J. B., & Baraković, S. (2019) · Sensors · doi:10.3390/s19235206 · Cited by 12
Tsanousa, A., Μεδίτσκος, Γ., Vrochidis, S., & Angelis, L. (2020) · Journal of Ambient Intelligence and Humanized Computing · doi:10.1007/s12652-020-01836-z · Cited by 13
“This may be an indication that devices affect the performance of a sensor, a finding in line with Stisen et al. (2015).”
BackgroundMethodResultInfluentialTransNet: Minimally Supervised Deep Transfer Learning for Dynamic Adaptation of Wearable Systems
Rokni, S. A., Nourollahi, M., Alinia, P., Mirzadeh, I., Pedram, M., & Ghasemzadeh, H. G. (2020) · ACM Transactions on Design Automation of Electronic Systems · doi:10.1145/3414062 · Cited by 15
Flexible, Self-Adaptive Sense-and-Compress SoC for Sub-microWatt Always-On Sensory Recording
Roose, J. D., Xin, H., Hallawa, A., Ascheid, G., Harpe, P. J. A., & Verhelst, M. (2020) · IEEE Solid-State Circuits Letters · doi:10.1109/lssc.2020.3018382 · Cited by 12
“To prove the systems self-adaptivity and automation capabilities, we apply our system on Human-Activity-Recognition (HAR) accelerometer data [8], constructing the dataset with 2”
MethodFaster DBSCAN via subsampled similarity queriesPreprint
Jiang, H., Jang, J., & Łącki, J. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2006.06743 · Cited by 12
“001 10 Still [47] 949,983 3 6 0.”
Learning Disentangled Representations for Time SeriesPreprint
Li, Y., Chen, Z., Zha, D., Du, M., Zhang, D., Chen, H., & Hu, X. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2105.08179 · Cited by 12
“• Heterogeneity Human Activity Recognition (HHAR) [37]: includes accelerometer data from 31 smartphones of different manufacturers and models positioned in various orientations.”
InfluentialDeriving Effective Human Activity Recognition Systems through Objective Task Complexity Assessment
Hiremath, S. K., & Plötz, T. (2020) · Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies · doi:10.1145/3432227 · Cited by 11
“1 Skoda car manufacturing related movements body-worn IMU; gestures [35] 2 PAMAP2 physical activity monitoring body-worn IMU; ADL [31] 3 MHealth body-motion (and vitals) monitoring body-worn IMU; ADL [4] 4 USC-HAD low-level activities in health-care phone-based IMU; ADL [44] 5 Opportunity cleanup and breakfast activit…”
A Review of Wearable Sensor-based Human Activity Recognition using Deep Learning
Zhu, Y., & Mo, L. (2022) · International Conference on Sensing, Measurement & Data Analytics in the era of Artificial Intelligence (ICSMD) · doi:10.1109/icsmd57530.2022.10058422 · Cited by 11
“[43] A. Stisen et al., “Smart devices are different: Assessing and mitigating mobile sensing heterogeneities for activity recognition,” SenSys 2015 - Proc.”
BackgroundToo Good To Be True: accuracy overestimation in (re)current practices for Human Activity Recognition
Tello, A., Degeler, V., & Lazovik, A. (2024) · IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops) · doi:10.1109/percomworkshops59983.2024.10503465 · Cited by 17
“…(225) ICAIIC - Ni et al. [19] 2020, (19) Sensors (3.8) Gupta [20] 2021, (58) International Journal of Information Management Data Insights - Li et al. [21] 2021, (6) UBICOMP 2021 - Mekruksavanich and Jitpattanakul [22] 2021, (165) Sensors (3.8) Bouchabou et al. [23] 2021, (20) Communications in…”
BackgroundInfluentialZheng, W., & Arakawa, Y. (2025) · IEEE Access · doi:10.1109/access.2025.3598088 · Cited by 13
“Physical activity includes routine bodily motions, such as standing, sitting, lying down, walking, walking downstairs, and walking upstairs, which are fundamental to maintaining basic physical functions [52], [53], [54], [55], [56].”
