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 2,701–2,800 of 2,807
Spatiotemporal Graph Neural Network for Lower-Limb Activity Recognition Using IMUs
Katherine, A. M. L., Wu, Q., Zuo, G., & Yu, S. (2026) · IEEE/ASME International Conference on Mechatronic and Embedded Systems and Applications · doi:10.1109/mesa70578.2026.11684188
“However, many deep IMU-based HAR models process multichannel sensor streams as flat temporal sequences, which may under-exploit structural relationships among lower-limb segments [12], [17].”
Integrating Clinical Parameters and Cardiotocograms for Real-Time AlertingPreprint
Schulz, C., Kaaz, V., Brodowski, L., Dutemeyer, V., & Deserno, T. M. (2026) · Research Square · doi:10.21203/rs.3.rs-9651298/v1
Goyal, K., Goyal, S., & Pratik, S. (2025) · 2025 IEEE 7th Symposium on Computers & Informatics (ISCI) · doi:10.1109/isci65687.2025.11167688
Performance Analysis of CNN-LSTM Model for Human Activity Recognition
Hemalatha, N., & Vikram, G. N. V. R. (2025) · IEEE International Symposium on Sustainable Energy, Signal Processing and Cyber Security (iSSSC) · doi:10.1109/isssc66652.2025.11387929
Aziz, K., Dua, S., & Gupta, P. (2024) · International Conference on Advancements in Power, Communication and Intelligent Systems (APCI) · doi:10.1109/apci61480.2024.10616954 · Cited by 2
“Recent advancements in deep learning have opened new avenues for research in activity recognition [5] demonstrated the effectiveness of deep convolutional and LSTM recurrent neural networks in improving the accuracy and reliability of activity recognition systems, showcasing the potential of deep…”
BackgroundAdil, F., Kazambayev, I., Mikryakov, A., Aubakirova, A., & Medetkhan, A. (2026) · IEEE International Conference on Smart Information Systems and Technologies (SIST) · doi:10.1109/sist61674.2026.11595990
“Machine learning approaches, including neural networks and deep learning methods, have demonstrated promising results in predicting physiological state and physical readiness in controlled experimental conditions [6], [12].”
Wan, J., Teng, T. C. I., & Zhang, X. (2023) · doi:10.22541/au.169985538.89262193/v1
Rotation Estimation Auxiliary Task for Wrist Inertial-Based Human Activity Recognition
Calatrava-Nicolás, F. M., Rey, V. F., Lukowicz, P., Stoyanov, T., & Mozos, Ó. M. (2026) · IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops) · doi:10.1109/percomworkshops68308.2026.11585279
Sensor Fusion-Based HAR for Disease Monitoring and Prediction
Biswas, S. (2026) · Signals and communication technology · doi:10.1007/978-3-032-09241-0_6
Gupta, A. K., Garg, R. R., Verma, A., & Sain, M. (2025) · International Journal of Humanoid Robotics · doi:10.1142/s0219843625400237
Deep Neural Networks for Human Behavior Understanding
Singh, R. K., & Nigam, S. (2019) · Springer eBooks · doi:10.1007/978-3-030-15887-3_32 · Cited by 1
Research On Driving Behavior Recognition By Smart Car SensorsPreprint
Hao, A., Qi, H., & Liang, H. (2023) · Research Square · doi:10.21203/rs.3.rs-3029615/v1
Abdulrahman, B., Zhu, Z., & Conway, A. J. (2026) · Electronics · doi:10.3390/electronics15143139
Zaidan, S. (2025) · Journal of education for pure science. · doi:10.32792/jeps.v15i2.477
“machine learning methods [2–4] .”
Beyond Action Segmentation: An IMU-based Fitness Tracking System with Cycle-Level Granularity
Li, T., Jin, B., Li, D., Chen, Y., Feng, X. T., Ma, Y., Liu, D., Wu, J., & Guo, C. (2026) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp55912.2026.11462265
OpenPack Challenge 2022 Report: Impact of Data Cleaning and Time Alignment on Activity Recognition
Matsubayashi, Y. (2023) · IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events (PerCom Workshops) · doi:10.1109/percomworkshops56833.2023.10150299
“We employed DeepConvLSTM [2] for the classification model because it is usually used in a baseline method in many human activity recognition studies.”
