Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition
Open accessThe DOI matches its Crossref record; 48 of the 49 references it cites were found in a registry or an index; 8 of the 23 sources its claims cite were checked for support.
Abstract
Introduces DeepConvLSTM, a deep learning architecture combining convolutional and LSTM recurrent units for multimodal wearable activity recognition, outperforming non-recurrent deep networks and handcrafted feature baselines by 4% to 9%.
Overview
CoChat Summary
TL;DR
The paper proposes DeepConvLSTM, a deep framework combining four convolutional layers with two LSTM recurrent layers for wearable activity recognition from raw multimodal sensor data. It reports improved F1 scores over a baseline CNN and prior results on the OPPORTUNITY and Skoda datasets, including 0.915 F1 for OPPORTUNITY gestures with the Null class and 0.958 F1 on Skoda.
Key findings
- On OPPORTUNITY, DeepConvLSTM achieved F1 scores of 0.930 for modes of locomotion without Null, 0.895 for modes with Null, 0.866 for gestures without Null, and 0.915 for gestures with Null; the baseline CNN scored 0.912, 0.878, 0.783, and 0.883, respectively. The paper reports a 6% average improvement over the best OPPORTUNITY challenge submissions and more than 9% improvement on gesture recognition without Null.
- On Skoda, DeepConvLSTM reached 0.958 F1, compared with 0.884 for the baseline CNN, 0.893 for CNN [43], and 0.861 for CNN [23], improving the best reported result by 6%.
- Multimodal fusion improved gesture recognition on OPPORTUNITY without Null: 0.689 F1 with accelerometers only, 0.611 with gyroscopes only, 0.745 with accelerometers + gyroscopes, 0.839 with accelerometers + gyroscopes + magnetic sensors, and 0.864 with the full OPPORTUNITY sensor set.
- DeepConvLSTM had 999,122 parameters versus 7,445,458 for the baseline CNN, a 600% increase for the baseline; training took 340.3 min versus 282.2 min, and classification took 6.68 s versus 5.43 s, with DeepConvLSTM classifying almost 15 min of data per second and achieving 900× real-time on the GPU used.
Methods
- DeepConvLSTM uses four convolutional layers with 64 feature maps each, followed by two dense LSTM layers with 128 cells and a softmax output; the baseline CNN shares the convolutional layers but uses two non-recurrent dense layers with 128 units. Models were trained supervised with cross-entropy loss, mini-batch gradient descent with RMSProp, learning rate 10e-3, decay factor 0.9, mini-batches of 100 segments, and dropout p=0.5 on dense-layer inputs.
- Inputs were sliding windows of 500 ms with 250 ms step size, using raw sensor data with minimal preprocessing: linear interpolation for missing values and per-channel normalization to [0,1]. Evaluation used weighted F1 to address class imbalance.
- Experiments used the OPPORTUNITY dataset (on-body sensors, 113 channels, 30 Hz; Task A: 5-class modes of locomotion/postures; Task B: 18-class sporadic gestures) and the Skoda dataset (one subject, 10 right-arm 3D accelerometers, decimated to 30 Hz, 10 manipulative gestures, about 3 h and about 70 repetitions per gesture). Analyses included comparison to baseline CNN and published results, confusion matrices, sensor-modality fusion, and hyperparameter evaluations of sequence length and number of convolutional layers.
Limitations
- The paper states that DeepConvLSTM has not yet been implemented on a wearable device; the GPU used clearly outperforms current high-end wearable computational power, and training is best envisioned server-side. It also states that pooling was omitted because the sliding-window setup limits downsampling, though pooling could be useful without that requirement, and that training time increases with the number of layers, creating a performance–computational budget trade-off.
- Evident from the text: Skoda evaluation uses one subject and only right-arm sensors, and OPPORTUNITY testing is limited to ADL4 and ADL5 for Subjects 2 and 3, which constrains generalizability. The text does not report cross-validation or statistical significance testing.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures



![Figure 4 Placement of on-body sensors used in the OPPORTUNITYdataset (left: inertial measurements units; right: 3-axis accelerometers) [7].](https://figures.cochat.ai/paper-figures/27c0e4b39d104fe2b5315587d736fd2d/deep-convolutional-and-lstm-recurrent-neural-networks-for-multimodal-wearable-figure-4.jpg)




Verification
Claims the paper makes
5 Valid · 5 Not valid · 13 Not checked · 99 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“The choice of the number of recurrent layers is made following the results presented in [33], where the authors showed that a depth of at least two recurrent layers is beneficial when processing sequential data.”
§ Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition > 3. Architecture > 3.1. DeepConvLSTM
What the sources say
- Supportedcites [33] Karpathy A. Johnson J. Li F.F. Visualizing and understanding recurrent networks Cornell Univ. Lab. 2015 arXiv:1506.02078 source
Section 4.1 compares recurrent networks with one, two, and three layers on two sequence datasets and states that a depth of at least two is consistently beneficial. This directly supports the attributed finding, within the paper’s tested settings. (LLM, conf=0.94)
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“This seems to indicate that although deep learning techniques are often employed with large amounts of data, (e.g., millions of frames in computer vision [22]), they may actually be applicable to problem domains where acquiring annotated data is very costly, such as in supervised activity recognition.”
§ Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition > 5. Results and Discussion > 5.4. Discussion
What the sources say
- Supportedcites [22] Pigou L. Oord A.V.D. Dieleman S. van Herreweghe M. Dambre J. Beyond Temporal Pooling: Recurrence and Temporal Convolutions for Gesture Recognition in Video arXiv Preprint 2015 arXiv:1506.01911 source
The citation is attached to the example of large-scale computer-vision data, not to the citing authors’ conclusion about costly annotation. Section 4.1 (page unavailable) reports 1,720,800 labeled video frames in the gesture-recognition dataset, supporting the cited example of data on the order of millions of frames. (LLM, conf=0.86)
hedgedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The baseline CNN also offers better results than the OPPORTUNITY submissions in the recognition of models of locomotion.”
§ Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition > 5. Results and Discussion > 5.1. Performance Comparison
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“From the results, we can also see how the model is able to learn from signals obtained from accelerometers, gyroscopes and magnetometers, fusing them without requiring any specific preprocessing.”
§ Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition > 6. Conclusions
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“(iv) it is applicable even if the gestures are longer than the observation window.”
§ Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition > 5. Results and Discussion > 5.4. Discussion
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: unclear — It is unclear why the paper says it holds.
84 more claims
Create a free account to see all 122 claims and to verify the 2 references still open.
- It may state two separate claims as one. 14
- Probably not a substantial claim. 23
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 doesn’t support the claim as written.
In the citing paper
“However, these complex network framework suffered from low ef ciency and can hardly meet real-time requirements in practice applications.”
From InnoHAR: A Deep Neural Network for Complex Human Activity Recognition (2019) · cites this paper as [19]
Checked by a reasoning model — may be wrong.
Experimental — accuracy not yet measuredThe cited source doesn’t support the claim as written.
In the citing paper
“However, due to the lack of filtration techniques applied to multimodal data, the system’s performance was low.”
From A Multimodal IoT-Based Locomotion Classification System Using Features Engineering and Recursive Neural Network (2023) · cites this paper as [15]
Checked by a reasoning model — may be wrong.
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“and yield better performances than pure CNNs or LSTM networks [12]”
From Comparison of Feature Learning Methods for Human Activity Recognition Using Wearable Sensors (2018) · cites this paper as [12]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“The proposed method outperforms other methods, such as those based on deep believe networks (DBNs) [10], SVM [10], and CNNs [13].”
From Deep Recurrent Neural Networks for Human Activity Recognition (2017) · cites this paper as [13]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“The popular DeepConvLSTM model [7] can only afford four, moderately-sized convolution layers to learn effective data representations, and two LSTM layers for sequence modeling.”
From Complex Deep Neural Networks from Large Scale Virtual IMU Data for Effective Human Activity Recognition Using Wearables (2021) · cites this paper as [7]
Experimental — accuracy not yet measured
20 more claims about this paper
Create a free account to see every claim other papers make about this one.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
§ Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition > 3. Architecture
“The length of the kernels is the same for every convolutional layer, being defined as Pl=5,∀l=2,⋯,5.”
§ Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition > 4. Experimental Setup > 4.2. Performance Measure
“Following the OPPORTUNITY challenge experimental setup, the length of the window is 500 ms, with a step size of 250 ms.”
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
“| Total | 996,800 +(128×nc)+nc | 996,800 +(128×nc)+nc | 7,443,136+(128×nc)+nc | 7,443,136+(128×nc)+nc |”
§ Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition > 3. Architecture > 3.3. Model Implementation and Training
Reported 996800 · recomputed 1037888
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
“| 6 | Wai,Waf,Wac,Wao: 7232×128 | 942,592 | W: 57,856×128 | 7,405,696 |”
§ Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition > 3. Architecture > 3.3. Model Implementation and Training
Reported 942592 · recomputed 3702784
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
“| 7 | Wai,Waf,Wac,Wao: 128×128 | 33,280 | W: 128×128 | 16,512 |”
§ Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition > 3. Architecture > 3.3. Model Implementation and Training
Reported 33280 · recomputed 131072
Citations
Cites
ImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155
Going deeper with convolutions
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298594 · Cited by 48,562 · cited 1×
“Other publications showed much deeper networks, such as “GoogLeNet”, which is a 27-layer deep neural network applied to image classification [45].”
BackgroundReducing the dimensionality of data with neural networks
Hinton, G. E., & Salakhutdinov, R. R. (2006) · Science · doi:10.1126/science.1127647 · Cited by 20,928
Speech recognition with deep recurrent neural networks
Graves, A., Mohamed, A.-R., & Hinton, G. (2013) · 2013 IEEE International Conference on Acoustics, Speech and Signal Processing · doi:10.1109/icassp.2013.6638947 · Cited by 8,893 · cited 1×
“Networks using LSTM cells have offered better performance than standard recurrent units in speech recognition, where they gave state-of-the-art results in phoneme recognition [20].”
BackgroundConvolutional networks for images, speech, and time series
LeCun, Y., & Bengio, Y. (1998) · HAL (Le Centre pour la Communication Scientifique Directe) · doi:10.5555/303568.303704 · Cited by 4,231
The class imbalance problem: A systematic study1
Japkowicz, N., & Stephen, S. (2002) · Intelligent Data Analysis · doi:10.3233/ida-2002-6504 · Cited by 3,295
DeepPose: Human Pose Estimation via Deep Neural Networks
Toshev, A., & Szegedy, C. (2014) · 2014 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2014.214 · Cited by 3,185 · cited 1×
“Deep CNNs have had a major impact on fields, like content recommendation [24], speech recognition [25] and in computer vision [26–28], where they have become the de facto standard approach.”
BackgroundOverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks
Sermanet, P., Eigen, D., Zhang, X., Mathieu, M., Fergus, R., & LeCun, Y. (2013) · arXiv · doi:10.48550/arxiv.1312.6229 · Cited by 2,680 · cited 1×
“Deep CNNs have had a major impact on fields, like content recommendation [24], speech recognition [25] and in computer vision [26–28], where they have become the de facto standard approach.”
BackgroundLee, H., Grosse, R., Ranganath, R., & Ng, A. Y. (2009) · Proceedings of the 26th Annual International Conference on Machine Learning · doi:10.1145/1553374.1553453 · Cited by 2,412 · cited 1×
“Deep learning techniques have outperformed many conventional methods in computer vision [13] and audio classification [14].”
MethodA review of wearable sensors and systems with application in rehabilitation
Patel, S., Park, H., Bonato, P., Chan, L., & Rodgers, M. (2012) · Journal of NeuroEngineering and Rehabilitation · doi:10.1186/1743-0003-9-21 · Cited by 2,396 · cited 1×
“, from simple hand gestures to complex activities, such as “cooking a meal”) and the context in which they occur from sensor data is at the core of smart assistive technologies, such as in smart homes [1], in rehabilitation [2], in health support [3,4], in skill assessment [5] or in industrial settings [6].”
Background
Research works printed but not linked: 1
- Gers F.A. Schraudolph N.N. Schmidhuber J. Learning precise timing with LSTM recurrent networks J. Mach. Learn. Res. 2003 3 115 143
Cited by
Items 2,301–2,400 of 2,807
Towards Smarter Activity Recognition: Leveraging Semantic Context in HAR Models
Pilatus, S., Strem, N., & Wichert, R. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-30038-6_7
Assessing Occupational Physical Workload Using IoT and Human Activity Recognition
Bouchekif, D. (2026) · Lecture notes in business information processing · doi:10.1007/978-3-032-35043-5_18
Learning the Group Structure of Deep Neural Networks with an Expectation Maximization Method
Yi, S., & Choi, J. (2018) · IEEE ... International Conference on Data Mining workshops · doi:10.1109/icdmw.2018.00106
“We compare our method with three baseline methods which are residual model [1], 1 regularization, Dynamic Network Surgery (DNS) [19], and additional convLSTM [28], [29] for the groundwater dataset in terms of the performance and the number of parameters.”
BackgroundMethodSelf-attention-based Human Activity Detection Using Wearable Sensors
Diwakar, S., Dwivedi, D., SINGH, S. P., & Sharma, M. K. (2023) · Lecture notes in electrical engineering · doi:10.1007/978-981-99-0969-8_66
Xu, Y., Li, J., Huang, Z., Cheng, T., Chen, C., Tan, B., Tan, Z., Chen, H., & Zhou, Y. (2026) · Frontiers in Sports and Active Living · doi:10.3389/fspor.2026.1858408
“Early deep architectures overcame some of these bottlenecks by combining convolutional and LSTM recurrent units to model the temporal dynamics of multimodal sensor activations (8).”
GenIMU: Simulate Realistic IMU Data Using Video Generation Models
Rieger, N., Bert Arnrich, A., & Konak, O. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-13312-0_25
Jeon, Y., Lee, J., Kim, T.-G., & Baek, Y. (2018)
Khan, N., Davidson, L., Sarkar, N., & Cole, K. R. (2026) · Digital Health · doi:10.1177/20552076261444097
“45 This automated approach has been shown to be more precise in HAR than traditional methods and has strengthened the generalization capabilities of recognition models.”
Li, H. (2019) · theses.fr (ABES) · doi:10.70675/3aa9be79z54f8z44bazabacz38bf1a61173f
“The work [86] proposed convolutional recurrent neural network to recognize daily activity; their algorithm gained an improvement of 6% compared to the stateof-the-art works.”
StrengthSense: A Dataset of IMU Signals Capturing Everyday Activities with Diverse Physical Demands
Yang, Z., Liu, Z., Zhang, Y., Leite, C. F. S., & Xiao, Y. (2026) · ACM International Symposium on Wearable Computers (ISWC) · doi:10.1145/3830727.3834811 · Cited by 2
Zhou, L., & Wang, P. (2026) · Systems and Soft Computing · doi:10.1016/j.sasc.2026.200641
Human Activity Tracking Using Mobile Sensor Data and an Optimised LSTM
Anandarao, S., Mastani, S. A., Geeta, T., Vamsi, B., & Reddy, P. S. S. (2024) · Lecture notes in networks and systems · doi:10.1007/978-981-97-0573-3_25
Effect of Sensor Positions on Wearable Sensor-Based Fall Detection Using Deep Residual Network
Mekruksavanich, S., & Jitpattanakul, A. (2025) · IEEE International Conference on Software Engineering and Service Science (ICSESS) · doi:10.1109/icsess67729.2025.11380725 · Cited by 1
“In parallel, recurrent neural networks (RNNs)—especially long short-term memory (LSTM) and gated recurrent unit (GRU) variants—excel at capturing temporal dependencies inherent in time-series data [10].”
