CrossHAR: Generalizing Cross-dataset Human Activity Recognition via Hierarchical Self-Supervised Pretraining
The DOI matches its Crossref record; 71 of the 71 references it cites were found in a registry or an index.
Abstract
The increasing availability of low-cost wearable devices and smartphones has significantly advanced the field of sensor-based human activity recognition (HAR), attracting considerable research interest. One of the major challenges in HAR is the domain shift problem in cross-dataset activity recognition, which occurs due to variations in users, device types, and sensor placements between the source dataset and the target dataset. Although domain adaptation methods have shown promise, they typically require access to the target dataset during the training process, which might not be practical in some scenarios. To address these issues, we introduce CrossHAR, a new HAR model designed to improve model performance on unseen target datasets. CrossHAR involves three main steps: (i) CrossHAR explores the sensor data generation principle to diversify the data distribution and augment the raw sensor data. (ii) CrossHAR then employs a hierarchical self-supervised pretraining approach with the augmented data to develop a generalizable representation. (iii) Finally, CrossHAR fine-tunes the pretrained model with a small set of labeled data in the source dataset, enhancing its performance in cross-dataset HAR. Our extensive experiments across multiple real-world HAR datasets demonstrate that CrossHAR outperforms current state-of-the-art methods by 10.83% in accuracy, demonstrating its effectiveness in generalizing to unseen target datasets.
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
“While effective, these methods suffer from two major limitations: (1) collecting data from each new user increases deployment costs, and (2) the two-stage training pipeline adds complexity and computational overhead.”
From Towards User-Generalizable Wearable-Sensor-Based Human Activity Recognition: A Multi-Task Contrastive Learning Approach (2025) · cites this paper as [11]
In the citing paper
“Prior works have explored personalized HAR frameworks [11,12,13,14,15,16], typically adopting a two-stage approach: pre-training on a general dataset followed by fine-tuning on data from the target user.”
From Towards User-Generalizable Wearable-Sensor-Based Human Activity Recognition: A Multi-Task Contrastive Learning Approach (2025) · cites this paper as [11]
In the citing paper
“While effective, these methods suffer from two major limitations: (1) collecting data from each new user increases deployment costs, and (2) the two-stage training pipeline adds complexity and computational overhead.”
From Towards User-Generalizable Wearable-Sensor-Based Human Activity Recognition: A Multi-Task Contrastive Learning Approach (2025) · cites this paper as [11]
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 1×
“• ResNet [24] (Residual Neural Network) is a popular deep learning model and has been applied in time series analysis.”
MethodKirillov, A., Mintun, E., Ravi, N., Mao, H., Rolland, C., Gustafson, L., Xiao, T., Whitehead, S., Berg, A. C., Lo, W.-Y., et al. (2023) · 2023 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00371 · Cited by 11,953
Transformers in Vision: A Survey
Khan, S., Naseer, M., Hayat, M., Zamir, S. W., Khan, F. S., & Shah, M. (2022) · ACM Computing Surveys · doi:10.1145/3505244 · Cited by 3,595 · cited 1×
“Recently, self-supervised pretraining has received increasing attention from the research community due to its great success in natural language processing [32] and computer vision [14, 19, 33].”
BackgroundTransformers in Vision: A Survey
Khan, S., Naseer, M., Hayat, M., Zamir, S. W., Khan, F. S., & Shah, M. (2021) · arXiv · doi:10.48550/arxiv.2101.01169 · Cited by 3,594 · cited 1×
“Recently, self-supervised pretraining has received increasing attention from the research community due to its great success in natural language processing [32] and computer vision [14, 19, 33].”
BackgroundThe Performance of LSTM and BiLSTM in Forecasting Time Series
Siami-Namini, S., Tavakoli, N., & Namin, A. S. (2019) · 2019 IEEE International Conference on Big Data (Big Data) · doi:10.1109/bigdata47090.2019.9005997 · Cited by 1,574 · cited 1×
“• LSTM [64] (Long Short-Term Memory) is a type of recurrent neural network architecture that is commonly used for time series classification tasks.”
