Understanding and improving recurrent networks for human activity recognition by continuous attention
The DOI matches its Crossref record; 31 of the 31 references it cites were found in a registry or an index.
Abstract
Deep neural networks, including recurrent networks, have been successfully applied to human activity recognition. Unfortunately, the final representation learned by recurrent networks might encode some noise (irrelevant signal components, unimportant sensor modalities, etc.). Besides, it is difficult to interpret the recurrent networks to gain insight into the models' behavior. To address these issues, we propose two attention models for human activity recognition: temporal attention and sensor attention. These two mechanisms adaptively focus on important signals and sensor modalities. To further improve the understandability and mean Fl score, we add continuity constraints, considering that continuous sensor signals are more robust than discrete ones. We evaluate the approaches on three datasets and obtain state-of-the-art results. Furthermore, qualitative analysis shows that the attention learned by the models agree well with human intuition.
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
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source says what the claim attributes to it.
In the citing paper
“Temporal attention was applied by Murahari et al. [138] whereas continuous sensor and temporal attention were evaluated by Zeng et al. [214], both leading to increased performance.”
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 [214]
The cited source says what the claim attributes to it.
In the citing paper
“[226] applies temporal attention and sensor attention into LSTM to improve the overall activity recognition accuracy by adaptively focusing on important time windows and sensor modalities.”
From Deep Learning in Human Activity Recognition with Wearable Sensors: A Review on Advances (2022) · cites this paper as [226]
In the citing paper
“The squeeze-and-excitation network (SENet) [15] is one of the representative methods, which brings clear performance gain across various deep CNN architectures via learning channel attention for each convolution block.”
From DanHAR: Dual Attention Network for multimodal human activity recognition using wearable sensors (2021) · cites this paper as [15]
In the citing paper
“SENet [15] for the first time proposes an effective mechanism to learn channel attention and achieves promising performance.”
From DanHAR: Dual Attention Network for multimodal human activity recognition using wearable sensors (2021) · cites this paper as [15]
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
Hochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.8.1735 · Cited by 102,318 · cited 8×
“Models PAMAP2 DG Skoda4 LSTM baseline ([12]) (LSTM without Attention) 0.”
BackgroundMethodResultInfluentialNeural Machine Translation by Jointly Learning to Align and Translate
Bahdanau, D., Cho, K., & Bengio, Y. (2014) · arXiv · doi:10.48550/arxiv.1409.0473 · Cited by 16,213 · cited 4×
“The attention approach is proposed to mitigate this problem for speech and natural language processing (NLP) [3, 14], and it offers means to explicate the inner workings of neural networks.”
BackgroundInfluentialApproximation by superpositions of a sigmoidal function
Cybenko, G. (1989) · Mathematics of Control, Signals, and Systems · doi:10.1007/bf02551274 · Cited by 13,883
Show, Attend and Tell: Neural Image Caption Generation with Visual Attention
Xu, K., Ba, J., Kiros, R., Cho, K., Courville, A., Salakhutdinov, R., Zemel, R., & Bengio, Y. (2015) · arXiv · doi:10.48550/arxiv.1502.03044 · Cited by 8,248
Activity 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 4×
“RECOGNITION We frame HAR as a sequence classification problem [4].”
BackgroundMethodInfluentialDeep 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 · cited 4×
“The baseline methods include standard LSTM [12], DeepConvLSTM [19], and LSTM-S (LSTM + sample-by-sample analysis) [11].”
BackgroundMethodInfluentialAttention-Based Models for Speech Recognition
Chorowski, J., Bahdanau, D., Serdyuk, D., Cho, K., & Bengio, Y. (2015) · arXiv · doi:10.48550/arxiv.1506.07503 · Cited by 2,021 · cited 1×
“Deep neural networks (DNNs) has had revolutionary impact in speech recognition [7], image classification [24], and machine translation [3].”
BackgroundA 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 2×
“INTRODUCTION Wearable-based human activity recognition (HAR) systems are an indispensable component in mobile ubiquitous computing and human-computer interaction [5].”
BackgroundIntroducing 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 first is the PAMAP2 dataset (Physical Activity Monitoring for Aging People 2) [21].”