Human activity recognition with a multibranch network based on CNN and LSTM
Yuan, R., Zhang, Y., Wang, L., & Li, S. (2024) · International Conference on Computer Technology, Information Engineering, and Electron Materials (CTIEEM) · doi:10.1117/12.3023366
“The sensor data comprises three-axis accelerometer and gyroscope data, sampled at a frequency of 50 Hz [12].”
MethodVyšata, O., Matyáš, D., Chandrashekhar, R. K., & Procházka, A. (2025) · IEEE International Conference on Digital Signal Processing (DSP) · doi:10.1109/dsp65409.2025.11074893
“Recent advances in machine learning and deep learning have opened new frontiers in the automated analysis of IMU-derived gait data [6].”
Context Analysis and Estimation of Mobile Users Considering the Time Series of Data
SHIMIZU, H., Suganuma, M., & Kameyama, W. (2020) · IEEE Global Conference on Consumer Electronics (GCCE) · doi:10.1109/gcce50665.2020.9292016
“The shorthand notation of the architectures following [4] are used in Table III to show all the parameters, where Cs(n) denotes a Conv with n feature maps and s stride, and F(n) denotes a Fc with n units.”
MethodConvolutional Neural Networks for Human Activity Recognition
Nurwulan, N. R. (2026) · Advances in Science, Technology & Innovation/Advances in science, technology & innovation · doi:10.1007/978-3-032-24784-1_24
An Intelligent Monitoring System for Sheep Behavior Based on ActiGraph Sensors
Ghadir, S., Ghadir, D., Berhe, T. M., Adami, D., Giordano, S., Pagano, M., Rossi, P., Sotgiu, F. D., Mossa, F., & Berlinguer, F. (2026) · Network · doi:10.3390/network6020031
“This hybrid design is widely adopted in sensor-based activity recognition, as it leverages both short-term feature learning and long-term temporal modeling [19,20].”
Self-supervised Deep Learning for Reading Activity ClassificationPreprint
Islam, M. R., Sakamoto, S., Yamada, Y., Vargo, A. W., Iwata, M., Iwamura, M., & Kise, K. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2012.03598
“3 Self-supervised Deep Learning In the past decade, the development and application of DL has successfully solved many problems in the eld of ubiquitous computing [17, 62, 10, 49], pervasive intelligence [61, 75], health [47, 18, 25], and well-being [59, 37].”
BackgroundChen, S., Wu, S. K., Zhu, L., & Yang, H. (2022) · Lecture notes in computer science · doi:10.1007/978-3-030-95384-3_7
AutoResearcHAR: 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
Koo, B., Bilau, I., Rodriguez, A. D., Yang, E., & Kwon, H. (2025) · medRxiv · doi:10.1101/2025.05.24.25328107
“The Deep Convolutional LSTM (DeepConvLSTM) model [51] was also employed as a deep learning baseline, applied directly to raw sensor time series instead of extracted features.”
Alotaibi, R. S., Reyad, O., & Karar, M. E. (2026) · International Journal of Telemedicine and Applications · doi:10.1155/ijta/1465162
“A combined CNN-LSTM classifier has been proposed in [40], achieving 93.80% accuracy.”
InfluentialContext-Aware Informative Sample Selection and Image Forgery Detection
Bappy, J. H. (2018) · eScholarship (California Digital Library)
Pappa, L., Karvelis, P. S., & Stylios, C. D. (2026) · Expert Systems with Applications · doi:10.1016/j.eswa.2026.133478
Ruzimboev, K. R., Avezmatov, I., Yusupov, S., Sattarova, S., Abdrimova, F., & Umar, B. (2026) · International Siberian workshops and tutorials on electron devices and materials · doi:10.1109/edm69524.2026.11631980
(2021) · doi:10.2196/preprints.30135 · Cited by 9
Deep Learning-Based Human Activity Recognition Using Smartphone Sensor
Singh, I., Anuragi, V., Jain, T., & Singh, T. (2025) · IEEE World Conference on Applied Intelligence and Computing (AIC) · doi:10.1109/aic66080.2025.11211906
“HAR systems create essential value for both health surveillance activities and home automation services and fitness monitoring applications and behavioral assessment tasks [4].”