BackgroundProceedings of the 2nd International Conference on Neural Information Processing Systems
Touretzky, D. S. (1989) · Cited by 1,439
Generalizing to unseen domains: A survey on domain generalization
Wang, J., Lan, C., Liu, C., Ouyang, Y., Qin, T., Lu, W., Chen, Y., Zeng, W., & Yu, P. (2022) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2022.3178128 · Cited by 1,046 · cited 4×
“Given these practical constraints, domain generalization [8, 35, 76, 84, 89], focuses on training models exclusively with source domain data to achieve robust performance on unseen target domains, becomes a more promising approach.”
BackgroundMethodInfluentialA survey of unsupervised deep domain adaptation
Wilson, G., & Cook, D. J. (2020) · ACM Transactions on Intelligent Systems and Technology · doi:10.1145/3400066 · Cited by 987 · cited 1×
“Domain adaptation , a widely studied technique to address the domain shift issue [25, 34, 80, 80], aims to bridge the gap between source and target domains by incorporating data from both during the training phase.”
BackgroundImageBind One Embedding Space to Bind Them All
Girdhar, R., El-Nouby, A., Liu, Z., Singh, M., Alwala, K. V., Joulin, A., & Misra, I. (2023) · 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52729.2023.01457 · Cited by 863 · cited 2×
“(ii) Furthermore, recent advancements in pretraining methodologies in the computer vision [11, 19] and natural language processing (NLP) [32] communities have shown remarkable effectiveness in improving generalization [31, 76].”
BackgroundLaMDA: Language Models for Dialog Applications
Thoppilan, R., De Freitas, D., Hall, J., Shazeer, N., Kulshreshtha, A., Cheng, H.-T., Jin, A., Bos, T., Baker, L., Du, Y., et al. (2022) · arXiv · doi:10.48550/arxiv.2201.08239 · Cited by 778 · cited 1×
“Motivated by the success of the foundation model in NLP [69] and computer vision communities [38], we will explore developing foundation models for sensor-based human activity recognition [40].”
Method
Research works printed but not linked: 22
- Si-An Chen, Chun-Liang Li, Nate Yoder, Sercan O Arik, and Tomas Pfister. 2023. Tsmixer: An all-mlp architecture for time series forecasting. arXiv preprint arXiv:2303.06053 (2023). · arXiv
- Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. 2020. A simple framework for contrastive learning of visual representations. In International conference on machine learning. PMLR, 1597--1607.
- Jiaxiang Dong, Haixu Wu, Haoran Zhang, Li Zhang, Jianmin Wang, and Mingsheng Long. 2023. SimMTM: A Simple Pre-Training Framework for Masked Time-Series Modeling. arXiv preprint arXiv:2302.00861 (2023). · arXiv
- Alexey Dosovitskiy Lucas Beyer Alexander Kolesnikov Dirk Weissenborn Xiaohua Zhai Thomas Unterthiner Mostafa Dehghani Matthias Minderer Georg Heigold Sylvain Gelly et al. 2020. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929 (2020). · arXiv
- Emadeldeen Eldele, Mohamed Ragab, Zhenghua Chen, Min Wu, Chee Keong Kwoh, Xiaoli Li, and Cuntai Guan. 2021. Time-series representation learning via temporal and contextual contrasting. arXiv preprint arXiv:2106.14112 (2021). · arXiv
- John Cristian Borges Gamboa. 2017. Deep learning for time-series analysis. arXiv preprint arXiv:1701.01887 (2017). · arXiv
- Hassan Ismail Fawaz, Germain Forestier, Jonathan Weber, Lhassane Idoumghar, and Pierre-Alain Muller. 2019. Deep learning for time series classification: a review. Data mining and knowledge discovery 33, 4 (2019), 917--963.
- Aryan Jadon, Avinash Patil, and Shruti Jadon. 2022. A Comprehensive Survey of Regression Based Loss Functions for Time Series Forecasting. arXiv preprint arXiv:2211.02989 (2022). · arXiv
- Junguang Jiang, Yang Shu, Jianmin Wang, and Mingsheng Long. 2022. Transferability in deep learning: A survey. arXiv preprint arXiv:2201.05867 (2022). · arXiv
- Jacob Devlin Ming-Wei Chang Kenton and Lee Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of naacL-HLT, Vol. 1. 2.
- Taesung Kim, Jinhee Kim, Yunwon Tae, Cheonbok Park, Jang-Ho Choi, and Jaegul Choo. 2021. Reversible instance normalization for accurate time-series forecasting against distribution shift. In International Conference on Learning Representations.