Background(2015) · Cited by 989
Research works printed but not linked: 1
- 11. Nils Y Hammerla, Shane Halloran, and Thomas Ploetz. 2016. Deep, convolutional, and recurrent models for human activity recognition using wearables. IJCAI (2016).
Cited by
Items 101–186 of 186
Qin, L., Zhang, L., Li, C., Song, C., Cheng, D., Wang, S., Wu, H., & Song, A. (2024) · IEEE Transactions on Artificial Intelligence · doi:10.1109/tai.2024.3489532 · Cited by 12
“Zeng et al. [18], Yang et al. [19], and Ronao et al. [20] at an earliest time have applied convolution neural networks (CNNs) to handle multivariate sensor time series for HAR.”
Multi-agent Transformer Networks for Multimodal Human Activity Recognition
Li, J., Yao, L., Li, B., Wang, X., & Sammut, C. (2022) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3511808.3557402 · Cited by 15
“Recent approaches utilized deep neural networks, such as CNN [2, 10, 26, 41, 50, 51, 54] and LSTM Networks [5, 18, 35, 38, 53], as feature extractors to learn the representation of the input sensory segments automatically, then map the representation to labels using another neural network [3].”
MethodBi-STAN: bilinear spatial-temporal attention network for wearable human activity recognition
Gao, C., Chen, Y., Jiang, X., Hu, L., Zhao, Z., & Zhang, Y. (2023) · International Journal of Machine Learning and Cybernetics · doi:10.1007/s13042-023-01781-1 · Cited by 13
“In recent years, fruitful achievements have been made in natural language processing [9, 24], computer vision [25, 26], ubiquitous computing [27–31], and other aspects.”
BackgroundWashSpot: Real-Time Spotting and Detection of Enacted Compulsive Hand Washing with Wearable Devices
Burchard, R., Scholl, P., Lieb, R., Laerhoven, K. V., & Wahl, K. (2022) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3544793.3563428 · Cited by 13
Activity-Aware Deep Cognitive Fatigue Assessment using Wearables
Alam, M. A. U. (2021) · Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) · doi:10.1109/embc46164.2021.9630729 · Cited by 12
“2) Activity Recognition Module: We develop a two step multi-label activity recognition framework which consists of two LSTM with Consistency Self-Attention (LSTM-CSA) [36] models, (1) gestural activity recognition and (2) postural activity recognition.”
MethodInfluentialYun, S., Kim, H.-J., Ryu, J.-K., & Kim, S.-C. (2023) · Healthcare · doi:10.3390/healthcare11070940 · Cited by 16
“machine learning problem for recognizing fi ne-grained di ff erences 2. of machine-learning 3. and analyzed the machine-generated a tt ention vectors during the inference”
LSTM with Uniqueness Attention for Human Activity Recognition
Zheng, Z., Shi, L., Wang, C., Sun, L. J., & Pan, G. (2019) · Lecture notes in computer science · doi:10.1007/978-3-030-30508-6_40 · Cited by 9
Yu, J., Zhang, L., Cheng, D., Huang, W., Wu, H., & Song, A. (2024) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2024.3457830 · Cited by 15
“…70% data of all these volunteers is chosen to generate the training set, while the remaining 30% data is held out for validation and test; PAMAP2 [44]: To maintain consistency, we assign the first and second runs from participant 6 for the test set, and the first and second runs from participant…”
BackgroundMethodResultInfluentialYoshimura, N., Maekawa, T., & Hara, T. (2021) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn52387.2021.9533888 · Cited by 12
“Zeng et al. and Ma et al. visualized a model with an attention mechanism by drawing an attention map [3], [4].”
BackgroundBaseman, C. M., Fayfman, M., Schechter, M. C., Ostadabbas, S., Santamarina, G., Plötz, T., & Arriaga, R. I. (2023) · Journal of Diabetes Science and Technology · doi:10.1177/19322968231213378 · Cited by 15
Enhanced Prediction Model for Human Activity Using an End-to-End Approach
Park, S., Lee, H. O., Hwang, Y. M., Ko, S., & Lee, B. (2022) · IEEE Internet of Things Journal · doi:10.1109/jiot.2022.3223674 · Cited by 11
“Besides, to explicitly utilize sensor modalities, an attention module was used with DLMs in [30], [31], [32], [34], and [35].”