InfluentialFramework for Human Activity Recognition on Smartphones and Smartwatches
Mitrevski, B., Petreski, V., Gjoreski, M., & Stojkoska, B. R. (2018) · Communications in computer and information science · doi:10.1007/978-3-030-00825-3_8 · Cited by 2
Gao, Y. H., Long, Y., Guan, Y., et al. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1902.08068
“osition, outliers, etc. 4.2 Baselines We compare the proposed DPD method to a range of alternative, conventional machine learning methods. Despite popularity, existing deep learning techniques (e.g., [22, 24, 39, 51, 52]) are not applicable to this problem because these methods need training data with sample/window-wi…”
BackgroundMethodInfluentialMekruksavanich, S., & Jitpattanakul, A. (2026) · Lecture notes in networks and systems · doi:10.1007/978-3-032-26217-2_10 · Cited by 1
Behavior recognition based on enhanced human pose estimation
Xu, Z., & jun, L. (2024) · International Conference on Image Processing and Artificial Intelligence (ICIPAI) · doi:10.1117/12.3035498
“The RNN [5] /LSTM [6] /GRU [7] network is used for extracting temporal information; 3D CNN [8] is a combination of temporal information and spatial dimension information extraction; GNN [9] /GCN [10] (about relationship extraction between nodes), transformer and other networks belong to backbone…”
Impact of Deep Learning Techniques in IoT
Vadhana, M. C., Bala, P. S., & Ramdinthara, I. Z. (2021) · Wiley eBooks · doi:10.1002/9781119761655.ch10
Adapting Pretrained Large Vision Models for Sensor-based Activity Recognition
Cai, Y.-Z., Feng, R.-X., Cai, K.-L., Liu, Y.-H., Guo, B.-S., & Hong, Z.-Q. (2026) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3832003
“With the advancement of deep learning techniques, sensor-based human activity recognition (HAR) has achieved significant performance improvements through a wide range of approaches, including CNN-based methods [43, 79], RNN-based methods [77], and attention-based methods [58, 112].”
Real-Time Human Activity Recognition on Embedded Equipment: A Comparative StudyPreprint
Houda, N., Lohr, C., & Leduc, B. (2023) · SSRN Electronic Journal · doi:10.2139/ssrn.4643597
Nandi, D., Singh, P. K., & Chowdhury, C. (2026) · Multimedia Tools and Applications · doi:10.1007/s11042-026-21373-9 · Cited by 1
“…results depicted an overall good performance Study Accuracy O’Halloran et al. [ 30] 0.91 Ha et al. [ 31 ] 0.92 Lyu et al. [ 32] 0.96 Ordóñez et al. [33] 0.94 Proposed FEST Architecture 0.98 where the overall accuracy achieved were 94.00%, 84.04% and 98.12% for UCI-HAR, KU-HAR and MHEALTH,…”
InfluentialKaithal, P. K., Sengar, A. S., Chourey, A., Gupta, C., Pandey, A., & Gangrade, S. (2025) · World Skills Conference on Universal Data Analytics and Sciences (WorldSUAS) · doi:10.1109/worldsuas66815.2025.11199300
“[4] [9]”
InfluentialDeep Learning for Fire and Smoke Detection in Outdoor Spaces
Favorskaya, M. N., & Jain, L. C. (2021) · Smart innovation, systems and technologies · doi:10.1007/978-981-33-4619-2_15
Gao, Y., & Liu, Z. (2026) · International Journal of e-Collaboration · doi:10.4018/ijec.420872
Wieland, F., & Nigg, C. (2022) · doi:10.2196/preprints.42337 · Cited by 8
Chen, X., & Zhao, F. (2026) · Internet Technology Letters · doi:10.1002/itl2.70292
Gonzalez, D. R., & Qu, Y. (2022) · European Journal of Electrical Engineering and Computer Science · doi:10.24018/ejece.2022.6.2.433
“In recent years, with the rapid development of deep learning technology, the classification performance of human activity recognition based on deep learning networks has increased substantially [8].”
BackgroundContinual Egocentric Activity Recognition via Foreseeable-Generalized Multi-Modal RepresentationsPreprint
He, C., Cheng, S., Qiu, Z., Xu, L., Meng, F., Wu, Q., & Li, H. (2023) · doi:10.36227/techrxiv.24041583
Shkel, A. M., & Wang, Y. (2021) · doi:10.1002/9781119699910.ch1
Aziz, K., Dua, S., & Gupta, P. (2023) · International Conference on Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES) · doi:10.1109/icses60034.2023.10465456 · Cited by 1
“Then, the machine learning algorithm is selected and various tests are carried out to evaluate the potency of the model [15].”
BackgroundMethodNugroho, H. A., & Wicaksana, H. S. (2019) · Jurnal Teknologi dan Sistem Komputer · doi:10.14710/jtsiskom.7.4.2019.166-171
“Data menggunakan akselerometer smartphone ataupun perangkat yang digunakan pada manusia seperti smart watch dan sport bracelet yang mempunyai akselerometer [7]-[9].”
BackgroundA Deep Transfer Learning Approach to Support Opportunistic Wearable Activity Recognition
Baños, O., Gil, D., Medina-Quero, J., Sanchez, A., & Villalonga, C. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-43085-5_37
Prediction of Blast Furnace Gas Utilization Rate Based on CNN-GRU
Chu, L. M., & Cui, G. M. (2023) · Guangdong-Hong Kong-Macao Greater Bay Area Artificial Intelligence and Big Data Forum (AIBDF) · doi:10.1145/3660395.3660484
“Therefore, Convolutional Neural Networks (CNN) are introduced to enhance the feature extraction capability of blast furnace data [9-11].”
BackgroundLearning Multi-level Features For Sensor-based Human Action RecognitionPreprint
Xu, Y. L., Shen, Z., Zhang, X., et al. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1611.07143
“Instead of dealing with sparsely labeled cases, MLPL proposed in this paper implements MIL to learn discriminative latent patterns of actions, by which high-level features would be acquired.”
MethodAbe, H., Hino, T., Sugihara, M., Ikeya, H., & Shimosaka, M. (2020) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros45743.2020.9341500
“From this characteristic, LSTM has been frequently applied to activityrecognition researches in recent years [8], [9], [7].”