- Diederik P Kingma and Jimmy Ba. 2014. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014). · arXiv
- Zikang Leng, Hyeokhyen Kwon, and Thomas Plötz. 2023. On the Benefit of Generative Foundation Models for Human Activity Recognition. arXiv preprint arXiv:2310.12085 (2023). · arXiv
- Wang Lu, Jindong Wang, Xinwei Sun, Yiqiang Chen, and Xing Xie. 2023. Out-of-distribution Representation Learning for Time Series Classification. In The Eleventh International Conference on Learning Representations. https://openreview.net/forum?id=gUZWOE42l6Q
- Qianli Ma, Zhen Liu, Zhenjing Zheng, Ziyang Huang, Siying Zhu, Zhongzhong Yu, and James T Kwok. 2023. A Survey on Time-Series Pre-Trained Models. arXiv preprint arXiv:2305.10716 (2023). · arXiv
- Riccardo Presotto, Sannara Ek, Gabriele Civitarese, François Portet, Philippe Lalanda, and Claudio Bettini. 2023. Combining Public Human Activity Recognition Datasets to Mitigate Labeled Data Scarcity. arXiv preprint arXiv:2306.13735 (2023). · arXiv
- Xue Wang Liang Sun Rong Jin Tian Zhou Peisong Niu. 2023. One Fits All: Power General Time Series Analysis by Pretrained LM. In NeurIPS.
- Ilya O Tolstikhin, Neil Houlsby, Alexander Kolesnikov, Lucas Beyer, Xiaohua Zhai, Thomas Unterthiner, Jessica Yung, Andreas Steiner, Daniel Keysers, Jakob Uszkoreit, et al. 2021. Mlp-mixer: An all-mlp architecture for vision. Advances in neural information processing systems 34 (2021), 24261--24272.
- Dmitry Ulyanov, Andrea Vedaldi, and Victor Lempitsky. 2017. Instance Normalization: The Missing Ingredient for Fast Stylization. arXiv:1607.08022 [cs.CV] · arXiv
- Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. Advances in neural information processing systems 30 (2017).
- Zhiguang Wang, Weizhong Yan, and Tim Oates. 2017. Time series classification from scratch with deep neural networks: A strong baseline. In 2017 International joint conference on neural networks (IJCNN). IEEE, 1578--1585.
- Hongyi Zhang, Moustapha Cisse, Yann N Dauphin, and David Lopez-Paz. 2017. mixup: Beyond empirical risk minimization. arXiv preprint arXiv:1710.09412 (2017). · arXiv
Cited by
Qureshi, T. S., Shahid, M. H., Farhan, A. A., & Alamri, S. (2025) · Artificial Intelligence Review · doi:10.1007/s10462-025-11275-x · Cited by 69
GOAT: A Generalized Cross-Dataset Activity Recognition Framework with Natural Language Supervision
Miao, S., & Chen, L. (2024) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3699736 · Cited by 27
LLM4HAR: Generalizable On-device Human Activity Recognition with Pretrained LLMs
Hong, Z., Song, Y., Li, Z., Yu, A., Zhong, S., Ding, Y. C., He, T., & Zhang, D. (2025) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3711896.3737226 · Cited by 34
Self-citationBian, S., Liu, M., Rey, V. F., Geißler, D., & Lukowicz, P. (2025) · ACM International Symposium on Wearable Computers (ISWC) · doi:10.1145/3715071.3750410 · Cited by 20
PhyMask: An Adaptive Masking Paradigm for Efficient Self-Supervised Learning in IoT
Kara, D., Kimura, T., Chen, Y., Li, J., Wang, R., Chen, Y., Wang, T., Liu, S., & Abdelzaher, T. (2024) · ACM Conference on Embedded Networked Sensor Systems (SenSys) · doi:10.1145/3666025.3699325 · Cited by 20
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
“Researchers have utilized fine-tuning and domain adaptation methods in [12] to overcome the effect of device variability in cross-dataset scenarios, where one dataset is used to train a model and another is used to test it.”
Li, J., Xiu, C., Yang, D., & Wang, F. (2025) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2025.3575979 · Cited by 9
“Tag is placed horizontally on the chest when walking [50], [51].”