BackgroundMethodPlug-and-play multi-dimensional attention module for accurate Human Activity Recognition
Liang, J., Zhang, L., Bu, C., Yang, G., Wu, H., & Song, A. (2024) · Computer Networks · doi:10.1016/j.comnet.2024.110338 · Cited by 9
Distilled mid-fusion transformer networks for multi-modal human activity recognition
Li, J., Yao, L., Li, B., & Sammut, C. (2025) · Knowledge-Based Systems · doi:10.1016/j.knosys.2025.114160 · Cited by 14
The effect of hyperparameter search on artificial neural network in human activity recognition
Sütő, J. (2021) · Open Computer Science · doi:10.1515/comp-2020-0227 · Cited by 9
“[19] developed a RNN model (from Long Short-Term Memories – LSTM) for HAR because this type of model has a considerable pattern recognition capability in sequential data streams.”
MethodExploring Multimodal Fusion for Continuous Protective Behavior Detection
Cen, G., Wang, C., Olugbade, T., Williams, A. C. D. C., & Bianchi‐Berthouze, N. (2022) · International Conference on Affective Computing and Intelligent Interaction and workshops · doi:10.1109/acii55700.2022.9953851 · Cited by 11
“In [20], attention was used in an LSTM-based architecture to weight input data from multiple sensors at each single timestep as well as weighting the temporal sequence at the LSTM’s output layer.”
Huang, Q., Xie, W., Li, C., Wang, Y.-F., & Liu, Y.-W. (2023) · Applied Sciences · doi:10.3390/app131910560 · Cited by 7
Yao, M., Cheng, D., Zhang, L., Liu, L., Song, S. J., Wu, H., & Song, A. (2024) · Applied Soft Computing · doi:10.1016/j.asoc.2024.112444 · Cited by 8
SMOTE-Enhanced CNN-Bi-LSTM for wearable sensor-based human activity recognition
Arafa, A., Harfoush, H., El-Fishawy, N. A., & Radad, M. A. (2026) · Pervasive and Mobile Computing · doi:10.1016/j.pmcj.2026.102161 · Cited by 8
ALANet:Autoencoder-LSTM for pain and protective behaviour detection
Yuan, X., & Mahmoud, M. (2020) · International Conference on Automatic Face and Gesture Recognition · doi:10.1109/fg47880.2020.00063 · Cited by 12
“[15] applied temporal attention to the hidden layer of LSTM, filtered out the irrelevant parts of the data and applied sensor attention to the input layer of LSTM to obtain the information of important sensors.”
MethodFew-Shot and Weakly Supervised Repetition Counting With Body-Worn Accelerometers
Nishino, Y., Maekawa, T., & Hara, T. (2022) · Frontiers in Computer Science · doi:10.3389/fcomp.2022.925108 · Cited by 8
“An attention mechanism (Xiao et al., 2015; Zeng et al., 2018; Yu et al., 2019) is introduced into the detection block to focus on the target action because unrelated actions can be included within an input segment.”
BackgroundAttriNet: learning mid-level features for human activity recognition with deep belief networks
Nair, H., Tan, C. S., Zeng, M., Mengshoel, O. J., & Shen, J. P. (2019) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3341162.3345600 · Cited by 7
“Although deep learning approaches have been applied in activity recognition task [25, 32, 33, 35], the area is under-researched [18].”
