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 1,001–1,100 of 2,807
Deep Learning architecture for temperature forecasting in an IoT LoRa based system
Ouahab, I. B. A., Abdelhakim, B. A., Astito, A., Zafar, B. A., & Bouhorma, M. (2019) · International Conference on Networking, Information Systems & Security (NISS) · doi:10.1145/3320326.3320375 · Cited by 17
“Such as a human activity recognition application, in [6] the authors proposed a generic deep learning framework for activity recognition based on convolutional and LSTM recurrent units.”
BackgroundRecognition of multiple overlapping activities using compositional CNN-LSTM model
Okita, T., & Inoue, S. (2017) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3123024.3123095 · Cited by 15
“This paper focuses on the problem that these architectures [6, 3] as well as most of other architectures (the semi-CRF model [5] and the Bayesian model using the importance sampling [1]) recognize one activity at a time.”
BackgroundMethodWearable Electronic Glove and Multilayer Para-LSTM-CNN-Based Method for Sign Language Recognition
Wang, D., Wang, M., Zhang, Z., Liu, T., Meng, C., & Guo, S. (2024) · IEEE Internet of Things Journal · doi:10.1109/jiot.2024.3454215 · Cited by 23
Hayashi, T., Nishida, M., Kitaoka, N., Toda, T., & Takeda, K. (2017) · IEICE Transactions on Fundamentals of Electronics Communications and Computer Sciences · doi:10.1587/transfun.e101.a.199 · Cited by 20
GraphConvLSTM: Spatiotemporal Learning for Activity Recognition with Wearable Sensors
Han, J., He, Y., Liu, J., Zhang, Q., & Jing, X. (2019) · Globecom · doi:10.1109/globecom38437.2019.9013934 · Cited by 30
Hui, M., Luo, N., Iu, H. H., et al. (2020) · IEEE Access · doi:10.1109/access.2020.2982043 · Cited by 17
Online Learning of Wearable Sensing for Human Activity Recognition
Zhang, Y., Gao, B., Yang, D., Woo, W. L., & Wen, H. (2022) · IEEE Internet of Things Journal · doi:10.1109/jiot.2022.3188785 · Cited by 23
“In addition, the deep neural network (DNN) [14], convolutional neural network (CNN) [15]–[17], restricted Boltzmann machine (RBM) [18], [19], and recurrent neural network (RNN) [20] models are widely used [13].”
MethodEbner, M., Fetzer, T., Bullmann, M., Deinzer, F., & Grzegorzek, M. (2020) · Sensors · doi:10.3390/s20226559 · Cited by 20
“This classification accuracy is composed of the class accuracies 95.8% for the positive, and 89.1% for the outlier class.”
BackgroundInfluentialXiang, T., Jin, Y., Liu, Z., Clifton, L., Clifton, D. A., Zhang, Y., Zhang, Q., Ji, N., & Zhang, Y. (2025) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2025.3548771 · Cited by 20
“Our proposed McBP-Net, with its CNN-LSTM architecture, effectively captures both spatial and temporal dependencies [41].”
Mao, Y., Yan, L., Guo, H., Hong, Y., Huang, X., & Yuan, Y. (2023) · Applied Sciences · doi:10.3390/app131810529 · Cited by 18
Human Activity Recognition Using Place-Based Decision Fusion in Smart Homes
Cumin, J., Lefebvre, G., Ramparany, F., et al. (2017) · Lecture notes in computer science · doi:10.1007/978-3-319-57837-8_11 · Cited by 17
Moccia, S., Solbiati, S., Khornegah, M., Bossi, F., & Caiani, E. G. (2022) · Computer Methods and Programs in Biomedicine · doi:10.1016/j.cmpb.2022.106753 · Cited by 20
Domínguez-Vidal, J. E., Rodríguez, N., & Sanfeliu, A. (2024) · International Journal of Social Robotics · doi:10.1007/s12369-024-01103-7 · Cited by 22
“What we have done is to use more and better sensors and more complex and powerful inference systems in order to make better and better inferences of their intent [3–5].”