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
UniMTS: Unified Pre-training for Motion Time Series
Zhang, X., Teng, D., Chowdhury, R., Li, S., Hong, D., Gupta, R., & Shang, J. (2024) · Advances in Neural Information Processing Systems 37 · doi:10.52202/079017-3414 · Cited by 8
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
“…HAR approaches [18, 23, 38, 53, 55, 59] into two groups: (i) Source data-based method denotes those that are requiring the training (source) data of the downloaded HAR model [18, 55] and (ii) Source-free HAR denotes those that solving non-IID do not need source-domain data [23, 38, 53, 59].”
InfluentialAn Indoor Fusion Positioning Method Based on UWB/IMU/MAG in Harsh Environment
Li, J., Xiu, C., Ji, G., Wang, F., Wang, Y., Chakwizira, J., & Yang, D. (2025) · IEEE Sensors Journal · doi:10.1109/jsen.2025.3591547 · Cited by 8
“For IMU/MAG fusion localization, IMU positioning algorithms tend to accumulate localization errors over time because of unstable data in all axes [8], [9], which is corrected to some extent by integrating MAG data.”
AutoLife: Automatic Life Journaling with Smartphones and LLMs
Xu, H., Zeng, Z., Tong, P., Li, M., & Srivastava, M. B. (2025) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3770683 · Cited by 26
GNSSFormer: Enhancing GNSS Single Point Positioning Performance Based on Transformer for Smartphone
Li, J., Xiu, C., Yu, A., Hong, Z., Wang, F., Wang, H., Wang, Y., Chakwizira, J., & Yang, D. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3620847 · Cited by 8
“Additionally, the same algorithm has been studied across datasets between di ff erent smartphones [34], which is something that needs to be considered in the future to validate the proposed algorithm.”
Self-citationBernardo, J. B. L., Taparugssanagorn, A., Miyazaki, H., Pati, B. M., & Thapa, U. (2024) · Applied Sciences · doi:10.3390/app142210461 · Cited by 9
Guo, P., & Nakayama, M. (2025) · Sensors · doi:10.3390/s25226988 · Cited by 6
“Prior works have explored personalized HAR frameworks [11–16], typically adopting a two-stage approach: pre-training on a general dataset followed by fine-tuning on data from the target user.”
Behavior-aware Sparse Trajectory Recovery in Last-mile Delivery with Multi-scale Attention Fusion
Wang, H., Wang, S., Lin, L., Yang, Y., Wang, S., & Wen, H. (2024) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3627673.3680079 · Cited by 6
“Trajectory data plays a crucial role in various applications such as vehicle rebalancing and charging [34, 35], route planning [41], behavior-aware coordinate extraction [11], community road map sensing [12], and human activity recognition [9].”
Self-citationAddrLLM: Address Rewriting via Large Language Model on Nationwide Logistics Data
Yang, Q., Hong, Z., Cao, D., Wang, H., Xie, Z., He, T., Liu, Y., Yang, Y., & Zhang, D. (2025) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3690624.3709425 · Cited by 15
Self-citationScalable Area Difficulty Assessment with Knowledge-enhanced AI for Nationwide Logistics Systems
Xie, Z., Lyu, W., Song, Y., Wang, H., Yang, G., Liu, Y., He, T., Zhang, D., & Wang, G. (2025) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3690624.3709407 · Cited by 11
“Inspired by these findings, we address a real-world challenge in logistics [17, 18, 42, 44, 45].”
Self-citationWei, 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
ActiFormer: Sign-Aware Linear Attention for Sensor-Based Human Activity Recognition
Sun, Q., Fu, Z., Zhang, L., Chen, G., Huang, W., Wu, H., & Song, A. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3633705 · Cited by 5
“Simultaneously, when compared to models focused on domain generalization and self-supervised pretraining (such as CrossHAR [59]) and federated contrastive learning frameworks (such as FCLFD [61]), ActiFormer still achieves a superior F 1-score of 92.64% on the PAMAP2 dataset.”
