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 2,001–2,100 of 2,807
Estimating Frontal Body Landmarks from Thermal Sensors Using Residual Neural Networks
Polo-Rodríguez, A., Lupión, M., Ortigosa, P. M., et al. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-07704-3_27 · Cited by 2
Martin, D., Roberts, D. L., & Bozkurt, A. Y. (2024) · Animals · doi:10.3390/ani14233403 · Cited by 2
“Previous experiments have shown that deep convolutional LSTMs have been successful for multimodal activity recognition using accelerometry [28,29].”
A Survey of Motion Data Processing and Classification Techniques Based on Wearable Sensors
Xiaoqiong, X., Xuemei, 熊. X., Keda, Z., & Chao, L. (2023) · IgMin Research · doi:10.61927/igmin123 · Cited by 2
Zhang, P., Shao, X., Zhao, Y., & Zhao, F. (2026) · Expert Systems · doi:10.1111/exsy.70190 · Cited by 2
Jeong, J., Oh, Y., Kim, D., & Yoon, H. (2026) · Knowledge-Based Systems · doi:10.1016/j.knosys.2025.115251 · Cited by 4
Recurrent Control Nets as Central Pattern Generators for Deep Reinforcement Learning
Liu, V., Adeniji, A., Lee, N., et al. (2019) · Cited by 2
Smart Wearable Device for Elderly Dementia People
Surendran, D., & Kiruthika, V. R. (2020) · International Journal of Engineering and Advanced Technology · doi:10.35940/ijeat.c6338.049420 · Cited by 2
Sensor-Based Human Activity Recognition for Elderly In-patients with a Luong Self-Attention Network
Nithin, G. R., Chhabra, M., Hao, Y., Wang, B., & Zheng, R. (2021) · IEEE/ACM Conference on Connected Health: Applications, Systems and Engineering Technologies · doi:10.1109/chase52844.2021.00019 · Cited by 3
Pragmatic Big Data and smart manufacturing
Benhenni, A. L. (2017) · International Congress of Metrology (CIM) · doi:10.1051/metrology/201709002 · Cited by 2
A NOVEL DEEP LEARNING OPTIMIZATION ALGORITHM FOR HUMAN MOTIONS ANOMALY DETECTION
Omae, Y., Mori, M., Akiduki, T., & Hirotaka, T. (2019) · International journal of innovative computing, information & control · Cited by 2
One-D Convolution Neural Network Models for Human Activity Recognition using mHealth Datasets
Yogesh, K. M., Arpitha, S., & Gad, I. (2023) · International Conference on Frontiers of Artificial Intelligence and Machine Learning (FAIML) · doi:10.1145/3616901.3616933 · Cited by 2
“For instance, Ordonez and Roggen developed a Deep ConvLSTM architecture that combines both convolutional and recurrent layers [15].”
BackgroundTzamalis, P., Nikoletseas, S., & Spirakis, P. G. (2023) · International Conference on Distributed Computing in Smart Systems and the Internet of Things (DCOSS-IoT) · doi:10.1109/dcoss-iot58021.2023.00071 · Cited by 2
groupShapley: Efficient prediction explanation with Shapley values for feature groupsPreprint
Jullum, M., Redelmeier, A., & Aas, K. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.12228 · Cited by 2
“Another possible application of groupShapley is to time series classification settings (Kvamme et al., 2018; Ordóñez and Roggen, 2016).”
BackgroundAccuracy Improvement of Complex Sensor-based Activity Recognition Using Hybrid CNN
Mekruksavanich, S., Jantawong, P., & Jitpattanakul, A. (2022) · doi:10.1109/incit56086.2022.10067453 · Cited by 2
“It has tremendous potential in many application domains, including healthcare, care for the elderly, intelligent homes, amusement, and sporting events [1].”
BackgroundLuo, B., Qiu, Q., Zhu, T., & Liu, Z. (2024) · IFIP advances in information and communication technology · doi:10.1007/978-3-031-57919-6_7 · Cited by 2
Walking-posture Classification from Single-acceleration-sensor Data using Deep Learning
Hayama, T., & Arakawa, R. (2020) · IIAI International Congress on Advanced Applied Informatics (IIAI-AAI) · doi:10.1109/iiai-aai50415.2020.00087 · Cited by 2
“Ordonez et al. [12] developed a machine learning-based method to classify daily-life activities, such as “walk” and “sit-down,” using an accelerator attached to a human waist.”