MethodActivity recognition via Riemannian optimization and multimodal Transformers
Soleimani, F., Khodabandelou, G., Chibani, A., & Amirat, Y. (2026) · Machine Learning with Applications · doi:10.1016/j.mlwa.2026.101021
Predicting the Movement Intention and Controlling the Grip of a Myoelectrical Active Prosthetic Arm
Dellagostin, J., Cukla, A. R., Bisogno, F. E., et al. (2022) · Lecture notes in networks and systems · doi:10.1007/978-3-030-96308-8_102
Mekruksavanich, S., Phaphan, W., & Jitpattanakul, A. (2026) · International Conference on Big Data Analytics and Practices (IBDAP) · doi:10.1109/ibdap69915.2026.11677760
“Convolutional and recurrent networks—hybrid convolutional long short-term memory designs in particular—showed that both the spatial and temporal structure within IMU data can be extracted automatically and with precision [1]–[3].”
Design and Development of Smart Wearable Technology Enhanced Learning for Specially Abled Students
Shreyanth, S., Harshitha, D. S., Agarwal, P., et al. (2024) · Springer Proceedings in Humanities and Social Sciences · doi:10.1007/978-981-97-3701-7_10
How Much Context Matters? A Comparison for Skeleton-Based Activity Recognition
Tschöpe, M., Friedrich, B., Bian, S., & Lukowicz, P. (2026) · International Conference on Activity and Behavior Computing (ABC) · doi:10.1109/abc68169.2026.11567225
“In 2016, Ord´o˜nez et al. [14] introduced the ConvLSTM architecture, one of the first models to combine convolutional and recurrent layers for sequential sensor data.”
Deep Learning for Virtual Reality User Identification: A Benchmark
Frizzo, D., Genilotti, F., Petrovic, D., et al. (2026) · International Conference on Control, Decision and Information Technologies (CoDIT) · doi:10.1109/codit70676.2026.11631300
Yang, C., Li, Z., Xi, Y., Liu, H., & Zhang, J. (2026) · International Conference on Advanced Electronic Materials, Computers and Software Engineering (AEMCSE) · doi:10.1109/aetcse69203.2026.11504161
“For example, Ordóñez F. J. and Roggen D. proposed the DeepConvLSTM framework, which combines convolutional and LSTM layers for human activity recognition using multimodal wearable sensors, achieving superior ability to differentiate similar activities[5].”
CNN-FIFO-LSTM: An Energy-Efficient Hybrid Architecture for Real-Time Fall Detection on Edge Devices
Cong, D. L., Van, C. C., Pham, H. A., Truong, C. D., Ngoc, M. N., & Tran, L. T. T. (2026) · IEEE Access · doi:10.1109/access.2026.3718746
“Ordóñez and Roggen introduced a convolutional-LSTM framework for multimodal wearable activity recognition, and DeepSense similarly combined convolutional and recurrent modules to learn local sensor interactions and temporal relationships in mobile-sensing time series [23], [24].”
Beyond Thresholds: A General Approach to Sensor Selection for Practical Deep Learning-based HAR
Cooper, G., & Marculescu, R. (2024) · IEEE/ACM International Conference on Internet-of-Things Design and Implementation (IoTDI) · doi:10.1109/iotdi61053.2024.00005
“7, 8, 9 for examples.”
BackgroundNishitsuji, R., Kawasaki, D., Nakashima, T., Sueyoshi, K., Hisamoto, H., & Endo, T. (2026) · Soft Computing · doi:10.1007/s00500-026-11387-8
“Machine-learning techniques are employed in various fields (Ordóñez and Roggen 2016; Kratzwald et al. 2018; Zhao et al. 2019).”