InfluentialBian, S., Liu, M., Ray, L. S. S., Zhou, B., Guo, B., Yu, Z., Plötz, T., Lukowicz, P., Yuan, S., & Rey, V. F. (2026) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3810230 · Cited by 16
Rey, V. F., Rezende, P. M. B., Zhou, B., Suh, S., & Lukowicz, P. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.129288 · Cited by 14
FineSat: Enhancing GNSS Signals for High-precision Sensing
Yu, A., Wang, X., Wang, P., Li, J., Ma, X., Hong, Z., Wang, H., Yi, D., & Zhang, D. (2025) · IEEE International Conference on Pervasive Computing and Communications · doi:10.1109/percom64205.2025.00022 · Cited by 6
“Gesture recognition represents an important approach in human-computer interaction [38], [39], [42], [43].”
Self-citationD 3 -TR: Data-driven Daily Delivery Task Rescheduling for Cost-effective Last-mile Delivery
Zhang, L., Xiang, Y., Lyu, W., Hong, Z., Wang, H., Zhang, D., Liu, Y., & He, T. (2025) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3746252.3761550 · Cited by 7
Self-citationTowards Customizable Foundation Models for Human Activity Recognition with Wearable Devices
Qiu, M., Weng, C., Fan, M., & Wu, K. (2025) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3749479 · Cited by 4
“For example, a recent work, CrossHAR [14], proposed a hierarchical self-supervised pretraining framework that generalizes well across datasets, making it a potential foundation model solution.”
DisCovHAR: Contrastive Attention for Human Activity Recognition Under Distribution Shifts
Chen, L., Odema, M., & Faruque, M. A. A. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3551263 · Cited by 5
“Similarly, IN has been found to capture the shape and trend of time-series sensor data, which is beneficial for recognizing different activities and improves generalization toward unseen domains [36], [37].”
VibWalk: Mapping Lower-limb Haptic Experiences of Everyday Walking
Ying-Lei, S., Tang, D., Hu, W., Yoon, S. H., & Shao, Y. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3714254 · Cited by 3
HARNode: A Time-Synchronised, Open-Source, Multi-Device, Wearable System for Ad Hoc Field Studies
Lepold, P., Röddiger, T., & Beigl, M. (2025) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3714394.3756255 · Cited by 3
“Current work on human activity recognition (HAR) in the wild is constrained by data scarcity – often addressed using additional synthetic data (e.g. Hong et al. [15]), communication and synchronisation limits [10], high deployment cost [16], and ad hoc sensor placement [14][19].”
Wonderwall: A Virtual-to-Real Foundation Model for IMU-based HAR
Miao, S., & Chen, L. (2026) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3789688 · Cited by 9
Zhang, J., Peng, C., Liu, Z., Feng, L., Wu, Y., Dong, Y., & Xu, D. (2025) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2025.3572220 · Cited by 6
“We further incorporate three more recently proposed cross-domain activity recognition approaches: CrossHAR [53], ContrastSense [55], and Mob-HAR [54].”
Li, Y., Luo, Z., Zhu, G., Chen, T., Zhao, H., Zhong, C., Jin, D., Dang, Y., Yang, F., & Li, X. (2026) · Scientific Reports · doi:10.1038/s41598-026-37312-5 · Cited by 2
Sun, Q., Li, Z., Wang, K., Fu, Z., Chen, G., Zhang, L., Wu, H., & Song, A. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2026.3694938 · Cited by 2
Self-citationTanigaki, K., Maekawa, T., & Hara, T. (2026) · IEEE International Conference on Pervasive Computing and Communications · doi:10.1109/percom67906.2026.11524533 · Cited by 3
“The Few-shot Performance score is computed as the arithmetic mean of the F1 scores across all data-percentage settings, equally evaluating generalization capability at each setting: [19, 64, 65] has highlighted the importance of this task.”
Real-Time Abnormal Address Detection for Mobile Devices in Location-Based Services
Hong, Z., Yang, H., Wang, H., Lyu, W., Yang, Y., Wang, G., Liu, Y., Wang, Y., & Zhang, D. (2025) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2025.3545875 · Cited by 3
Self-citationCOMODO: Cross-Modal Video-to-IMU Distillation for Efficient Egocentric Human Activity Recognition
Chen, B., Wongso, W., Li, Z., Khaokaew, Y., Xue, H., & Salim, F. D. (2026) · Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies · doi:10.1145/3810218 · Cited by 1
“[24] introduced hierarchical self-supervised pretraining to improve cross-dataset generalization, and [15] handled missing modality conditions through a combination of modality dropout training and cross-modal contrastive learning.”