EEGNAS: Neural Architecture Search for Electroencephalography Data Analysis and Decoding
Rapaport, E., Shriki, O., & Puzis, R. (2019) · Communications in computer and information science · doi:10.1007/978-981-15-1398-5_1 · Cited by 21
Sensor-based Detection and Classification of Soccer Goalkeeper Training Exercises
Haladjian, J., Schlabbers, D., Taheri, S., Tharr, M., & Bruegge, B. (2020) · ACM Transactions on Internet of Things · doi:10.1145/3372342 · Cited by 18
“…Long-Short-Term-Memory (LSTM) - have been adapted to exploit the time-dependency of the sensor signals, attaining the highest recognition results so far on large public benchmark datasets [Li et al. 2018; Murahari and Plötz 2018; Neverova et al. 2016; Ordóñez and Roggen 2016; Zeng et al. 2018].”
BackgroundDebele, Y., Shi, H.-Y., Wondosen, A., et al. (2023) · Drones · doi:10.3390/drones7070437 · Cited by 17
Medina-Quero, J., Orr, C., Zang, S., Nugent, C., Salguero, A., & Espinilla, M. (2018) · International Conference on Ubiquitous Computing and Ambient Intelligence (UCAmI) · doi:10.3390/proceedings2191225 · Cited by 17
“Some of these systems are capable of identifying simple to complex human activities, as well as the context in which they occur, from sensor data; this is currently a core aspect of smart assistive 15 technologies [3].”
BackgroundWei, C., & Chang, H.-C. (2021) · Sensors · doi:10.3390/s21155234 · Cited by 18
Estimating Vehicle Movement Direction from Smartphone Accelerometers Using Deep Neural Networks
Hernández, S., Pozo, R. F., & Gómez, L. A. H. (2018) · Sensors · doi:10.3390/s18082624 · Cited by 18
StresSense: Real-Time detection of stress-displaying behaviors
Khan, N. S., Qadir, S., Anjum, G., & Uddin, N. (2024) · International Journal of Medical Informatics · doi:10.1016/j.ijmedinf.2024.105401 · Cited by 20
Choi, J. H., & Lee, J. (2019) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3341162.3344871 · Cited by 17
“For instance, Ordóñez and Roggen [15] employed an early fusion-based approach, which integrates the sensor data at the initial stage, to classify actions of the OPPORTUNITY dataset.”
MethodRønsch, G. Ø., Külahçı, M., & Dybdahl, M. (2021) · International Journal of Production Research · doi:10.1080/00207543.2021.1893853 · Cited by 20
“Earlier work on raw signal data processing has shown good results using LSTM (Long Short Term Memory) networks (Nagorny et al. 2017) and CNN (Convolutional Neural Networks) (Ordóñez and Roggen 2016).”
MethodSun, Y., Xu, X., Tian, X., Zhou, L., & Li, Y. (2024) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2024.108705 · Cited by 26
Towards Backdoor Attack on Deep Learning based Time Series Classification
Ding, D., Zhang, M., Huang, Y., Pan, X., Feng, F., Jiang, E., & Yang, M. (2022) · IEEE International Conference on Data Engineering (ICDE) · doi:10.1109/icde53745.2022.00100 · Cited by 32
Feature learning using convolutional denoising autoencoder for activity recognition
Noor, M. H. M. (2021) · Neural Computing and Applications · doi:10.1007/s00521-020-05638-4 · Cited by 24
LSTM Based Hybrid Method for Basin Water Level Prediction by Using Precipitation Data
Liu, S., Puwen, L., & Koyamada, K. (2021) · Journal of Advanced Simulation in Science and Engineering · doi:10.15748/jasse.8.40 · Cited by 17
“LSTMs were initially introduced by Hochreiter and Schmidhuber in 1997, and were refined and applied into various domains by other researchers after that[13].”
MethodUsing Domain Knowledge for Interpretable and Competitive Multi-Class Human Activity Recognition
Scheurer, S., Tedesco, S. A., Brown, K. N., & O’Flynn, B. (2020) · Sensors · doi:10.3390/s20041208 · Cited by 18
“7%) on the Opportunity data (4% better than the deep convolutional LSTM by Ordóñez and Roggen [30]), a CNN the best score (93.”
BackgroundInfluentialDANA: Dimension-Adaptive Neural Architecture for Multivariate Sensor Data
Malekzadeh, M., Clegg, R. G., Cavallaro, A., & Haddadi, H. (2021) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3478074 · Cited by 27
“Alternatively, to turn deep architectures based on CNNs and FNNs/RNNs adaptive to changes in input-data dimensions, a preferable solution is to include an adaptive layer between the CNN and the FNN/RNN.”