BackgroundSelf-citationRobust Foot Motion Recognition Using Stride Detection and Weak Supervision-Based Fast Labeling
Hua, R., & Wang, Y. (2021) · IEEE Sensors Journal · doi:10.1109/jsen.2021.3075151 · Cited by 9
Optimal sensor channel selection for resource-efficient deep activity recognition
Leite, C. F. S., & Xiao, Y. (2021) · ACM/IEEE International Conference on Information Processing in Sensor Networks (IPSN) · doi:10.1145/3412382.3458278 · Cited by 13
Improving Wearable-Based Activity Recognition Using Image Representations
Guinea, A. S., Sarabchian, M., & Mühlhäuser, M. (2022) · Italian National Conference on Sensors · doi:10.3390/s22051840 · Cited by 9
“Details F 1 -Score Guan and Plötz [46] Ensembles of deep LSTM networks 0.9260 AttnSense [47] Attention with CNN and a GRU network 0.9310 Zeng et al. [36] LSTM + Continuous Temporal Attention 0.9381 DeepConvLSTM [7] Convolutional and LSTM network 0.9580 Our approach Image Representation of Time…”
BackgroundMethodInfluentialA lightweight deep learning with feature weighting for activity recognition
Ige, A. O., & Noor, M. H. M. (2022) · Computational Intelligence · doi:10.1111/coin.12565 · Cited by 10
“To the best of our knowledge, attention mechanism has been proposed in the literature to address this challenge, as seen in Ma et al. 16 where the authors proposed a GRU with attention mechanism, also in Zeng et al. 17 where the researchers proposed long short-term memory (LSTM) with attention, and in Gao et al. 18 wh…”
BackgroundHybrid Attention with CNN-BiLSTM and CBAM for Efficient Wearable Activity Recognition
Mekruksavanich, S., Jantawong, P., Phaphan, W., & Jitpattanakul, A. (2024) · Joint International Conference on Digital Arts, Media and Technology with ECTI Northern Section Conference on Electrical, Electronics, Computer and Telecommunications Engineering (ECTI DAMT & NCON) · doi:10.1109/ectidamtncon60518.2024.10480050 · Cited by 7
“For instance, Zeng et al. [25] introduced an LSTM network with attention that identified spatial and temporal dependencies in multimodal sensor signals.”
BackgroundWang, L., Zhang, L., Yu, J., Cheng, D., Yao, M., Wu, H., & Song, A. (2024) · IEEE Sensors Journal · doi:10.1109/jsen.2024.3510811 · Cited by 11
“Consistent with previous studies [34], we adhere to a protocol for data splitting: the first and second runs from Participant 6 are designated for the test set, the first and second runs from Participant 5 for the validation set, and the remaining data for the training set.”
InfluentialDeep Learning for Sensor-Based Activity Recognition: Recent Trends
Ahad, M. A. R., Antar, A. D., & Ahmed, M. (2020) · Intelligent systems reference library · doi:10.1007/978-3-030-51379-5_9 · Cited by 7
Freezing of Gait Detection Using Discrete Wavelet Transform and Hybrid Deep Learning Architecture
Thư, N. T., & Han, D. S. (2021) · International Conference on Ubiquitous and Future Networks · doi:10.1109/icufn49451.2021.9528547 · Cited by 9
“[10] propose a continuous attention-based long short-term memory network (LSTM) to adaptively focus on essential parts of the signal, thus, help to improve the detection accuracy.”
BackgroundMulti-Branch CNN GRU with attention mechanism for human action recognition
Verma, U., Tyagi, P., & Aneja, M. K. (2023) · Engineering Research Express · doi:10.1088/2631-8695/acd98c · Cited by 13
Beyond "Taming Electric Scooters": Disentangling Understandings of Micromobility Naturalistic Riding
Tabatabaie, M., He, S., Wang, H., & Shin, K. G. (2024) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3678513 · Cited by 10
“Conventional approaches [67, 86] usually derive correlations across the input features and the output model results, without interpreting the latent interactions of learning [41] with the physical aspects of the scenes.”
BackgroundDisagreement-based class incremental random forest for sensor-based activity recognition
Hu, C., Chen, Y., Hu, L., Yu, H., & Lu, D. (2021) · Knowledge-Based Systems · doi:10.1016/j.knosys.2021.108044 · Cited by 7
Ye, N., Zhang, L., Xiong, D., Wu, H., & Song, A. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3555985 · Cited by 7
“These efforts include developing lightweight network architectures [6], [15], network quantization [16], channel pruning [17], tensor decomposition [18], and channel skip-ping [19].”
InfluentialOptimizing Accuracy-Efficiency Trade-Offs of On-Device Activity Inference With Star Operation
Chen, G., Fu, Z., Sha, Y., Xiong, D., Zhang, L., Wu, H., & Song, A. (2025) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2025.3619549 · Cited by 6
“Zeng et al. [23] combined attention with CNNs to dynamically focus on critical temporal segments, while Zhang et al. [7] introduced multi-head convolutional attention to amplify discriminative sensor patterns.”