InfluentialKnee Injured Recovery Analysis Using Extreme Learning Machine
Ferreira, J. P., Ribeiro, B., Vieira, A., et al. (2020) · Communications in computer and information science · doi:10.1007/978-3-030-42520-3_6
Multimodal Smart-Skin for Real-Time Sitting Posture Recognition with Cross-Session Validation
Mutiara, G. A., Alfarisi, M. R., Mayadewi, P., Meisaroh, L., & Periyadi, P. (2026) · Multimodal Technologies and Interaction · doi:10.3390/mti10040039
Namura, N., Karakida, R., Kaneishi, D., Tada, M., Komuta, M., Sun, X., & Kanoga, S. (2025) · Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) · doi:10.1109/embc58623.2025.11253053
Multi-modal bearing fault diagnosis method fusing vision and complementary timing features
Zhang, X., Wang, R., & Zhang, L. (2026) · Proceedings of the Institution of Mechanical Engineers Part C Journal of Mechanical Engineering Science · doi:10.1177/09544062261421260 · Cited by 2
An Approach to Classify Human Finger Gestures Based on UWB Antenna Parameters
Tiwari, B., Balyan, V., Gupta, S. H., & Ojha, M. K. (2026) · International Journal of RF and Microwave Computer-Aided Engineering · doi:10.1155/mmce/8836225
Human Activity Recognition Using Feedforward Backpropagation Neural Network
Debs, F. A., Jurdy, M. E., Fares, E., Haidar, G. A., Owayjan, M. J., & Achkar, R. (2024) · International Conference on Innovation and Intelligence for Informatics, Computing, and Technologies (3ICT) · doi:10.1109/3ict64318.2024.10824544 · Cited by 1
Zhang, X., Ding, M., Teng, T., Wan, J., & Ding, H. (2026) · Sensors · doi:10.3390/s26123803
“As the main feature extractor, we first use a multi-layer stacked convolutional network [16].”
InfluentialShift-Aware Operator Selection for Label-Free Test-Time Calibration in Cross-Domain Dataset
Deshagouni, M., & Akther, S. (2026) · IEEE International Conference on Big Data Computing Service and Applications (BigDataService) · doi:10.1109/bigdataservice70481.2026.00051
“Deep learning has become the dominant modelling paradigm for this task, with convolutional, recurrent, attention-based, and hybrid architectures achieving strong within-dataset accuracy on standard benchmarks [6]–[9].”
Wang, Q., Ma, J., Fu, M., Wang, J., Bai, Y. B., Jiang, Z., & Li, H. (2026) · iScience · doi:10.1016/j.isci.2026.115252
Hassan, M., Liying, C. A. O., Rai, H. M., & Ghani, A. (2026) · Information Fusion · doi:10.1016/j.inffus.2026.104779
Zhou, Q., Zhang, Y., Imani, F., & Tang, J. (2026) · Journal of Manufacturing Systems · doi:10.1016/j.jmsy.2026.09.002
Downsampling wearable sensor data packets by measuring their information value
Belmonte, M., Casson, A. J., & Peek, N. B. (2019) · Proceedings of IEEE Sensors/Proceedings of IEEE Sensors ... · doi:10.1109/sensors43011.2019.8956721 · Cited by 1
Zhou, H. (2026) · IEEE Access · doi:10.1109/access.2026.3734272
“Convolutional-recurrent hybrids such as DeepConvLSTM [12] and their attention-augmented successors — multilevel residual attention [13], channel–temporal dual attention [14], and multi-level fusion for the Internet of Healthcare Things [15] — have raised in-distribution accuracy considerably;…”
An Empirical Evaluation of Model Compression Techniques for On-Device Human Activity Recognition
Erivwode, O. T., Bali, A., & Lago, P. (2026) · Conference proceedings - Canadian Conference on Electrical and Computer Engineering · doi:10.1109/ccece68150.2026.11610122
A Standardized Benchmark for Skeleton-Based Rehabilitation Assessment Using Deep Learning
Ismail-Fawaz, A., Devanne, M., Berretti, S., Weber, J., & Forestier, G. (2026) · International Conference on Automatic Face and Gesture Recognition · doi:10.1109/fg67764.2026.11556986 · Cited by 6
Autenticação Contínua Usando Sensores Inerciais dos Smartphones e Aprendizagem Profunda
Paz, I. J. V., Bragança, H. L. D. S., & Souto, E. (2022) · Anais do XXII Simpósio Brasileiro de Segurança da Informação e de Sistemas Computacionais (SBSeg 2022) · doi:10.5753/sbseg.2022.225312
“Por essa razão, este trabalho usa uma arquitetura de rede neural profunda proposta por Ordóñez and Roggen (2016), denominada de DeepConvLSTM, conforme mostra a Figura 2.”