BackgroundMethodInfluentialSaha, S. S., Sandha, S. S., Pei, S., Jain, V., Wang, Z., Li, Y., Sarker, A., & Srivastava, M. B. (2022) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3534586 · Cited by 28
“The bulk of emerging innovations builds upon recent advances in machine learning (ML) for sensor-based activity detection and gesture recognition [2, 11, 52, 68] because classical signal processing techniques do not yield high enough accuracy for complex activity recognition.”
BackgroundRazzaq, M. A., Medina-Quero, J., Cleland, I., Nugent, C., Akhtar, U., Bilal, H. S. M., Rehman, U. U., & Lee, S. (2020) · Multimedia Systems · doi:10.1007/s00530-020-00664-7 · Cited by 17
Multi-focus image fusion using edge discriminative diffusion filter for satellite images
Fiza, S., & Safinaz, S. (2024) · Multimedia Tools and Applications · doi:10.1007/s11042-024-18174-3 · Cited by 16
Agac, S., & İncel, Ö. D. (2024) · Computers & Electrical Engineering · doi:10.1016/j.compeleceng.2024.109274 · Cited by 19
Zhang, Y., Yamamoto, M., Suzuki, G., & Shioya, H. (2022) · IEEE Access · doi:10.1109/access.2022.3141767 · Cited by 17
“Behavior forecasting and the recognition of activity, speech, and emotion have also been performed by fusing convolutional neural network (CNN) [15] and LSTM [16]–[19] models.”
MethodDeep feature learning and selection for activity recognition
Mohammad, Y. F. O., Matsumoto, K., & Hoashi, K. (2018) · ACM Symposium on Applied Computing · doi:10.1145/3167132.3167234 · Cited by 21
“the lowest number of feature (except for OPP − ARM ) but at the expense of having the lowest classification accuracy by a large margin.”
BackgroundHuman Activity Recognition Based on Multichannel Convolutional Neural Network With Data Augmentation
Shi, W., Fang, X., Yang, G., & Huang, J. (2022) · IEEE Access · doi:10.1109/access.2022.3192452 · Cited by 24
Optimization-Free Test-Time Adaptation for Cross-Person Activity Recognition
Wang, S., Wang, J., Xi, H., Zhang, B., Zhang, L., & Wei, H. (2023) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3631450 · Cited by 41
“To overcome the limitations, researchers have utilized deep learning techniques to improve HAR performance in an end-to-end manner [32, 46, 54].”
MethodYang, Z., Li, K., & Huang, Z. (2024) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2024.108110 · Cited by 19
Subramanian, R. R., & Vasudevan, V. (2021) · Studies in computational intelligence · doi:10.1007/978-3-030-68291-0_11 · Cited by 18
Wearable affect and stress recognition: A reviewPreprint
Schmidt, P., Reiss, A., Duerichen, R., & Laerhoven, K. V. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1811.08854 · Cited by 31
“From a methodical point of view NN offer interesting approaches to transfer[105] or semi-supervised[150] learning.”
BackgroundSyed, S., Morseth, B., Hopstock, L. A., & Horsch, A. (2021) · Scientific Reports · doi:10.1038/s41598-021-87757-z · Cited by 22
Efficient light harvesting for accurate neural classification of human activities
Nicosia, A., Pau, D. P., Giacalone, D., et al. (2018) · International Conference on Consumer Electronics · doi:10.1109/icce.2018.8326103 · Cited by 16
“The first network is based on a shallow convolutional neural network [7] with only 5 layers; the first two layers are convolutional and they have the task of learning discriminative features from the tri-axial accelerometer, while the last layer is a softmax classifier returning the recognized activity among the avail…”
BackgroundA sea clutter detection method based on LSTM error frequency domain conversion
Yan, Y., & Xing, H. (2021) · Alexandria Engineering Journal · doi:10.1016/j.aej.2021.04.084 · Cited by 18
FilterNet: A Many-to-Many Deep Learning Architecture for Time Series Classification
(2020) · doi:10.3390/s20092498 · Cited by 17
“The p-C/L architecture is in some ways similar to the DeepConvLSTM [28] architecture, although the striding and pooling features of p-C/L, and its many-to-many approach, lead to substantially different performance characteristics.”