A Tailored Textile Sensor-based Wrap for Shoulder Complex Angles Monitoring
Zhang, R., Hu, Y., He, Y., Fang, S., Yu, B., Sun, X., & Wang, Q. (2023) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3594739.3610707 · Cited by 7
Identification of Human Movement Through a Novel Machine Learning Approach
Fahad, N., Sen, A., Jisha, S. S., Ahmad, S., Mokhlis, H., & Hossain, M. S. (2023) · Innovations in Power and Advanced Computing Technologies (i-PACT) · doi:10.1109/i-pact58649.2023.10434296 · Cited by 10
“They got accuracy for CNNs 91%, LSTMs 85.86%, and SVMs 84.07% [6].”
BackgroundInfluentialYao, M., Zhang, L., Cheng, D., Liu, J., Wang, L., Wu, H., & Song, A. (2024) · IEEE Sensors Journal · doi:10.1109/jsen.2024.3412395 · Cited by 7
“To ensure a fair comparison [44], we utilized sliding windows of length 171 with a 78% overlap to segment the dataset.”
MethodIncorporating Attention Mechanism Into CNN-BiGRU Classifier for HAR
Nafea, O., Abdul, W., & Muhammad, G. (2024) · IEEE Access · doi:10.1109/access.2024.3487860 · Cited by 7
“One recent attention mechanism was combined with a GRU network [6] and the other was combined with an LSTM network [7].”
BackgroundMethodSelf-relation attention networks for weakly supervised few-shot activity recognition
Deng, S., Guo, Z., Teng, D., Lin, B., Chen, D., Jia, T., & Wang, H. (2023) · Knowledge-Based Systems · doi:10.1016/j.knosys.2023.110720 · Cited by 7
IOTeeth: Intra-Oral Teeth Sensing System for Dental Occlusal Diseases Recognition
Hu, Z., Radmehr, A., Zhang, Y., Pan, S., & Nguyen, P. (2024) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3643516 · Cited by 12
“These works propose to model attention weights as learnable parameters for each sensor input, thus the model can dynamically adjust its emphasis by weighting the features of each sensor accordingly [39, 44, 90, 97, 99–101].”
BackgroundSpiroMask: Measuring Lung Function Using Consumer-Grade Masks
Adhikary, R., Lodhavia, D., Francis, C., Patil, R., Srivastava, T., Khanna, P., Batra, N., Breda, J., Peplinski, J., & Patel, S. (2023) · ACM Transactions on Computing for Healthcare · doi:10.1145/3570167 · Cited by 9
“• Classiication: We used a 1D Convolutional Neural Network (1D CNN) to classify speech, tidal breathing and noise [85, 89].”
MethodSheng, T., & Huber, M. (2025) · Sensors · doi:10.3390/s25134032 · Cited by 5
“Following prior work [64,73], sensor streams were downsampled from 100 Hz to 33.3 Hz and segmented into 5.12 s windows with a 1 s step size.”
InfluentialXiong, D., Zhang, L., Wang, S., Cheng, D., & Huang, W. (2025) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3729495 · Cited by 9
“Earlier researches have primarily concentrated on the use of traditional machine learning algorithms such as random forests, Naive Bayes, K-Nearest Neighbors (KNN), and Support Vector Machines (SVM) [4, 25] [47, 61, 64].”
Long kernel distillation in human activity recognition
Yao, M., Cheng, D., Zhang, L., Li, W., Wu, H., & Song, A. (2025) · Knowledge-Based Systems · doi:10.1016/j.knosys.2025.113397 · Cited by 6
From Wrist to Ankle: Understanding IMU Sensor Placement in Human Activity Recognition
Gouri Lakshmi, S., & Poulose, A. (2025) · Emerging Technologies for Intelligent Systems (ETIS) · doi:10.1109/etis64005.2025.10961378 · Cited by 8
“Similarly, Zeng et al. [15] employed LSTM networks for HAR and demonstrated that these models significantly outperform traditional methods, especially for complex activities that require long-term temporal dependency modelling.”