Zhao, Z., He, C., Jiang, X., & Xu, X. (2025) · Computers & Electrical Engineering · doi:10.1016/j.compeleceng.2025.110583 · Cited by 3
PhysioFit: A Inline Activity Recognition Leveraging Physiological Sensors
Subramanian, R. R., Gatikoppu, S. P. P. K., Chitikina, V. K., Jana, P., & Gollapalli, S. V. N. M. T. (2024) · International Conference on Automation, Computing and Renewable Systems (ICACRS) · doi:10.1109/icacrs62842.2024.10841482
Can LLMs Learn the Language of Inertial Sensors in Human Activity Recognition?
Bigelli, L., Bogliolo, A., Calisti, L., Capellacci, P., Contoli, C., Kania, N., & Lattanzi, E. (2025) · International Conference on Machine Learning Technologies (ICMLT) · doi:10.1109/icmlt65785.2025.11193221
“Over the years, machine learning and deep learning approaches have significantly improved HAR, achieving remarkable classification accuracy [4], [5].”
A video-based temporal movement quality assessment framework for paralympic wheelchair sports
Anil, G., & Narayanan, A. E. (2026) · Health Technologies · doi:10.58962/ht.306
“By integrating temporal learning models, it is possible to analyze movement patterns over a long period of time [9].”
InfluentialA Systematic Review of Abnormal Behavior Detection and Analysis in Driving SimulatorsPreprint
Tang, Y. M., Zhao, D., Chen, T., et al. (2024) · SSRN Electronic Journal · doi:10.2139/ssrn.4772738
Data Management and Mining Technologies to Manage and Analyze Data in IoT
Sharma, S., Sasubilli, S. M., & Bhargava, K. (2020) · Wiley eBooks · doi:10.1002/9781119778868.ch10
Sensors as Neural Computing Units
Alsaleem, F., Hasan, M. H., Nikfarjam, H., & Pourkamali, S. (2022) · Elsevier eBooks · doi:10.1016/b978-0-12-822548-6.00159-x
DeepMultiple: A Deep Learning Model for RFID-based Multi-object Activity Recognition
Shen, S., Yang, L., Chen, S., Dong, W., Yu, B. W., & Wang, Q. (2023) · Proceedings/Proceedings of the ... International Conference on Software Engineering and Knowledge Engineering · doi:10.18293/seke2023-138
Zhu, X., Gu, X., Wan, J., Cai, J., & Gu, C. (2026) · The Journal of Supercomputing · doi:10.1007/s11227-026-08669-1
Dual-Space Masked Reconstruction for Robust Self-Supervised Human Activity Recognition
Xiao, S., Deng, J., Tang, C., & Huang, Z. (2025) · Proceedings of the 34th ACM International Conference on Information and Knowledge Management · doi:10.1145/3746252.3761215
Gjoreski, H., Sazdov, B., Stankoski, S., et al. (2024) · Ambient intelligence and smart environments · doi:10.3233/aise240031
Construction of motion phase classification algorithm using ankle joint physical signals
Matsumoto, T., & Kano, Y. (2024) · Intelligent Data Analysis · doi:10.1177/1088467x241296750
A novel ranging method based on magnetic fields and neural networks
Mushtaq, S., Brunacci, V., Felicioni, S., et al. (2026) · Measurement · doi:10.1016/j.measurement.2026.122061
An Intelligent Collaboration Framework of Iot Applications Based on Event Logic GraphPreprint
Zhu, M., Han, Y., Liu, Z., et al. (2022) · SSRN Electronic Journal · doi:10.2139/ssrn.4105010
Application of Convolutional Neural Network in Feature Signal Recognition
Yu, H. (2023) · International Conference on Electronics and Communication, Network and Computer Technology (ECNCT) · doi:10.1109/ecnct59757.2023.10281136
“Various interpretable AI techniques have been proposed to address this issue, such as visualization methods [18] and attention mechanisms [19].”
MethodMadhukar, G., Jatoth, C., & Doriya, R. (2025) · Proceedings of the Institution of Mechanical Engineers Part D Journal of Automobile Engineering · doi:10.1177/09544070251337211 · Cited by 2
“29–32,81 High-dimensional sparse features are serialized by LSTM. Driving behaviors as well as vehicle running were utilized to reflect a driver’s personality, and deep learning technique was employed to extract critical characteristics without feature selection.”