BackgroundMethodResultInfluentialActivity Recognition Based on Thermopile Imaging Array Sensor
Liang, Q., Lei, Y., Zhai, X., Wan, Z., & Nie, H. (2018) · IEEE International Conference on Electro Information Technology · doi:10.1109/eit.2018.8500177 · Cited by 15
“(Ordonez & Roggen, 2016).”
Semi-Supervised Adversarial Learning Using LSTM for Human Activity Recognition
Yang, S., Baek, D.-G., & Thapa, K. (2022) · Sensors · doi:10.3390/s22134755 · Cited by 18
“HAR systems deployments to the real world in applications such as ambient assisted living (AAL), personal health [2], elderly care [3], defences [4], astronauts [5], and smart homes [6] are potentially increasing.”
BackgroundHashmi, A., Barukab, O., & Osman, A. H. (2024) · PLoS ONE · doi:10.1371/journal.pone.0302294 · Cited by 26
Exploring Trends and Clusters in Human Posture Recognition Research: An Analysis Using CiteSpace
Yan, L., & Du, Y. R. (2025) · Sensors · doi:10.3390/s25030632 · Cited by 25
Teng, Q., Tang, Y., & Hu, G. (2024) · IEEE Sensors Journal · doi:10.1109/jsen.2024.3364187 · Cited by 18
“Inspired by the success of CNNs and RNNs in time series domains, [32] developed a generic hybrid framework (i.e., DeepConvLSTM) that combines convolutional blocks with LSTM recurrent units for multimodal wearable activity recognition.”
MethodA CNN Based Transfer Learning Model for Automatic Activity Recognition from Accelerometer Sensors
Chikhaoui, B., Gouineau, F., & Sotir, M. (2018) · Lecture notes in computer science · doi:10.1007/978-3-319-96133-0_23 · Cited by 20
“CNNs are well suited for classification problems and very often outperforming existing approaches [11].”
BackgroundPrabono, A. G., Yahya, B. N., & Lee, S. (2020) · Computers & Industrial Engineering · doi:10.1016/j.cie.2020.106953 · Cited by 18
The Lifespan of Human Activity Recognition Systems for Smart Homes
Hiremath, S. K., & Plötz, T. (2023) · Sensors · doi:10.3390/s23187729 · Cited by 27
“Sensor-based HAR systems comprise on-body or wearable sensing [49], sensors placed on objects [25,50], and ambient or in-the-environment sensors [51].”
BackgroundDiSalvo, B. J., Bandaru, D., Wang, Q., Li, H., & Plötz, T. (2022) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3550328 · Cited by 23
“More sophisticated, recent analysis techniques such as Deep Learning based methods [38, 57] are not directly applicable in our envisioned scenario due to training data limitations and the highly idiosyncratic behaviors and EDA responses of individual students.”
Feature selection for physical activity recognition using genetic algorithms
Baldominos, A., Isasi, P., & Sáez, Y. (2017) · IEEE Congress on Evolutionary Computation (CEC) · doi:10.1109/cec.2017.7969569 · Cited by 18
Primitive activity recognition from short sequences of sensory data
Mohammad, Y. F. O., Matsumoto, K., & Hoashi, K. (2018) · Applied Intelligence · doi:10.1007/s10489-018-1166-6 · Cited by 16
“93 using a deep convolutional LSTM for the locomotion activities [26] and 0.”
BackgroundMethodInfluentialSensor-based Activity Recognition using Deep Learning: A Comparative Study
Trabelsi, I., Françoise, J., & Bellik, Y. (2022) · International Conference on Movement and Computing (MOCO) · doi:10.1145/3537972.3537996 · Cited by 23
“To improve accuracy of predictions, researchers in [10, 32, 38] merged functionalities of CNN and LSTM architectures to create a new one, the so-called CNN-LSTM.”
BackgroundMethodReal-time human behaviour monitoring using hybrid ambient assisted living framework
Patel, A. D., & Shah, J. H. (2020) · Journal of Reliable Intelligent Environments · doi:10.1007/s40860-020-00100-7 · Cited by 17
Yuki, Y., Nozaki, J., Hiroi, K., Kaji, K., & Kawaguchi, N. (2018) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3267305.3267533 · Cited by 17
Gait Activity Authentication Using LSTM Neural Networks with Smartphone Sensors
Zhang, M. (2019) · International Conference on Mobile Ad-Hoc and Sensor Networks (MSN) · doi:10.1109/msn48538.2019.00092 · Cited by 20
“The combination of CNNs and LSTMs in a unified framework has offered state-of-the-art results [8].”