BackgroundAn active smartphone authentication method based on daily cyclical activityPreprint
Mi, C., Xu, R., Lin, C., & Meng, R. Y. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1909.00045
Sun, J., Ru, J., Chen, Z., Qi, F., Ramos‐Mucci, L., Chen, S., Cribbs, A. P., Deng, L., & Wang, X. (2022) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2022.06.30.498304
“In order for maximization of method performance specific to the regulatory effect prediction, we attempted to seek prime solutions from the computer-vision and speech-recognition fields by utilizing both well-established models including AlexNet [44], MobileNetV2 [45], Transformer-based ConvMixer [46], ResNet18 [47],…”
MethodHybrid-DP Diffusion for Privacy-Preserving Wi-Fi CSI Sensing and Biometric Template Generation
Hu, B., Lu, C., Kwak, D., Lei, J., Xu, M., You, J., & Feng, Y. (2026) · IEEE Sensors Journal · doi:10.1109/jsen.2026.3690368
“Leveraging existing Wi-Fi infrastructure, CSI enables applications such as human activity recognition (HAR), gesture sensing, and indoor monitoring without requiring specialized hardware or wearable devices [1], [2].”
Guo, L., Wang, J., Wu, Q., Li, X., Zhang, B., Zhou, L., & Xiong, D. (2022) · doi:10.2196/preprints.40416 · Cited by 48
Toward Wearable Sensor-Based Human Activity Recognition: A Survey
Zou, H., Chen, Z., Pan, Y., & Li, J. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2026.3702134 · Cited by 3
“A representative hybrid model, the DeepConvLSTM by Ord ´o˜nez and Roggen [130], automatically learns feature representations through CNN layers and models temporal dependencies via LSTM.”
AN ADVANCED REVIEW ON HUMAN ACTIVITY RECOGNITION USING ARTIFICIAL INTELLIGENCE
Reddy, A. S. (2025) · Journal of Science Engineering Technology and Management Sciences · doi:10.63590/jsetms.2025.v02.i07.pp517-525
Overmind: Fast and Scalable Decentralized Learning Simulation for Mobile Environments
Lee, S., Yu, H., Zheng, X., & Julien, C. (2024) · IEEE International Conference on Mobile Ad-Hoc and Sensor Systems (MASS) · doi:10.1109/mass62177.2024.00033
“We utilize the DeepConvLSTM [2] model for OP, which consists of convolutional layers and LSTM recurrent units.”
MethodWISNet: A Deep Neural Network based Human Activity Recognition SystemPreprint
Sharen, H., Rukmani, P., Anbarasi, L. J., & Gandomi, A. H. (2022) · Research Square · doi:10.21203/rs.3.rs-1884556/v1
Latent Markov Chain Encoding for Abnormal Landing Event Detection
He, D., Deng, L., Sui, Y., et al. (2019) · International Conference on Electronic Information Technology and Computer Engineering (EITCE) · doi:10.1109/eitce47263.2019.9094793
Gong, F., & Yang, F. (2025) · Research Square · doi:10.21203/rs.3.rs-6309936/v1
Cross-Person Human Activity Recognition Method Based on the HARG Model
Zhang, X., Xin, P., Zhang, J., Liu, M., Yue, H., & Hou, M. (2023) · International Conference on Computer, Big Data and Artificial Intelligence (ICCBD+AI) · doi:10.1109/iccbd-ai62252.2023.00050
“Currently, the numerous models based on CNN, LSTM, and their combinations have been proposed [4], [5].”
BackgroundMethod(2021) · Elsevier eBooks · doi:10.1016/b978-0-12-820581-5.00017-1
A review on devices and learning techniques in domestic intelligent environment
Ye, J., Wang, M., Zhong, J., & Jiang, H. (2024) · Journal of Ambient Intelligence and Humanized Computing · doi:10.1007/s12652-024-04759-1
“Ord´on˜ez and Roggen (2016) firstly combined CNN and LSTM in a hybrid DL model, in which the convolution layers can capture local features and transmit them into LSTM layers to obtain temporal information.”
Deep Interaction Feature Fusion for Robust Human Activity Recognition
Oh, Y., Kim, S., & Bui, A. A. T. (2024) · Communications in computer and information science · doi:10.1007/978-981-97-9003-6_7 · Cited by 1
Moreau, H. (2021) · theses.fr (ABES) · doi:10.70675/1b594f01z9f35z47d2zb83fz69b74ff72897
Ghandi, M., Ismail, M., & Blaisdell, M. (2021) · ACADIA quarterly · doi:10.52842/conf.acadia.2021.100 · Cited by 1
A Computational Pipeline for Activity PredictionUsing Wearable Sensor DataPreprint
Chuah, J., Vancells-Lopez, L., Loya, A. K., et al. (2026) · Research Square · doi:10.21203/rs.3.rs-9359112/v1
“This approach is consistent with prior work in gait analysis, where individual steps are commonly treated as the unit of analysis despite underlying sequential structure [23, 24].”
Archana, M., Kavitha, S. N., & Vathsala, A. V. (2025) · Research Square · doi:10.21203/rs.3.rs-6124369/v1
Analysing Human Cognitive Proficiency Using Face Recognition in Deep Learning
Prasad, S., & Khilji, N. (2026) · Lecture notes in networks and systems · doi:10.1007/978-3-032-32476-4_30
Celik, Y., Kuduz, H., Engin, F. V., Kacar, F., Tarakçı, E., Pearson, L. T., Das, J., Stuart, S., Woo, W. L., & Godfrey, A. (2026) · Journal of NeuroEngineering and Rehabilitation · doi:10.1186/s12984-026-01910-1 · Cited by 4
“DL architectures such as Convolutional Neural Networks (CNNs) and hybrid CNN-Long Short-Term Memory (LSTM) models, have shown very high accuracy on benchmark datasets by learning complex spatial and temporal features from IMU data [15, 16].”