MethodResultGViT: Combining Convolutional and Transformer Layers for Spatial-Temporal EEG Analysis
Zhu, C., Xu, Y., & Qu, X. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-12660-3_32 · Cited by 2
“Building upon this foundation, temporal convolutional approaches such as InceptionTime [7] and architectures optimized for sequential modeling like Deep-ConvLSTM [15] began to show improvements in capturing long-range temporal dependencies.”
Zhang, C., Hu, Y., Bu, R., Xiong, Z., Liu, M., Li, B., Zhao, L., Song, Y., & Li, C. (2024) · Urban Climate · doi:10.1016/j.uclim.2024.102261 · Cited by 3
Distance graph routing capsule network with vote elimination for human activity recognition
Feraydounraad, E., Hoseini, K., & Nazerfard, E. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.128236 · Cited by 3
Machine learning interpretation of inter-well radiowave survey data
Aleshin, I. M., & Malygin, I. V. (2019) · Computer Research and Modeling · doi:10.20537/2076-7633-2019-11-4-675-684 · Cited by 2
T-WaveNet: Tree-Structured Wavelet Neural Network for Sensor-Based Time Series AnalysisPreprint
Liu, M., Zeng, A., Lai, Q., & Xu, Q. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2012.05456 · Cited by 6
“[Guan and Pl¨otz, 2017; Yao et al. , 2017; Chen et al. , 2016; Ord´o˜nez and Roggen, 2016] combine CNNs with Recurrent neural networks (RNNs) to extract local and long-term temporal dependency features.”
BackgroundRecognition of Smartphone User Activity: From A Cyclical Perspective.Preprint
Mi, C., Xu, R., Lin, C., & Meng, R. Y. (2019) · arXiv (Cornell University) · Cited by 2
Inertial Sensor Based Human Activity Identification System Using CNN- LSTM Deep Learning Technique
Supriya, S., Shukla, A., & Manchanda, M. (2023) · IEEE Uttar Pradesh Section International Conference on Electrical, Electronics and Computer Engineering (UPCON) · doi:10.1109/upcon59197.2023.10434925 · Cited by 2
Islam, A. T., Bader, S., & Kirste, T. (2022) · International Workshop on Sensor-based Activity Recognition and Interaction (iWOAR) · doi:10.1145/3558884.3558897 · Cited by 3
“Studies in this area propose ANN-based motion classifiers using various computational units, including convolutional units [3], long short-term memory (LSTM) recurrent units [8], and variations of these computational units, e.g. in [11, 17].”
BackgroundPromptTSS: A Prompting-Based Approach for Interactive Multi-Granularity Time Series Segmentation
Chang, C., Lo, M.-C., Peng, W.-C., & Chen, T.-F. (2025) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3746252.3761142 · Cited by 7
“Most segmentation models are static, meaning they are trained on specific datasets and lack the ability to adapt to evolving patterns [7, 28].”
An Open Dataset of Pediatric Activity and Energy Expediture and Deep Learning Approach for Analysis
Qu, X., Wu, J., Zhang, T., et al. (2019) · IEEE International Conference on Healthcare Informatics (ICHI) · doi:10.1109/ichi.2019.8904472 · Cited by 2
Raja, M. C., & Mahri, M. M. B. A. (2025) · International Conference on Innovative Data Communication Technologies and Application (ICIDCA) · doi:10.1109/icidca66325.2025.11280566 · Cited by 3
Introducing Local Distance-Based Features to Temporal Convolutional Neural Networks
Iwana, B. K., Mori, M., Kimura, A., & Uchida, S. (2018) · International Conference on Frontiers in Handwriting Recognition · doi:10.1109/icfhr-2018.2018.00025 · Cited by 2
“Neural networks with fusion schemes have been used for combining title and image data [29], and for data from multiple sensors [30], such as multi-source remote sensing [31], activities of daily life [32], and pedestrian detection [33].”