WeakCounter: Acceleration-based Repetition Counting of Actions with Weakly Supervised Learning
Nishino, Y., Maekawa, T., & Hara, T. (2021) · International Symposium on Wearable Computers · doi:10.1145/3460421.3480431 · Cited by 5
A Shallow Convolution Network Based Contextual Attention for Human Activity Recognition
Xu, C., Mao, Z., Fan, F., Qiu, T., Shen, J., & Gu, Y. (2023) · Lecture notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering · doi:10.1007/978-3-031-34776-4_9 · Cited by 12
Mekruksavanich, S., Jantawong, P., Phaphan, W., & Jitpattanakul, A. (2024) · Joint International Conference on Digital Arts, Media and Technology with ECTI Northern Section Conference on Electrical, Electronics, Computer and Telecommunications Engineering (ECTI DAMT & NCON) · doi:10.1109/ectidamtncon60518.2024.10480100 · Cited by 5
“For instance, an attention-based LSTM network [25] was proposed to effectively capture dependencies in multimodal sensor signals across both spatial and temporal dimensions.”
BackgroundLearning Bodily and Temporal Attention in Protective Movement Behavior DetectionPreprint
Wang, C., Peng, M., Olugbade, T. A., Lane, N., Williams, A. C. D. C., & Bianchi-Berthouze, N. (2019) · arXiv.org · doi:10.48550/arxiv.1904.10824 · Cited by 13
“In addition, we compare our work with the approach used in related HAR studies [19] [20] [21], where the sensor attention was computed before the extraction of temporal information.”
MethodInfluentialWang, L., Yan, Y.-C., Huang, D., Pan, Y.-Y., Chak, F.-C.-A., Luo, K., & Li, J.-Q. (2025) · IEEE transactions on consumer electronics · doi:10.1109/tce.2025.3540853 · Cited by 5
“To comprehensively compare the performance of the AHARNet, we compared it with various structural models, such as LSTM [35], [36], CNN [19], [34], ConvLSTM [20], Attention [37], [38], [39], and Transformer [22].”
Ensemble early exit network on human activity recognition using wearable sensors
Yu, J., Zhang, L., Cheng, D., Bu, C., Liu, L., Wu, H., & Song, A. (2025) · Computer Networks · doi:10.1016/j.comnet.2025.111409 · Cited by 6
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
Huan, R., Dong, G., Cui, J., Jiang, C., Chen, P., & Liang, R. (2025) · IEEE Sensors Journal · doi:10.1109/jsen.2025.3600835 · Cited by 6
“Zeng et al. [25] introduced two attention mechanisms, namely, time attention and sensor attention, to refine LSTM networks for behavior recognition.”
KATN: Key Activity Detection via Inexact Supervised Learning
You, X., Zhang, L., Yu, H., Yuan, M., & Li, X. (2021) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3494957 · Cited by 5
“The attention mechanism has also been utilized in HAR task [32][47][28] and weakly labeled tasks[39][31].”
MethodSemNet: Learning semantic attributes for human activity recognition with deep belief networks
Venkatachalam, S., Nair, H., Zeng, M., Tan, C. S., Mengshoel, O. J., & Shen, J. P. (2022) · Frontiers in Big Data · doi:10.3389/fdata.2022.879389 · Cited by 4
Self-citationYe, N., Zhang, L., Xiong, D., Wu, H., & Song, A. (2024) · IEEE Internet of Things Journal · doi:10.1109/jiot.2024.3458441 · Cited by 5
“Zeng et al. [9] and Yang et al. [5] have first proposed to apply CNNs in the HAR environment, which could automatically capture local invariant features through raw time series sensor signals collected from tri-axial accelerometers.”
InfluentialXia, Q., Wada, A., Yoshii, T., Namioka, Y., & Maekawa, T. (2022) · Lecture notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering · doi:10.1007/978-3-030-94822-1_26 · Cited by 3
Human Activity Recognition Using Deep Models and Its Analysis from Domain Adaptation Perspective
Gurov, N., Khan, A., Hussain, R., & Khattak, A. M. (2019) · Lecture notes in computer science · doi:10.1007/978-3-030-29852-4_15 · Cited by 4
Sequence-Information Recognition Method Based on Integrated mDTW
Sun, B., & Chen, C. (2024) · Applied Sciences · doi:10.3390/app14198716 · Cited by 2
Bharot, N., Verma, P., Vidyarthi, A., Gupta, D. K., & Breslin, J. G. (2025) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2025.3604710 · Cited by 2
Li, Z., Chen, G., Liu, X., Zhang, L., Sun, Q., Wang, K., Wu, H., & Song, A. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3650552 · Cited by 2
“In contrast, ResNet-SaF ( γ = 4), equipped with the SaF mechanism, achieves the highest accuracy of 96.18% under exactly the same C /γ channel configuration (16 , 24 , 32 , 40).”