Self-citationUniMTS: Unified Pre-training for Motion Time Series
Zhang, X., Teng, D., Chowdhury, R. R., Li, S., Hong, D., Gupta, R. K., & Shang, J. (2024) · arXiv · Cited by 9
“To enhance semantics learning, we use large language models to augment text descriptions (Section 3.3.2).”
InfluentialMembership Inference Against Self-supervised IMU Sensing Applications
Zhao, T., Wang, N., & Wang, X. (2025) · ACM Conference on Embedded Networked Sensor Systems (SenSys) · doi:10.1145/3715014.3722060 · Cited by 4
High-Order Moments Conditional Domain Adaptation Networks for Wearable Human Activity Recognition
Ghosh, I., Chugh, G., Faridee, A. Z. M., & Roy, N. (2025) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3746252.3761225 · Cited by 4
NeighSqueeze: Compact Neighborhood Grouping for Efficient Billion-Scale Heterogeneous Graph Learning
Feng, X., Zhong, S., Hang, J., Zhang, Y., Yang, G., Wang, H., Zhang, D., & Wang, G. (2025) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3746252.3761569 · Cited by 2
Self-citationFraudulent Delivery Detection with Multimodal Courier Behavior Data in Last-Mile Delivery
Wang, S., Duan, S., Zhong, S., Hong, Z., Zhou, Z., Lin, H., Zuo, W., Zhang, D., & Ding, Y. C. (2025) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3746252.3761560 · Cited by 1
“With the wide-spread use of IoT devices and smartphones, last-mile delivery has incorporated an increasing number of sensor devices, such as GPS [7] and indoor/outdoor status detectors [13, 14].”
Self-citationSennan, S., Ramasubbareddy, S., Cho, Y., & Pandey, B. K. (2025) · Scientific Reports · doi:10.1038/s41598-025-26807-2 · Cited by 2
“It is important to note that the CrossHAR results shown in Table 8 were taken directly from the original study by Hong et al. 24 .”
UrbanPOI: Updating Urban POIs With Large-Scale Crowdsourcing for Location-Based Services
Hong, Z., Guo, B., Lyu, W., Wang, H., Liu, Y., Wang, G., He, T., & Zhang, D. (2026) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2026.3670149 · Cited by 1
“Current smartphones have many sensors that could potentially benefit the performance of POI updating, e.g., utilizing the barometer sensor to infer better indoor status [47], [48].”
Self-citationVsens: Incorporating XR into the Process of Collecting Virtual IMU Data
Liang, F., Min, T., Xia, C., & Sugiura, Y. (2026) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3790110 · Cited by 1
“During HAR system development, small deviations in sensor position or orientation can lead to substantial differences in acceleration and angular velocity patterns, thereby altering model performance and generalizability [6, 18].”
He, T., Lin, C., He, Q., Wu, Y., & Chen, J. (2026) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2026.3676474 · Cited by 1
“Hong et al. designed a local-global hierarchical self-supervised learning method, which achieved human activity recognition after fine-tuning using a small amount of labeled data [22].”
Self-citationBeyond Navigation: Unlocking the High-Precision Sensing Potential of GNSS Signals
Yu, A., Ma, X., Hong, Z., Wang, P., Wang, X., Li, J., Wang, H., Ding, Y., & Zhang, D. (2026) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2026.3699634 · Cited by 1
“Gesture recognition represents an important approach in human-computer interaction [1], [19], [50], [54], [55].”
Self-citationMODEL$\mathrm{X}_{c}$: A Novel Data-Efficient Personalization for Fall Detection
Yasmin, A., Mahmud, T., Alamgeer, S., Islam, T. Z., & Ngu, A. H. H. (2026) · IEEE International Conference on Digital Health (ICDH) · doi:10.1109/icdh72779.2026.00038 · Cited by 1
“For cross-dataset HAR, source-only generalization methods include CrossHAR by Hong et al. [24], which uses physically informed sensor augmentation and hierarchical self-supervised pretraining for robustness to unseen datasets without target data, and Guo et al.’s multi-task contrastive framework…”
Ye, X., & Wang, K. I. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5891594 · Cited by 2
Haresamudram, H., Tang, C. I., Suh, S., Lukowicz, P., & Ploetz, T. (2024) · arXiv · Cited by 31
“Alternatively, Hong et al. [84], improve cross dataset performance via masked modeling in conjunction with contrastive regularization.”