Rakhimova, L. S. (2026) · ITECX Journal International Academic Journal · doi:10.68209/mescwd5
A Multi-Scale CNN-Hierarchical RNN Based Network for Human Gait Recognition Using Smartphone Sensors
Das, S., & Sahoo, P. R. (2025) · IEEE International Conference on Applied Electromagnetics, Signal Processing, & Communication (AESPC) · doi:10.1109/aespc67542.2025.11326877 · Cited by 1
Miękisiak, G., Szczepański, T., & Palczewska, A. (2026) · Italian National Conference on Sensors · doi:10.3390/s26196102
On-Chip Sensors Commercialized for Medical Diagnostics Applications
Shukla, V. K., & Tanushri (2026) · Springer eBooks · doi:10.1007/978-981-95-5519-2_14
A Comprehensive Review on Human Activity and Fitness Tracker using Different Approaches
Mishra, Y. (2025) · International Journal for Research in Applied Science and Engineering Technology · doi:10.22214/ijraset.2025.70544
Transfer learning for wearable computers
Akbari, A., Alinia, P., Ghasemzadeh, H. G., & Jafari, R. (2020) · Elsevier eBooks · doi:10.1016/b978-0-12-819246-7.00016-4
Mekruksavanich, S., Mathiang, K., Phaphan, W., & Jitpattanakul, A. (2026) · International Conference on Big Data Analytics and Practices (IBDAP) · doi:10.1109/ibdap69915.2026.11677998
“…patterns are extracted effectively by convolutional neural networks (CNNs), while recurrent architectures—long short-term memory (LSTM) networks and gated recurrent units (GRUs) in particular—capture dependencies across time steps, matching the dynamic character of human movement [5]–[9].”
Qiao, H., Rizk, H., & Maekawa, T. (2025) · Companion of the 2025 ACM International Joint Conference on Pervasive and Ubiquitous Computing · doi:10.1145/3714394.3754402
Su, S., Hu, M., Wei, L., & Yan, Z. (2026) · Biomedical Signal Processing and Control · doi:10.1016/j.bspc.2026.109699
Adaptive Multi-state Pipe Framework Based on Set Pair Analysis
Shi, L., Xu, H., Zhang, B., Sun, X., Li, J., & Fan, S. (2020) · International Journal of Machine Learning and Computing · doi:10.18178/ijmlc.2020.10.6.1002
“[15] proposed a general deep activity recognition framework based on convolution and Long Short-term Memory (LSTM) cycle units.”
MethodGonzález-Cañete, F. J., & Casilari, E. (2026) · Internet of Things · doi:10.1016/j.iot.2026.101969
Heading Judgment for the Waist-Mounted MIMU Using LSTM
Gu, S., & Guo, C. X. (2019) · IEEE International Conference on Parallel and Distributed Systems (ICPADS) · doi:10.1109/icpads47876.2019.00140
“Stack two LSTM: At present, there are many deep learning networks, CNN, RNN, C-RNN, etc.”
MethodInfluentialA Hybrid Domain Adaptation-Based Method for State of Health Prediction of Lithium-Ion Batteries
Liu, B., Xu, J., & Xia, W. (2023) · Lecture notes in electrical engineering · doi:10.1007/978-981-99-1027-4_73 · Cited by 1
Development of a Holistic System for Activity Classification Based on Multimodal Sensor Data
Rolletschke, T., Thomanek, R., Roschke, C., & Ritter, M. (2018) · Electronic workshops in computing · doi:10.14236/ewic/hci2018.167
Metadata
- Authors
- Ordóñez, F., & Roggen, D.
- Affiliations
- University of Sussex (GB)
- Corresponding author
- Ordóñez, Francisco
- Venue
- Sensors
- Publisher
- MDPI AG
- Published
- 2016-01-18
- Volume
- 16
- Issue
- 1
- Pages
- 115
- ISSN
- 1424-8220 (ISSN-L)
- PMC
- PMC4732148
- OpenAlex
- W2270470215
- Cited by
- 2,761
- References
- 49
- Citations by year
- 2026: 284 · 2025: 256 · 2024: 284 · 2023: 270 · 2022: 319 · 2021: 379 · 2020: 347 · 2019: 288 · 2018: 219 · 2017: 93 · 2016: 21
- FWCI
- 140.52
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Sensors (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- Sussex Research Online (University of Sussex) (submittedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Figshare (submittedVersion, cc-by)
- Funders
- Google Faculty Research Award (Is deep learning useful for wearable activity recognition?)
- EPSRC (EP/N007816/1)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Context-Aware Activity Recognition Systems
- Physical Sciences › Engineering › Biomedical Engineering › Gait Recognition and Analysis
- Physical Sciences › Computer Science › Signal Processing › Time Series Analysis and Forecasting
- Keywords
- human activity recognition, wearable sensors, multimodal sensor fusion, convolutional neural networks, LSTM neural network, temporal dynamics, automatic feature extraction, hyperparameter optimization
- MeSH terms
- Machine Learning, Clothing, Humans, Human Activities, Signal Processing, Computer-Assisted, Databases, Factual, Neural Networks, Computer, Monitoring, Ambulatory
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.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
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.