MethodResearch on Wearable Monitoring System for Freezing of Gait in Parkinson’s Disease
Li, B., Sun, Y., Yang, X., Yao, Z., Zhou, X., Ma, Z., Bai, T., & Wang, P. (2021) · IEEE Information Technology, Networking, Electronic and Automation Control Conference (ITNEC) · doi:10.1109/itnec52019.2021.9587310 · Cited by 2
“used the combination of sensors (including accelerometers, gyroscopes and magnetometers) to obtain the motion signals of the subjects, and used the deep learning technology to identify the motion patterns (such as opening the door, opening the drawer, and cleaning the table) [6].”
MethodBarshan, B., & Koşar, E. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3577925 · Cited by 5
“These are three single (1D CNN, 2D CNN, and LSTM) and three hybrid models (standard 1D CNN-LSTM, Ordóñez & Roggen’s 1D CNN-LSTM [27], and the alternative 1D CNN-LSTM [28]).”
Mekruksavanich, S., Phaphan, W., & Jitpattanakul, A. (2025) · Lobachevskii Journal of Mathematics · doi:10.1134/s1995080224608166 · Cited by 5
“Ordonez et al. [6] proposed a robust CNN-LSTM network for identifying activities utilizing multiple wearable sensors.”
NEAT Activity Detection using Smartwatch at Low Sampling Frequency
Dewan, A., Gunturi, V. M. V., Naik, V. S., & Dutta, K. K. (2021) · IEEE SmartWorld, Ubiquitous Intelligence & Computing, Advanced & Trusted Computing, Scalable Computing & Communications, Cloud & Big Data Computing, Internet of People and Smart City Innovation (SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI) · doi:10.1109/swc50871.2021.00014 · 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
Towards a Clustering Guided Hierarchical Framework for Sensor-Based Activity Recognition
Wang, A., Zhao, S., Keh, H., Chen, G., & Roy, D. S. (2021) · Sensors · doi:10.3390/s21216962 · Cited by 4
PerceptionNet: A Deep Convolutional Neural Network for Late Sensor FusionPreprint
Kasnesis, P., Patrikakis, C. Z., & Venieris, I. S. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1811.00170 · Cited by 2
“Finally, in order to test the general applicability of our approach we used the same CNN architecture on the PAMAP2 dataset and compared it with the conventional CNN-EF [14] and the Convolutional LSTM [15] methods.”
BackgroundMethodInfluentialSenseMLP: a parallel MLP architecture for sensor-based human activity recognition
Li, W., Guo, J., & Wu, H. (2024) · Multimedia Systems · doi:10.1007/s00530-024-01384-y · Cited by 2
“Various architectures, including Convolutional Neural Networks (CNNs) [9–16], hybrid CNN-RNN models [17–21], attention-based models [22–28], and self-attention/transformer-based models [29–35], have been proposed and shown to enhance HAR accuracy.”
Violence Detection Through Surveillance System
Deshmukh, A. J., Warang, K., Pente, Y., & Marathe, N. R. (2020) · Advances in intelligent systems and computing · doi:10.1007/978-981-15-8289-9_49 · Cited by 2
Gattulli, V., Impedovo, D., Piccinno, A., & Pirlo, G. (2025) · SN Computer Science · doi:10.1007/s42979-025-04187-8 · Cited by 2
An Algorithm for Mining the Living Habits of Elderly People Living Alone Based on AIoT
Wu, J., Lu, Y., & Jiang, Y. (2025) · Sensors · doi:10.3390/s25072299 · Cited by 4
“Francisco Javier Ordóñez proposed a DNN framework called DeepConvLSTM [31], a DNN framework that combines convolutional and recursive layers, and instead of using a pooling layer, models the temporal dynamics of feature maps by using a two-layer LSTM.”
IMU Sensing–Based Hopfield Neuromorphic Computing for Human Activity Recognition
Yu, Z., Zahid, A., Ansari, S., Abbas, H. T., Heidari, H., Imran, M. A., & Abbasi, Q. H. (2022) · Frontiers in Communications and Networks · doi:10.3389/frcmn.2021.820248 · Cited by 4
“Francisco et al. (Ordóñez and Roggen, 2016) uses the LSTM algorithm to improve the CNN structure and introduces the concept of time series of activity sensor data.”