BackgroundIf only we had more data!: Sensor-Based Human Activity Recognition in Challenging Scenarios
Plötz, T. (2023) · 2023 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events (PerCom Workshops) · doi:10.1109/percomworkshops56833.2023.10150267 · Cited by 16
“The success of end-to-end models such as DeepConvLSTM [15] then seemed to justify an even more drastic paradigm shift from optimizing individual components of the ARC, even if including learning representations as in [26], to fully automated approaches.”
BackgroundOn-Device Deep Learning Inference for Efficient Activity Data Collection
Mairittha, N., Mairittha, T., & Inoue, S. (2019) · Sensors · doi:10.3390/s19153434 · Cited by 23
“Researchers have been adopting deep learning methods for activity recognition [11,25,26].”
BackgroundMethodMekruksavanich, S., & Jitpattanakul, A. (2025) · Applied System Innovation · doi:10.3390/asi8030057 · Cited by 22
“Integrating models that incorporate CNNs and LSTMs utilizes both spatial and temporal data [14].”
Signal processing for miniature mass spectrometer based on LSTM-EEMD feature digging
Zhan, C., Ju, Z., Xie, B., Chen, J., Ma, Q., & Li, M. (2024) · Talanta · doi:10.1016/j.talanta.2024.126904 · Cited by 16
A hybrid and context-aware framework for normal and abnormal human behavior recognition
Mojarad, R., Chibani, A., Attal, F., Khodabandelou, G., & Amirat, Y. (2023) · Soft Computing · doi:10.1007/s00500-023-09188-4 · Cited by 19
Tahir, S. B. U. D., Jalal, A., & Kim, K. (2021) · International Bhurban Conference on Applied Sciences and Technology (IBCAST) · doi:10.1109/ibcast51254.2021.9393204 · Cited by 15
Xia, S., Chu, L., Pei, L., Yang, J., Yu, W., & Qiu, R. C. (2024) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2024.3381171 · Cited by 23
“[43]proposedtheNeuralNetworkViterbi(NNV),which generatesthepseudogroundtruthiteratively.”
BackgroundSiraj, M. S., & Ahad, M. A. R. (2020) · 2020 Joint 9th International Conference on Informatics, Electronics & Vision (ICIEV) and 2020 4th International Conference on Imaging, Vision & Pattern Recognition (icIVPR) · doi:10.1109/icievicivpr48672.2020.9306630 · Cited by 15
“Our approach is inspired by [23, 24] except for using a sequential layer of CNNs and GRUs similar to DeepSense as demonstrated by Yao et al.”
MethodEnd-to-end Learning for Measuring in-meal Eating Behavior from a Smartwatch
Kyritsis, K., Diou, C., & Delopoulos, A. N. (2018) · Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) · doi:10.1109/embc.2018.8513627 · Cited by 16
“In the recent years, end-to-end learning approaches that jointly train the convolutional and recurrent parts of Deep Neural Networks (DNNs) have been successfully applied in practical recognition problems such as multi-sensor wearable activity recognition [12] and audio-based emotion recognition [13].”
BackgroundInertial-Vision: Cross-Domain Knowledge Transfer for Wearable Sensors
Abebe, G., & Cavallaro, A. (2017) · IEEE International Conference on Computer Vision Workshops · doi:10.1109/iccvw.2017.165 · Cited by 16
Deep Learning for Detecting Human Activities From Piezoelectric-Based Kinetic Energy Signals
Manjarrés, J., Lan, G., Gorlatova, M. A., Hassan, M., & Pardo, M. (2021) · IEEE Internet of Things Journal · doi:10.1109/jiot.2021.3093245 · Cited by 16
Applying Deep Learning and Wearable Devices for Educational Data Analytics
Zhou, Z., Tam, V. W. L., Lui, K., Lam, E. Y. M., Yuen, A. H. K., Hu, X., & Law, N. W. Y. (2019) · IEEE International Conference on Tools with Artificial Intelligence (ICTAI) · doi:10.1109/ictai.2019.00124 · Cited by 22
“By combining the convolutional neural network(CNN) and the recurrent units of the LSTM model, the model proposed by Ordonez and Roggen outperformed the non-recurrent networks structures by 4% as shown in the obtained experimental results.”