RAG-HAR: Retrieval Augmented Generation-based Human Activity Recognition
Sivaroopan, N., Karunarathna, H., Madarasingha, C., Jayasumana, A., & Thilakarathna, K. (2025) · Annual IEEE International Conference on Pervasive Computing and Communications · doi:10.1109/percom67906.2026.11524560 · Cited by 9
“We evaluate RAG-HAR on six widely used HAR bench-marks—HHAR [37], PAMAP2 [32], MHEALTH [5], GO-TOV [30], SKODA [36], and USC-HAD [42]—which span 6 to 60 sensor channels and 6 to 16 activities, covering diverse activity types such as locomotion (e.g., walking, running), daily living tasks (e.g.,…”
Distributed Learning in Healthcare
Tuladhar, A., Rajashekar, D., & Forkert, N. D. (2022) · Integrated science · doi:10.1007/978-3-031-11199-0_10 · Cited by 11
Gestural and Touchscreen Interaction for Human-Robot Collaboration: A Comparative Study
Bongiovanni, A., Luca, A. D., Gava, L., Grassi, L., Lagomarsino, M., Lapolla, M., Marino, A., Roncagliolo, P., Macciò, S., Carfì, A., et al. (2023) · Lecture notes in networks and systems · doi:10.1007/978-3-031-22216-0_9 · Cited by 11
OpenHAR: A Matlab Toolbox for Easy Access to Publicly Open Human Activity Data Sets
Siirtola, P., Koskimäki, H., & Röning, J. (2018) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3267305.3267503 · Cited by 14
Madokoro, H., Nix, S., Woo, H., & Sato, K. (2021) · Applied Sciences · doi:10.3390/app112411807 · Cited by 11
On Robustness of Cloud Speech APIs: An Early Characterization
Mathur, A., Isopoussu, A., Kawsar, F., Smith, R. E., Lane, N. D., & Bianchi‐Berthouze, N. (2018) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3267305.3267505 · Cited by 11
“[23] studied sampling rate heterogeneity in inertial sensors of smart devices and how it impacts the accuracy of shallow HAR classifiers.”
BackgroundDailyLLM: Context-Aware Activity Log Generation Using Multi-Modal Sensors and LLMs
Tian, Y., Ren, X., Wang, Z., Güngör, O., Yu, X., & Rosing, T. (2025) · IEEE International Conference on Mobile Ad-Hoc and Sensor Systems (MASS) · doi:10.1109/mass66014.2025.00060 · Cited by 13
“HHAR [26] Motion [27] Shoaib [ The SOTA results for the activity benchmark dataset come from LLaSA [14], while the SOTA results (NMF [37] and ConvNet [38]) for the scenario dataset from the DCASE competition leaderboard [30], [31].”
Toptsis, M., Karkanis, N., Giannakoulas, A., & Kaifas, T. N. F. (2026) · Electronics · doi:10.3390/electronics15020295 · Cited by 13
LogoRA: Local-Global Representation Alignment for Robust Time Series Classification
Zhang, H., Zhang, Y.-F., Zhang, Z., Wen, Q., & Wang, L. (2024) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2024.3459908 · Cited by 15
Temporal Source Recovery for Time-Series Source-Free Unsupervised Domain Adaptation
Wang, Y., Gong, P., Wu, M., Ott, F., Li, X., Xie, L., & Chen, Z. (2026) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2026.3681762 · Cited by 16
“To comprehensively evaluate TemSR, we selected seven public TS datasets, including UCI-HAR [49], Sleep-EDF [50], Machine Fault Diagnosis (MFD) [51], HHAR [52], WISDM [53], EMG [54], and Opportunity HAR [55].”
Electronic Assessment of Physical Decline in Geriatric Cancer Patients
Fallahzadeh, R., Ghasemzadeh, H. G., & Shahrokni, A. (2018) · Current Oncology Reports · doi:10.1007/s11912-018-0670-5 · Cited by 10
Adaptive 1-dimensional time invariant learning for inertial sensor-based gait authentication
Permatasari, J., Connie, T., Ong, T. S., & Teoh, A. B. J. (2022) · Neural Computing and Applications · doi:10.1007/s00521-022-07741-0 · Cited by 10
“Therefore, the prediction results are evaluated with conventional machine learning techniques; the system will most probably fail due to the disparity between the domain and target data [21].”
BackgroundLearning to Generalize in Heterogeneous Federated Networks
Chen, C., Ye, T., Wang, L., & Gao, M. Z. (2022) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3511808.3557378 · Cited by 18
“This is a real-world human activity recognition dataset released by [32].”
BackgroundHussein, D., Belkhouja, T., Bhat, G., & Doppa, J. R. (2022) · International Conference on Computer Aided Design · doi:10.1145/3508352.3549430 · Cited by 18
“The sensor disturbances, in turn, lead to a degradation of the classification accuracy [1, 18, 28, 29].”