InfluentialSelf-citationSenseMLP: a parallel MLP architecture for sensor-based human activity recognition
Li, W., Guo, J.-M., & Wu, H. (2024) · Multimedia Systems · doi:10.1007/s00530-024-01384-y · Cited by 1
“In [22, 23], researchers introduced temporal and sensor modality attentions coupled with LSTM networks, resulting in improved performance on benchmark human activity datasets.”
Su, J., Shao, H., Deng, X., & Jiang, Y. (2024) · Computers, materials & continua/Computers, materials & continua (Print) · doi:10.32604/cmc.2024.057604 · Cited by 3
Parallel Attention Based Network for Human Activity Recognition Using Wearable Devices
Xu, C., Fan, F., Ke, G., Guo, C., Wu, Q., & Shen, J. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-78201-5_29 · Cited by 3
Person Identification Based on Accelerations Sensed in Smartphones with LSTM
Takahashi, Y., Nakamura, K., Kamiyama, T., Oguchi, M., & Yamaguchi, S. (2021) · Journal of Information Processing · doi:10.2197/ipsjjip.29.707 · Cited by 2
“Zeng et al. proposed a method for recognizing human activities based on LSTM, continuous temporal attention, and continuous sensor attention [14].”
MethodDomain models for data sources integration in HAR
Hamidi, M., & Osmani, A. (2020) · Neurocomputing · doi:10.1016/j.neucom.2020.06.138 · Cited by 3
Pepsi, M. B. B., Kumar, N. S., Jeyashree, S., & Subitcha, M. (2024) · Computing · doi:10.1007/s00607-024-01392-w · Cited by 2
“Deep learning, [6] is a neural network with multiple hidden layers, a subtype of Machine Learning.”
Wolling, F., Kostolani, D., Trollmann, P., & Michahelles, F. (2025) · International Workshop on the Semantic Web · doi:10.1145/3715071.3750419 · Cited by 3
“Network architectures for HAR typically involve deep learning models spanning from recurrent neural networks [10, 50, 52], CNNs [23, 35, 47], and lately Transformers [12, 53].”
Pourhasan, M., & Fathi, A. (2026) · Journal of Supercomputing · doi:10.1007/s11227-026-08285-z
XAttentionHAR Ensemble: Leveraging Cross-Modal Attention for Enhanced Activity Recognition
Sahni, S., Jain, S., & Saritha, S. K. (2024) · International Journal of Pattern Recognition and Artificial Intelligence · doi:10.1142/s0218001424500265 · Cited by 1
Wang, C. (2019) · 2019 8th International Conference on Affective Computing and Intelligent Interaction Workshops and Demos (ACIIW) · doi:10.1109/aciiw.2019.8925091 · Cited by 1
“[22] proposed an attention-based LSTM framework, where a sensor attention module was used at input level for each timestep, with an additional temporal attention module at a deeper layer.”
MethodHaresamudram, H., Tang, C. I., Suh, S., Lukowicz, P., & Ploetz, T. (2024) · arXiv · Cited by 31
“Temporal attention was applied by Murahari et al. [130] whereas continuous sensor and temporal attention were evaluated by Zeng et al. [207], both leading to increased performance.”
ISWC 2018: Towards Smarter Wearable Technologies That Blend With Our Bodies
Peiris, R. L., Starner, T. E., Beigl, M., & Kunze, K. S. (2019) · IEEE Pervasive Computing · doi:10.1109/mprv.2019.2912252
“First, Zeng et al.(8) were nominated for their work on “Understanding and improving recurrent networks for human activity recognition by continuous attention.”
BackgroundA Minimal Social Weight Wearable Device for Thermal Regulation
Serrano, J. B., & Semwal, S. K. (2020) · Advances in intelligent systems and computing · doi:10.1007/978-3-030-39445-5_35
LEARNING EMBEDDINGS FOR WEARABLE-BASED HUMAN ACTIVITY ANALYSIS
Sheng, T. (2020) · Cited by 1
Sensing-HH: A Deep Hybrid Attention Model for Footwear Recognition
Yao, Y., Wen, Y., & Wang, J. (2020) · Electronics · doi:10.3390/electronics9091552
“[50] proposed two attention models for human activity recognition: temporal attention and sensor attention.”