BackgroundInfluentialFischer, A., Bedrikow, A. B., Tommelein, I. D., Nübel, K., & Fottner, J. (2023) · Algorithms · doi:10.3390/a16040212 · Cited by 19
An Information Gain-Based Model and an Attention-Based RNN for Wearable Human Activity Recognition
Liu, L., He, J., Ren, K., Lungu, J., Hou, Y., & Dong, R. (2021) · Entropy · doi:10.3390/e23121635 · Cited by 18
Outdoor activity classification using smartphone based inertial sensor measurements
Bodhe, R., Sivakumar, S., Sakarkar, G., Juwono, F. H., & Apriono, C. (2024) · Multimedia Tools and Applications · doi:10.1007/s11042-024-18599-w · Cited by 20
Shakerian, R., Yadollahzadeh‐Tabari, M., & Rad, S. Y. B. (2022) · IET Biometrics · doi:10.1049/bme2.12066 · Cited by 21
Smartphone-sensors Based Activity Recognition Using IndRNN
Li, S., Li, C., Li, W., Hou, Y., & Cook, C. (2018) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3267305.3267521 · Cited by 15
“Some methods have been reported in the literature generally falling into two categories: traditional methods [5, 10, 12] and deep learning based methods including convolutional neural networks (CNN) [14, 1, 3, 13] and Recurrent Neural Networks (RNN) [9, 8].”
MethodHuman activity recognition in smart environments employing margin setting algorithm
Igwe, O. M., Wang, Y., Giakos, G. C., & Fu, J. Y. (2020) · Journal of Ambient Intelligence and Humanized Computing · doi:10.1007/s12652-020-02229-y · Cited by 16
Action recognition for educational proposals applying concepts of Social Assistive Robotics
de Carvalho, K. B., Basílio, V. T., & Brandao, A. S. (2021) · Cognitive Systems Research · doi:10.1016/j.cogsys.2021.09.002 · Cited by 20
Mulás-Tejeda, E., Gómez-Espinosa, A., Cabello, J. A. E., Cantoral-Ceballos, J. A., & Molina-Leal, A. (2024) · Sensors · doi:10.3390/s24103004 · Cited by 16
“The use of memory cells, instead of recurrent units, helps ease the long-timescale learning of temporal relationships [31].”
BackgroundA Context-aware Hybrid Framework for Human Behavior Analysis
Mojarad, R., Attal, F., Chibani, A., & Amirat, Y. (2020) · IEEE International Conference on Tools with Artificial Intelligence (ICTAI) · doi:10.1109/ictai50040.2020.00078 · Cited by 15
“In [17], a deep framework using convolutional and LSTM recurrent units is proposed for HAR.”
MethodKolhar, S., & Jagtap, J. (2021) · Ecological Informatics · doi:10.1016/j.ecoinf.2021.101334 · Cited by 16
Self-Supervised WiFi-Based Activity Recognition
Bocus, M. J., Lau, H.-S., McConville, R., Piechocki, R. J., & Santos‐Rodríguez, R. (2022) · IEEE Globecom Workshops (GC Wkshps) · doi:10.1109/gcwkshps56602.2022.10008537 · Cited by 29
Wang, X., Wang, X., Lv, T., Jin, L., & He, M. (2021) · Sensors · doi:10.3390/s21206927 · Cited by 24
“A Hybrid Model and DeepConvLSTM The basic idea in [5,33] is the same, namely, to use a hybrid model combining the characteristics of a CNN and LSTM to address the HAR identification task.”
BackgroundMethodInfluentialVolleyball Setting Technique Assessment Using a Single Point Sensor
Holatka, A.-K., Suwa, H., & Yasumoto, K. (2019) · IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops) · doi:10.1109/percomw.2019.8730811 · Cited by 17
Zhao, B., Li, S., Gao, Y., Li, C., & Li, W. (2020) · Sensors · doi:10.3390/s20236984 · Cited by 16
“Francisco framework [17] using convolution and LSTM (long short-term memory) together where the convolution extracts the spatial feature and LSTM helps learn the long-term temporal information.”