MethodYang, L., Hao, Y., & Shinoda, K. (2023) · European Signal Processing Conference (EUSIPCO) · doi:10.23919/eusipco58844.2023.10289883 · Cited by 2
Sensor-based human activity recognition via zero-shot learning
Wang, W. (2019) · DR-NTU (Nanyang Technological University) · doi:10.32657/10356/104645 · Cited by 5
“Especially in recent years, with the fast development of the deep learning techniques, many methods [70, 113, 171] based on deep learning are proposed.”
BackgroundSensor Data Augmentation by Resampling for Contrastive Learning in Human Activity RecognitionPreprint
Wang, J., Zhu, T., Gan, J., Chen, L., Ning, H., & Wan, Y. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2109.02054 · Cited by 2
Image-Mediated Data Augmentation for Low-Resource Human Activity Recognition
Wang, Z., Qu, Y., Tao, J., & Song, Y. (2019) · International Conference on Compute and Data Analysis (ICCDA) · doi:10.1145/3314545.3314565 · Cited by 3
Modaldrop: Modality-Aware Regularization for Temporal-Spectral Fusion in Human Activity Recognition
Zeng, X., Chen, Y., Xu, B., & Zhang, T. (2023) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49357.2023.10095880 · Cited by 3
“The proposed method in this paper can be readily applied to any backbone, including CNN, Transformer [9], or DeepConvL-STM [10].”
MethodInfluentialLearning Sensor Interdependencies for IMU-to-Segment Assignment
Kaichi, T., Maruyama, T., Tada, M., & Saitô, H. (2021) · IEEE Access · doi:10.1109/access.2021.3105801 · Cited by 2
“Following the methods of successful studies that have applied DNNs to IMU data [18], [37], we augment the sensor signals by adding zero-mean Gaussian noise to the accelerations.”
MethodStraightforward Recognition of Daily Objects in Smart Environments from Wearable Vision Sensor
Medina-Quero, J., Cruciani, F., Seidenari, L., et al. (2019) · IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops) · doi:10.1109/percomw.2019.8730860 · Cited by 2
Domain models for data sources integration in HAR
Hamidi, M., & Osmani, A. (2020) · Neurocomputing · doi:10.1016/j.neucom.2020.06.138 · Cited by 3
A New Model for Assembly Task Recognition: A Case Study of Seru Production System
Torkul, O., Selvi, İ. H., Şişci, M., & Diren, D. D. (2024) · IEEE Access · doi:10.1109/access.2024.3484955 · Cited by 2
“On the other hand, in some studies [88], [95], [101], the label assigned to the window was determined to be the same as the last sample of the window.”
ResultSensor-Based Automatic Recognition of Construction Worker Activities Using Deep Learning Network
Tezcan, Ö., Akçay, C., Sari, M., & Cavus, M. (2025) · Sensors · doi:10.3390/s25133988 · Cited by 6
“…although recent DL-based HAR systems using public datasets have reported high accuracies—e.g., 98% using LSTM on Parkinson’s data [15], and 93–95% with CNN or hybrid models [60,61]—our results match or exceed these despite being based on more variable and noisy real-world construction data.”
InfluentialMulti-Channels LSTM Networks for Fence Activity Classification
Hu, K., ZHENG, C., He, W. P., Bao, X., & Wang, Y. (2018) · IEICE Transactions on Information and Systems · doi:10.1587/transinf.2018edl8004 · Cited by 2
Wu, X., & Zhao, B. (2025) · Lecture notes in computer science · doi:10.1007/978-981-96-9891-2_38 · Cited by 2
Behavior recognition research based on reinforcement learning for dynamic key feature selection
Zhang, T., Ma, C., Sun, H., Yan, L., Wang, B., & Fang, Y. (2022) · International Symposium on Advances in Informatics, Electronics and Education (ISAIEE) · doi:10.1109/isaiee57420.2022.00054 · Cited by 3
“…LDA [17] 69% QDA [17] 53% NCC [17] 51% 1NN [17] 87% 3NN [17] 85% UP [17] 64% NStar [17] 64% SStar [17] 86% BDN [18] 73% CNN [18] 85.1% DeepConvLSTM [19] 89.5% Baseline 89.16% DKFSN 89.74% As can be seen in Table II, the method proposed in this paper has significant advantages over the baseline…”
MethodInfluentialZhang, W., & Zhou, S. (2017) · Sensors · doi:10.3390/s17061214 · Cited by 2
“DeepMap+ attempts to not damage the user-friendliness of the measurement device, so we drop the advanced multi-modal sensing [10,12,25] to free the mobile device from overloaded sensors although multi-sensors can bring us useful environment contexts.”