Augmentation Appproaches to Refine Wearable Human Activity Recognition
Salunkhe, S., Shinde, S., Patil, P., & Burchard, R. (2024) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3675094.3678464 · Cited by 2
TA-DA! - Improving Activity Recognition using Temporal Adapters and Data Augmentation
Hopp, M., Hartleb, H., & Burchard, R. (2024) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3675094.3678454 · Cited by 3
He, T., Xia, Z., & Chan, S.-H. G. (2025) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2025.3586618 · Cited by 2
“Recently, deep learning has shown powerful capabilities in INS signal processing, effectively classifying various pre-defined behaviors of phone users, including their actions, gestures, and gaits [26], [27], [28], [29], [30], [31], [32], [33].”
Self-citationAttrAppGen: User Attribute-Aware App Usage Trace Generation With Large Language Models
Duan, S., Chen, W., Liu, S., Lyu, F., & Zhang, Y. (2025) · IEEE Transactions on Network Science and Engineering · doi:10.1109/tnse.2025.3604790 · Cited by 1
“Such massive volumes of app usage data have significant implications for stakeholders, including smartphone manufacturers, network operators, app service providers, and end consumers [8], [9], [10], [11].”
Inertial Signal Forecasting With Foundation Model Techniques
Wieland, C., & Pankratius, V. (2025) · IEEE Sensors Journal · doi:10.1109/jsen.2025.3603456 · Cited by 3
“CrossHAR [29] applies cross-dataset evaluation by merging unlabeled signals from three datasets for self-supervised pre-training.”
Experience Paper: Adopting Activity Recognition in On-demand Food Delivery Business
Xu, H., Zhang, Y., Gao, W., Shen, G., & Li, M. (2025) · ACM Annual International Conference on Mobile Computing and Networking (MobiCom) · doi:10.1145/3680207.3765261 · Cited by 5
HandiSense: Left/Right Operating Hand Recognition for Multi-finger Touchscreen Interactions
Xu, X., Cui, M., Ding, D., & Ling, Z. (2025) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3770677
“• CrossHAR [25] is a hierarchical self-supervised learning framework designed to improve cross-user and cross-device generalization through robust pretraining.”
Zhang, Z., Zhao, D. Y., Zhu, G., Li, X. Q., Li, C., Liu, B., Li, Y., & Ma, H. (2026) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2026.3672513 · Cited by 1
“However, CrossHAR [37] and Imagebind [42] are limited to fixed modalities or require a specific modality to bind all multi-modal datasets.”
Zhang, T., Yi, C.-Z., Song, Y.-H., Wei, B.-C., & Chen, Z.-B. (2026) · IEEE Sensors Journal · doi:10.1109/jsen.2026.3680607
“In recent years, end-to-end architectures—ranging from CNN/RNN/TCN backbones to attention-based and transformer-style models—have been increasingly explored for sensor-based HAR to improve long-range temporal modeling and representation capacity [18], [21].”
Liu, X., & Wang, G. (2026) · Mathematics · doi:10.3390/math14111932
GenHAR: Generalizing Cross-domain Human Activity Recognition for Last-mile Delivery
Hong, Z., Li, Z., Fan, X., Yang, G., Guo, B., Wang, H., He, T., & Zhang, D. (2026) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3770854.3783921 · Cited by 2
Self-citationThink Before Action: Learning Suction Health for Reliable Grasp Evaluation in Logistics
Hao, Z., Hong, Z., Fang, J. H., Ma, J., Li, X., Zhang, D., Wang, H., He, T., & Yin, Z. (2026) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3770855.3818492
Self-citationLearning From Decoupled Sensor Axes: Enhancing Cross-Dataset Human Activity Recognition
Tang, Z.-P., He, S.-B., Wang, C., Chen, Q., Pan, J., & Gu, C.-J. (2026) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2026.3723993
“Pre-training strategies have also been explored: HAR-DoReMi [18] optimizes data mixtures for pre-training, and CrossHAR [12] employs hierarchical self-supervised learning to develop domain-invariant features.”