MethodA hybrid CNN and BLSTM network for human complex activity recognition with multi-feature fusion
Huan, R., Zhan, Z., Ge, L., Chi, K., Chen, P., & Liang, R. (2021) · Multimedia Tools and Applications · doi:10.1007/s11042-021-11363-4 · Cited by 18
Motion-To-BMI: Using Motion Sensors to Predict the Body Mass Index of Smartphone Users
Yao, Y., Ling, S., & Jin, Y. (2020) · Sensors · doi:10.3390/s20041134 · Cited by 16
“Recently, the combination of CNNs and LSTM in a unified stack framework has already offered state-of-the-art results in speech recognition [24] and some motion sensor-based tasks [25].”
BackgroundRecommendation system for human physical activities using smartphones
Kadri, N., Ellouze, A., & Ksantini, M. (2020) · doi:10.1109/iccis49240.2020.9257671 · Cited by 16
“In the case of deep learning, we can talk about convolutional neural networks [4] and recurrent neural networks [5].”
BackgroundMethodDeep Learning in the Recognition of Activities of Daily Living Using Smartwatch Data
Cavalcante, A. F., Kunst, V. H. D. L., Chaves, T., de Souza, J. D. T., Ribeiro, I. M., Quintino, J. P., da Silva, F. Q. B., Santos, A. L. M., Teichrieb, V., & Gama, A. E. F. D. (2023) · Sensors · doi:10.3390/s23177493 · Cited by 22
Human Activity Recognition Machine With an Anchor-Based Loss Function
Jin, L., Wang, X., Chu, J., & He, M. (2021) · IEEE Sensors Journal · doi:10.1109/jsen.2021.3130761 · Cited by 16
IoT System Based on parameter optimization of Deep Learning using Genetic Algorithm
Slim, S. O., Elfattah, M. M. A., Atia, A., & Mostafa, M.-S. (2021) · International journal of intelligent engineering and systems · doi:10.22266/ijies2021.0430.20 · Cited by 16
Hua, R., & Wang, Y. (2020) · IEEE Sensors Letters · doi:10.1109/lsens.2020.2984152 · Cited by 15
iEat: automatic wearable dietary monitoring with bio-impedance sensing
Liu, M., Zhou, B., Rey, V. F., Bian, S., & Lukowicz, P. (2024) · Scientific Reports · doi:10.1038/s41598-024-67765-5 · Cited by 22
Li, X., Xu, H., Wang, Y., Zeng, J., Li, Y., Li, X., Ai, W., Zheng, H., & Duan, Y. (2024) · IEEE Internet of Things Journal · doi:10.1109/jiot.2024.3491362 · Cited by 19
“Ordóñez and Roggen [20] proposed a combination of CNN and LSTM to extract spatial-temporal features.”
Wang, Y., Chen, Q., Ding, M., & Li, J. (2019) · Sensors · doi:10.3390/s19235302 · Cited by 15
Tian, Y., Zhang, J., Li, L., & Liu, Z. (2021) · IEEE Access · doi:10.1109/access.2021.3100580 · Cited by 17
Martínez-Hernández, U., Awad, M. I., & Dehghani-Sanij, A. A. (2021) · Neurocomputing · doi:10.1016/j.neucom.2021.10.044 · Cited by 21
“Inertial measurement units, over the full body, were employed for recognition of ADLs using CNNs alone, and combined with Recurrent Neural Networks [5, 43].”
MethodCombining Symbolic Reasoning and Deep Learning for Human Activity Recognition
Rueda, F. M., Lüdtke, S., Schröder, M., Yordanova, K., Kirste, T., & Fink, G. A. (2019) · IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops) · doi:10.1109/percomw.2019.8730792 · Cited by 20
“[6], [16], [19], [21].”
BackgroundMethodInfluentialSenyurek, V. Y., Imtiaz, M. H., Belsare, P. V., Tiffany, S. T., & Sazonov, E. E. (2019) · Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) · doi:10.1109/embc.2019.8856395 · Cited by 15
Multi-Instance Learning Algorithm Based on LSTM for Chinese Painting Image Classification
Li, D., & Zhang, Y. (2020) · IEEE Access · doi:10.1109/access.2020.3027982 · Cited by 14
Modality-wise relational reasoning for one-shot sensor-based activity recognition
Kasnesis, P., Chatzigeorgiou, C., Patrikakis, C. Z., & Rangoussi, M. (2021) · Pattern Recognition Letters · doi:10.1016/j.patrec.2021.03.003 · Cited by 14
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.