BackgroundHuman Activity Recognition in a Free-Living Environment Using an Ear-Worn Motion Sensor
Boborzi, L., Decker, J., Rezaei, R., Schniepp, R., & Wuehr, M. (2024) · doi:10.20944/preprints202404.0250.v1 · Cited by 9
Fujii, A., Kajiwara, D., & Murao, K. (2020) · Smart innovation, systems and technologies · doi:10.1007/978-981-15-8269-1_8 · Cited by 3
A Neural Network Architecture for Children’s Audio–Visual Emotion Recognition
(2023) · doi:10.3390/math11224573 · Cited by 3
Mithsara, W. K. M. (2022) · doi:10.1109/csase51777.2022.9759717 · Cited by 2
Magni, S., Bremm, R. P., Verros, K., He, X., Lecossois, S., Jelke, F., Husch, A., Gonçalves, J. M., & Hertel, F. (2026) · Biomedical Signal Processing and Control · doi:10.1016/j.bspc.2026.110845 · Cited by 2
Abedin, A., Ehsanpour, M., Shi, Q., Rezatofighi, H., & Ranasinghe, D. C. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2007.07172 · Cited by 4
“Combining these concepts, [12] proposes DeepConvLSTM by pairing convolutional and recurrent networks in order to model the temporal correlations at a more abstract representation level.”
BackgroundMethodHoussein, E. H., Mahdy, M. A., Kayed, M., Ouyang, H., & Mohamed, W. M. (2025) · Archives of Computational Methods in Engineering · doi:10.1007/s11831-025-10462-x · Cited by 4
Detecting hate speech in roman Urdu using a convolutional-BiLSTM-based deep hybrid neural network
Zohaib, M., Farooque, G., Alsulami, M., et al. (2025) · PeerJ Computer Science · doi:10.7717/peerj-cs.3342 · Cited by 2
Action Recognition Using Pose Data in a Distributed Environment over the Edge and Cloud
Takasaki, C., Takefusa, A., Nakada, H., & Oguchi, M. (2021) · IEICE Transactions on Information and Systems · doi:10.1587/transinf.2020dap0009 · Cited by 3
“…[22] recognized and predicted human actions by generating human pose motion captures using a model with an encoder and decoder before and after LSTM. Ord˜o˜n ez et al. [23] conducted recognition using data from an accelerometer, a gyroscope and a magnetometer with a mode that combined CNN and LSTM.”
MethodDemonstrating User Authentication via Electrical Muscle Stimulation
Chen, Y., Yang, Z., Abbou, R., Lopes, P., Zhao, B. Y., & Zheng, H. (2021) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3411763.3451563 · Cited by 11
“In addition, we use a Convolutional Neural Network (CNN) for challenge classifcation task [21].”
MethodA Sensor-Independent Multimodal Fusion Scheme for Human Activity Recognition
Alexiadis, A., Nizamis, A., Giakoumis, D., Votis, K., & Tzovaras, D. K. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-09282-4_3 · Cited by 2
Haresamudram, H., Tang, C. I., Suh, S., Lukowicz, P., & Plötz, T. (2024) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3675094.3677562 · Cited by 8
“Furthermore, we will present DeepConvLSTM [17], which is the established supervised deep learning method for HAR, to contrast against heuristic-based features as well as learning without any label information.”
Cao, J., Guo, F., Lai, X., Zhou, Q., & Dai, J. (2020) · Communications in computer and information science · doi:10.1007/978-981-15-7670-6_43 · Cited by 4
Multimodal Fusion-Based Human Action Recognition Using Wi-Fi CSI and Smartwatch Sensors
Wang, J., Cai, X., Dai, X., Xia, M. Y., Shi, C., & Chen, W. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2026.3686886 · Cited by 10
“[40], TimeMixer [41], TimeXer [42], DeepConvLSTM [43], attend and discriminate [44], and Widar3.”