BackgroundMethodCo-MEAL: Cost-Optimal Multi-Expert Active Learning Architecture for Mobile Health Monitoring
Saeedi, R., Sasani, K., & Gebremedhin, A. H. (2017) · ACM International Conference on Bioinformatics, Computational Biology, and Health Informatics (ACM-BCB) · doi:10.1145/3107411.3107430 · Cited by 11
“g displaced data, misplaced data) [28, 30, 31].”
BackgroundJing, L., & Cheng, Z. (2017) · Conference proceedings/Conference proceedings - IEEE International Conference on Systems, Man, and Cybernetics · doi:10.1109/smc.2017.8122979 · Cited by 11
“For example, on Nokia Symbian and Maemo phones, the sampling rate of accelerometer vary between 25 ̃38 depending on the CPU load [9], which have unignorable influence on the stability of the HAR system [7].”
BackgroundAnomaly Detection Based on Fixed and Wearable Sensors in Assisted Living Environments
Mandarić, K., Skočir, P., Vuković, M., & Ježić, G. (2019) · International Conference on Software, Telecommunications and Computer Networks (SoftCOM) · doi:10.23919/softcom.2019.8903796 · Cited by 10
“[10] used 36 smartphones and smartwatches (13 different device models from four manufacturers) to analyze human activity recognition.”
BackgroundSemi-Unsupervised Learning: Clustering and Classifying using Ultra-Sparse Labels
Willetts, M., Roberts, S., & Holmes, C. C. (2020) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata50022.2020.9378265 · Cited by 15
“We use Fashion-MNIST, MNIST, and HAR [14]; masking the first half of classes partially and the second half fully in the same way as in the sub-section above.”
MethodDimensionality Reduction for Smart IoT Sensors
Vizárraga, J., Casas, R., Marco, Á., & Buldain, J. D. (2020) · Electronics · doi:10.3390/electronics9122035 · Cited by 10
“Some databases involved a large number of activities [2,28], many test subjects [29], a large number of sensors [1], or a large amount of data [1,3,29].”
BackgroundMethodSimilarity Embedding Networks for Robust Human Activity Recognition
Li, C., Tong, C. L., Niu, D., Jiang, B., Zuo, X. M., Cheng, L., Xiong, J., & Yang, J. (2021) · ACM Transactions on Knowledge Discovery from Data · doi:10.1145/3448021 · Cited by 14
“The USC-HAD [43] dataset was collected in 2012 by MotionNode, a wired sensing platform, with a sampling rate of 100 Hz, while the HHAR [34] dataset was collected in 2015 with eight different types of smartphones with sampling rates varying from 50 to 200 Hz.”
BackgroundMethodHow Much Unlabeled Data is Really Needed for Effective Self-Supervised Human Activity Recognition?
Dhekane, S. G., Haresamudram, H., Thukral, M., & Plötz, T. (2023) · Proceedings of the 2023 International Symposium on Wearable Computers · doi:10.1145/3594738.3611366 · Cited by 9
“For HHAR, PAMAP2, and RealWorld, 296, 354, and 1,204 minutes of data represent the maximum amount of data that can be used to perform both pre-training and fine-tuning.”
InfluentialSelf-Supervised Learning for Activity Recognition Based on Datasets With Imbalanced Classes
Hamad, R. A., Woo, W. L., Wei, B. W., & Yang, L. (2025) · IEEE Transactions on Emerging Topics in Computational Intelligence · doi:10.1109/tetci.2025.3526504 · Cited by 11
“The Heterogeneity Activity Recognition dataset [48] is recorded using two different smartphone sensors i.e. gyroscope and accelerometer as well as smartwatch sensors Authorized licensed use limited to the terms of the applicable license agreement with IEEE.”
MobHAR: Source-free Knowledge Transfer for Human Activity Recognition on Mobile Devices
Xue, M., Zhu, Y., Xie, W., Wang, Z., Chen, Y., Jiang, K., & Zhang, Q. (2025) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3712620 · Cited by 14
“As shown in Table 2, we summarize the factors contributing to the distribution shift based on five diverse datasets: HHAR [44], UCI [39], Motion [30], Shoaib [42], and USC [60].”