BackgroundDescription of Structural Biases and Associated Data in Sensor-Rich EnvironmentsPreprint
Hamidi, M., & Osmani, A. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2104.04885
“Our approach can be related to that of [155], with the difference that selection is performed beforehand, in a separate procedure, before incorporating them into the main model.”
BackgroundMethodDetect the Interactions that Matter in Matter: Geometric Attention for Many-Body SystemsPreprint
Frank, T., & Chmiela, S. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.02549 · Cited by 2
“[37] Ming Zeng, Haoxiang Gao, Tong Yu, Ole J Mengshoel, Helge Langseth, Ian Lane, and Xiaobing Liu.”
BackgroundXie, W., Huang, Q., Li, C., Wang, Y., & Liu, Y. (2023) · ACM International Conference on Multimedia in Asia (MMAsia) · doi:10.1145/3595916.3626425
Feature representation learning for human activity recognition
Khaertdinov, B. (2024) · Research Publications (Maastricht University) · doi:10.26481/dis.20240618bk · Cited by 1
Liu, X., Zhang, L., Huang, W., Xiong, D., Fu, Z., Wu, H., & Song, A. (2025) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2025.3566808
“Because of the obvious advantages of recurrent neural networks (RNNs) in modeling time series signals, Zeng et al. [12] have integrated both of attention modules into recurrent backbone networks to highlight prominent temporal intervals and sensor modalities, hence offering better understandability…”
Self-citationMulti-scale Efficient Spatial Attention on Human Activity Recognition Using Wearable Sensors
Fan, J., Jindal, A., & Atapour–Abarghouei, A. (2025) · Communications in computer and information science · doi:10.1007/978-981-96-6963-9_11
Multi-Branch Cross-Receptive Field Network for Wearable Human Action Recognition
Liu, L., Ni, C., Bu, C., & Huang, Y. J. (2026) · Chinese Control and Decision Conference (CCDC) · doi:10.1109/ccdc69976.2026.11560634
“Unlike image data, raw sensor signals [9] require preprocessing, including noise removal, signal segmentation, and resampling.”
Kostolani, D., Wolling, F., Schlund, S., & Michahelles, F. (2026) · Frontiers of Computer Science · doi:10.3389/fcomp.2026.1873283
LiteTSC: A Lightweight Neural Network for Time Series Classification on Edge Devices
Wang, T., & Guo, Z. (2026) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn69854.2026.11729938
“For HAR, attention mechanisms have been applied to weight the importance of different time steps [15] and sensor channels [16].”
MoPFormer: Motion-Primitive Transformer for Wearable-Sensor Activity Recognition
Zhang, H., Zhuang, Z., Wang, X., Yang, X., & Zhang, Y. (2025) · arXiv · doi:10.48550/arxiv.2505.20744
“Zeng et al. [54] developed attention mechanisms from two perspectives: they first proposed sensor attention on the input to extract salient sensory patterns, and then applied temporal attention to LSTMs to filter out inactive data segments.”
Metadata
- Authors
- Zeng, M., Gao, H., Yu, T., Mengshoel, O. J., Langseth, H., Lane, I., & Liu, X.
- Affiliations
- Carnegie Mellon University (US)
- The Norwegian University of Science and Technology
- Google (United States) (US)
- Venue
- Proceedings of the 2018 ACM International Symposium on Wearable Computers
- Publisher
- ACM
- Published
- 2018-10-08
- Pages
- 56-63
- OpenAlex
- W2894702700
- Cited by
- 185
- References
- 30
- Citations by year
- 2026: 6 · 2025: 18 · 2024: 32 · 2023: 25 · 2022: 35 · 2021: 37 · 2020: 19 · 2019: 13
- FWCI
- 5.20
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- ACM International Symposium on Wearable Computers (ISWC) (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Context-Aware Activity Recognition Systems
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Human Pose and Action Recognition
- Physical Sciences › Engineering › Biomedical Engineering › Gait Recognition and Analysis
- Keywords
- human activity recognition, recurrent neural networks, attention mechanism, temporal attention, interpretability, noise robustness, continuity constraints
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.