InfluentialUniMoRe : Uni fied Mo tion Re presentation via Gait IMU Pretraining
Chakraborty, S., Kumar, A., & Nigam, A. (2026) · IEEE Sensors Journal · doi:10.1109/jsen.2026.3718949
“Contrastive alignment approaches PRIMUS [1], IMU2CLIP [21], CroSSL [22], CrossHAR [23], exploit synchronized multimodal data to learn shared embedding spaces.”
Ghedini, C. G., Silva, A., Xavier, R., & Colombini, E. L. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7439359
Characterizing the Performance Gap in Human Activity Recognition for Older Adults
Khayami, H., Hwang, S., Ohn-Bar, E., Conroy, D. E., Lazar, A., Choe, E. K., & Kacorri, H. (2026) · ACM International Symposium on Wearable Computers (ISWC) · doi:10.1145/3830727.3834835
“Recent work has increasingly examined cross-dataset generalization [25, 34, 38, 39] and most studies report a substantial source-to-target drop while treating the domain gap as a single aggregate phenomenon [33, 61].”
Adapting Pretrained Large Vision Models for Sensor-based Activity Recognition
Cai, Y., Feng, R., Cai, K., Liu, Y., Guo, B., & Hong, Z. (2026) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3832003
“Such heterogeneity often leads to substantial performance degradation in cross-domain HAR scenarios [38, 66].”
InfluentialSelf-citationAutoResearcHAR: A Multi-Agent Framework for Dataset Adaptive IMU Based Human Activity Recognition
Ray, L. S. S., Rey, V. F., Bian, S., Suh, S., Lukowicz, P., & Zhou, B. (2026) · ACM International Symposium on Wearable Computers (ISWC) · doi:10.1145/3830727.3834823
Dong, C., Zhang, Q., & Wang, D. (2026) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3832005 · Cited by 1
StitchSense: Smart-Garment Sensing as a Source Domain for Watch-Based Activity Recognition
Datta, I., Leng, Z., Bhayana, R., & Ploetz, T. (2026) · ACM International Symposium on Wearable Computers (ISWC) · doi:10.1145/3830727.3834837 · Cited by 1
Kanimozhiraman, Maniraj, S. P., Geetha, R., Jegatheesan, A., Sujitha, B. B., & Jose Anand, A. (2026) · International Conference on Innovations in Mechanical, Robotics, Computing and Biomedical Engineering (ICIMRBE) · doi:10.1109/icimrbe70745.2026.11710115
“This issue is further exacerbated in the context of sustainability-oriented applications, due to minimal difference between ecologically meaningful classes, such as sparse vegetation and degraded forest, or wetland transition shemes and mixed crop fields [7-8].”
Lee, Y.-R., & Tseng, V. S. M. (2026) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn69854.2026.11728704
“CrossHAR [16] is a domain generalization method that does not access target-domain samples during training.”
TulipTender : A Mobile 3D Parkinson's Disease Assessment System Using a Single Camera
Xue, M., Zhang, Y., Wong, H. Y., Yang, Y., Chan, G., Cheung, Y.-F., Ip, N. Y., & Zhang, Q. (2026) · ACM Annual International Conference on Mobile Computing and Networking (MobiCom) · doi:10.1145/3795866.3844160
Metadata
- Authors
- Hong, Z., Li, Z., Zhong, S., Lyu, W., Wang, H., Ding, Y., He, T., & Zhang, D.
- Affiliations
- Rutgers, The State University of New Jersey (US)
- Jingdong (China) (CN)
- The University of Texas at Dallas (US)
- Venue
- Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
- Published
- 2024-05-13
- Volume
- 8
- Issue
- 2
- Pages
- 1-26
- ISSN
- 2474-9567 (ISSN-L)
- OpenAlex
- W4396919944
- Cited by
- 67
- References
- 66
- Citations by year
- 2026: 26 · 2025: 35 · 2024: 6
- FWCI
- 9.38
- Type
- journal article
- Language
- en
- Open access
- closed
- Locations
- Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies (publishedVersion)
- Funders
- National Science and Technology Major Project (2021ZD0114200)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Context-Aware Activity Recognition Systems
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Physical Sciences › Computer Science › Signal Processing › Time Series Analysis and Forecasting
- Keywords
- human activity recognition, cross-dataset generalization, domain shift, wearable sensors, smartphone sensors, domain adaptation, representation learning
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.