Investigating (re)current state-of-the-art in human activity recognition datasets
Bock, M., Hoelzemann, A., Moeller, M., & Laerhoven, K. V. (2022) · Frontiers in Computer Science · doi:10.3389/fcomp.2022.924954 · Cited by 10
“This type of plot can be recreated for every dataset using a script we provide in our repository (https://github.com/mariusbock/recurrent_state_of_the_art) and was used as basis for our analysis of each dataset. further investigate the overall shape of a dataset we created color-coded plots along with the sensor data…”
BackgroundMethodInfluentialKandpal, M., Sharma, B., Barik, R. K., Chowdhury, S., Patra, S. S., & Dhaou, I. B. (2023) · International Conference on Advanced Innovations in Smart Cities (ICAISC) · doi:10.1109/icaisc56366.2023.10085688 · Cited by 11
“Long Short-Term Memory (LSTM) network is a kind of RNN that has shown significant promise in speech recognition tasks [19-24].”
BackgroundSaha, S. S., Du, Y., Sandha, S. S., Garcia, L. A., Jawed, M. K., & Srivastava, M. B. (2023) · IEEE/ION Position Location and Navigation Symposium · doi:10.1109/plans53410.2023.10139997 · Cited by 13
Khaked, A. A., Oishi, N., Roggen, D., & Lago, P. (2023) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3594739.3610742 · Cited by 9
“affect measurement quality from device to device [24].”
Li, Y., Qin, X., Geng, J., Chen, R., Hou, Y., Gong, Y., Pan, M., & Zhang, P. (2024) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2024.3365477 · Cited by 9
Simon, L., Terhorst, Y., Cohrdes, C., Pryss, R. C., Steinmetz, L., Elhai, J. D., & Baumeister, H. (2024) · Sleep Medicine X · doi:10.1016/j.sleepx.2024.100114 · Cited by 11
An enhanced ResNet deep learning method for multimodal signal-based locomotion intention recognition
Sun, H., Gu, X., Zhang, Y., Sun, F., Zhang, S., Wang, D., & Yu, H. (2024) · Biomedical Signal Processing and Control · doi:10.1016/j.bspc.2024.107254 · Cited by 12
Task phase recognition for highly mobile workers in large building complexes
Stisen, A., Mathisen, A., Sørensen, S., Blunck, H., Kjargaard, M. B., & Prentow, T. S. (2016) · Annual IEEE International Conference on Pervasive Computing and Communications · doi:10.1109/percom.2016.7456504 · Cited by 10
Self-citationFatima, S. (2021) · International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS) · doi:10.1145/3441852.3476475 · Cited by 9
“Through a systematic literature review, we find that there are three main smartwatch datasets, none of which represent older adults: WISDM [38, 39], UCI-HHAR [34], and the Extrasensory [36] dataset.”
BackgroundContinual adaptation of federated reservoirs in pervasive environments
Caro, V. D., Gallicchio, C., & Bacciu, D. (2023) · Neurocomputing · doi:10.1016/j.neucom.2023.126638 · Cited by 11
Metadata
- Authors
- Stisen, A., Blunck, H., Bhattacharya, S., Prentow, T. S., Kjærgaard, M. B., Dey, A., Sonne, T., & Jensen, M. M.
- Affiliations
- Aarhus University (DK)
- Bell Laboratories, Dublin, Ireland
- Carnegie Mellon University (US)
- ORCID
- Venue
- Proceedings of the 13th ACM Conference on Embedded Networked Sensor Systems
- Publisher
- ACM
- Published
- 2015-11
- Pages
- 127-140
- OpenAlex
- W2057907879
- Cited by
- 714
- References
- 49
- Citations by year
- 2026: 45 · 2025: 81 · 2024: 99 · 2023: 91 · 2022: 82 · 2021: 82 · 2020: 61 · 2019: 62 · 2018: 54 · 2017: 35 · 2016: 21 · 2015: 1
- FWCI
- 15.18
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- ACM Conference on Embedded Networked Sensor Systems (SenSys) (publishedVersion)
- Funders
- Danish Advanced Technology Foundation (076-2011- 3.)
- The Danish Council for Strategic Research (11-11533)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Context-Aware Activity Recognition Systems
- Social Sciences › Transportation › Human Mobility and Location-Based Analysis
- Physical Sciences › Computer Science › Computer Science Applications › Mobile Crowdsensing and Crowdsourcing
- Keywords
- human activity recognition, mobile sensing, sensor heterogeneity, heterogeneous devices, smartphone sensors, smartwatches, large-scale deployment, feature representation, classification methods, motion sensors
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.