An Immune Inspired Algorithm for Fault Tolerant Enhanced Multimodal Machine Learning
Cross, M., & Radu, V. (2021) · IEEE International Conference on Bioinformatics and Biomedicine (BIBM) · doi:10.1109/bibm52615.2021.9669293 · Cited by 2
Kou, Z., & Fang, Y. (2019) · Instrumentation Mesure Métrologie · doi:10.18280/i2m.180611 · Cited by 2
“calculated, using ̂ and ̂ , and the test sample y can be allocated to the class with the smallest residual [20].”
MethodLearning Unit State Recognition Based on Multi-channel Data FusionPreprint
Tian, F., Yue, J., Wan, X., Chao, K., & Zheng, Q. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1806.07372 · Cited by 4
“In video channel, both spatial and temporal characteristics of video image sequence are extracted by a convolutional neural network(CNN)[15] conjunction with a long short-term memory(LSTM)[16] model.”
MethodHnoohom, N., Mekruksavanich, S., & Jitpattanakul, A. (2025) · Lecture notes in networks and systems · doi:10.1007/978-3-031-99958-1_20 · Cited by 2
A novel multi-scale attention network for human activity recognition using wearable sensor data
Zhou, W., Yuan, H., Cao, L., Chen, X., & Dong, Y. (2026) · Biomedical Signal Processing and Control · doi:10.1016/j.bspc.2026.110351 · Cited by 1
Factored Latent-Dynamic Conditional Random Fields for single and multi-label sequence modeling
Neogi, S., & Dauwels, J. (2021) · Pattern Recognition · doi:10.1016/j.patcog.2021.108236 · Cited by 34
Calatrava-Nicolás, F. M., Ray, L. S. S., Rey, V. F., Lukowicz, P., & Mozos, Ó. M. (2025) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3714394.3756194 · Cited by 2
Human Activity Recognition based on Smartphone Sensors- A Comparative Study
Dhammi, L., & Tewari, P. (2021) · International Conference on Reliability, Infocom Technologies and Optimization (Trends and Future Directions) (ICRITO) · doi:10.1109/icrito51393.2021.9596305 · Cited by 1
Pitto, L., Simon, F. R., Ertel, G. N., Gauchard, G. C., & Mornieux, G. (2025) · Sensors · doi:10.3390/s25010279 · Cited by 2
A neurorobotics approach to behaviour selection based on human activity recognition
Ranieri, C. M., Moioli, R. C., Vargas, P. A., & Romero, R. A. F. (2022) · Cognitive Neurodynamics · doi:10.1007/s11571-022-09886-z · Cited by 2
“Similar architectures have been proposed for inertial data, processing either raw data (Ordóñez and Roggen 2016; Garcia et al. 2019) or descriptors obtained through feature extraction methods (Steven Eyobu and Han 2018; Ashry et al.”
MethodLi, 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
“Following these e ff orts, Ord ´ o ˜ nez et al. [22] proposed DeepConvLSTM, an eight-layer architecture that combines convolutional layers for local feature extraction with long short-term memory (LSTM) units to capture long-range temporal dependencies.”
Brunyé, T. T., Petrimoulx, M. V., & Cantelon, J. A. (2026) · Sensors · doi:10.3390/s26072034 · Cited by 3
Mekruksavanich, S., & Jitpattanakul, A. (2025) · IEEE Access · doi:10.1109/access.2025.3605337 · Cited by 1
“Ordóñez and Roggen [17] proposed a deep convolutional long short-term memory (LSTM) architecture that combines the spatial abstraction capabilities of CNNs with the sequential modeling strengths of LSTM networks, outperforming conventional models on standard benchmark datasets.”
Recurrent Neural Networks Architectures for Accidental Fall Detection on Wearable Embedded Devices
Musci, M., & Piastra, M. (2021) · Springer eBooks · doi:10.1007/978-3-030-71676-9_4 · Cited by 1
Robust Interactive HMI for Occupancy Estimation in Smart Buildings (WIP)
Dissem, M., Amayri, M., & Bouguila, N. (2024) · IEEE Consumer Communications & Networking Conference (CCNC) · doi:10.1109/ccnc51664.2024.10454743 · Cited by 1
“While others are based on computer vision and vision-based data [2], [3].”
BackgroundRadar for assisted living in the context of Internet of Things for Health and beyondPreprint
Kernec, J. L., Fioranelli, F., Yang, S., et al. (2018) · EasyChair preprint · doi:10.29007/tznx · Cited by 1
Vennela, K., Balaji, B., Chinnaiah, M. C., Lam, S.-K., Janardhan, N., Raju, P. S., Krishna, D. H., Vani, G. D., & Basha, M. (2026) · Sensors · doi:10.3390/s26144644 · Cited by 1
Vukićević, A. M., & Petrović, M. (2023) · Lecture notes in networks and systems · doi:10.1007/978-3-031-29717-5_1 · Cited by 3
Multi-modal Deep Learning for Assessing Surgeon Technical Skill on a Surgical Knot-tying TaskPreprint
Kasa, K., Burns, D., Goldenberg, M. G., Selim, O., Whyne, C. M., & Hardisty, M. (2022) · doi:10.36227/techrxiv.20085425.v1 · Cited by 1
Dual-Alignment Based Generalized Zero-shot Learning for Human Activity Recognition
Yang, Z., Gu, Y., Weng, W., Guo, S., & Chen, Y. (2022) · IEEE Smart World Congress (SWC) · doi:10.1109/smartworld-uic-atc-scalcom-digitaltwin-pricomp-metaverse56740.2022.00071 · Cited by 1
“So we extract the temporal and spatial features in the original activity data through the backbone network composed of CNNs and LSTMs [31].”
MethodHyperbolic Transformers with LLMs for Multimodal Human Activity Recognition
Soleimani, F., Khodabandelou, G., Chibani, A., & Amirat, Y. (2025) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros60139.2025.11247069 · Cited by 3
Discrimination of Normal Side Gait Changes in Gait Asymmetry Using a Multi-Kernel Convolution Model
Chen, Z., & Wu, J. (2023) · International Conference on Intelligent Systems and Knowledge Engineering (ISKE) · doi:10.1109/iske60036.2023.10481181 · Cited by 1
Deep CHORES: Estimating Hallmark Measures of Physical Activity Using Deep LearningPreprint
Mardini, M., Wanigatunga, S. N. A. A., Saldana, S. J., Casanova, R., & Manini, T. M. (2020) · PubMed Central · doi:10.48550/arxiv.2007.13114 · Cited by 4
“Additionally, the performance of the models built on Skoda dataset 34,36 ranged between accuracy = [88.”
BackgroundMethodInfluentialDong, L., Xu, X., He, G., Xu, Y., Rasol, J., Tao, C., & Zhang, Z. (2025) · Alexandria Engineering Journal · doi:10.1016/j.aej.2025.04.019 · Cited by 1
Kalimuthu, S., Perumal, T., Yaakob, R., Marlisah, E., & Raghavan, S. (2024) · AIP conference proceedings · doi:10.1063/5.0179747 · Cited by 2
Niksa-Rynkiewicz, T., Benardos, P., Witkowska, A., & Vosniakos, G.-C. (2025) · Scientific Reports · doi:10.1038/s41598-025-06726-y · Cited by 1
“data, such as those from cameras and microphones, alongside inertial sensor data and radio frequencies, which can enhance the effectiveness of movement classification 5,6 .”
Recurrent Control Nets for Deep Reinforcement LearningPreprint
Liu, V., Adeniji, A., Lee, N., Zhao, J., & Srouji, M. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1901.01994 · Cited by 1
Activity Recognition using 1D convolution from Accelerometers Data
Wang, J., & Liu, X. H. (2020) · Journal of Physics Conference Series · doi:10.1088/1742-6596/1550/3/032161 · Cited by 1
A Neural Network Architecture for Children’s Audio-Visual Emotion Recognition
Matveev, A., Matveev, Y., Frolova, O., Nikolaev, A., & Lyakso, E. (2023) · doi:10.20944/preprints202310.0722.v1 · Cited by 3
Predicting Driving Conditions at Mountain Crossings Using Deep Learning
Skjermo, J., Arnesen, P., Södersten, C., & Dahl, E. (2020) · IEEE International Conference on Intelligent Transportation Systems (ITSC) · doi:10.1109/itsc45102.2020.9294191 · Cited by 1
Human Activity Recognition in Smart Homes
Soozaei, A. S. (2019) · Otago University Research Archive (University of Otago) · Cited by 1
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.