InfluentialResearch on Travel Mode Identification Based on Trajectory Data
Zhang, R., Li, D., & Huang, Y. M. (2024) · Lecture notes in civil engineering · doi:10.1007/978-981-97-9644-1_31
Predição de Congelamento de Marcha por Meio da Detecção de Eventos de Pré-Congelamento
Amoedo, J. P. A., Gonçalves, P. H. N., Albuquerque, R., Santos, E. M. D., Giusti, R., & Renato, C. (2023) · doi:10.5753/sbcas.2023.229854
“O modelo de rede utilizado neste trabalho é o DeepConvLSTM, proposto em [Ordóñez and Roggen 2016] e que se encontra disponibilizado publicamente1.”
CardioHARNet: A Lightweight Hybrid Deep Model Using Raw IMU Signals for Human Activity Recognition
Akter, M., & Son, J. (2026) · International Conference on Artificial Intelligence in Information and Communication (ICAIIC) · doi:10.1109/icaiic68212.2026.11454240
Kim, H., & Akther, S. (2026) · Communications in computer and information science · doi:10.1007/978-3-032-22190-2_12
Watanabe, K., Kano, K., Hossain, T., Katayama, S., Urano, K., Yonezawa, T., & Kawaguchi, N. (2025) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/3714394.3756198 · Cited by 2
“• DeepConvLSTM [15]: A deep learning model that classifies tasks based on features extracted from each 2.56s segment using CNN followed by LSTM. • CNN-Transformer : Our proposed task recognition model without pretraining, which classifies tasks based on each 51.2s window (composed of 20 consecutive…”
Nedeltcheva, G. N., Stefanov, T., Chikurtev, D., & Kovatcheva, E. P. (2026) · International Conference on Engineering, Technologies and Systems (TechSys) · doi:10.3390/engproc2026150005
“Sustaining analytical rigor requires implementation of robust data validation protocols, anomaly detection mechanisms, and intelligent imputation strategies [40].”
Human Activity Recognition System Using Artificial Neural Networks
Almeida, V. F. D., & Andreão, R. V. (2022) · World Congress on Medical Physics and Biomedical Engineering, September 7 - 12, 2009, Munich, Germany · doi:10.1007/978-3-030-70601-2_192
RECNN: A Novel Deep Learning Model to Predict 8-state Protein Secondary Structure
Cheng, L., Xia, Y., Lu, Y., Shen, J., Qiu, Y., Zhao, H., & Lu, W. (2025) · Communications in computer and information science · doi:10.1007/978-981-96-1907-8_17 · Cited by 1
Chen, X., Tao, L., & Li, Z. (2025) · Knowledge-Based Systems · doi:10.1016/j.knosys.2025.114833
Sadeghsalehi, H. (2026) · Array · doi:10.1016/j.array.2026.100778 · Cited by 4
Mekruksavanich, S., & Jitpattanakul, A. (2026) · Computers · doi:10.3390/computers15090595
“Four of them are representative deep learning architectures for time-series HAR: TCN [7], DeepConvLSTM [6], InceptionTime [8], and a Transformer-based encoder [9].”
Velantina, V., & Manikandan, V. (2024) · Algorithms for intelligent systems · doi:10.1007/978-981-97-0180-3_54
Stoeve, M., Dorschky, E., Richer, R., Rohleder, N., & Eskofier, B. M. (2026) · Scientific Reports · doi:10.1038/s41598-026-58429-7
“Input sequences were segmented (with the number of segments optimized via Optuna), and dropout was applied before the final softmax layer 45,46 .”
Hong, S., Kim, H., & Lee, B. (2026) · Automation in Construction · doi:10.1016/j.autcon.2026.107157
Wear2Toy: Leveraging Wearable Devices for Hand-Held Objects Activity Recognition
Porta, N. L., Maio, C. D., Faraci, F. D., Glaser, B., Ramelli, G. P., & Papandrea, M. (2026) · IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops) · doi:10.1109/percomworkshops68308.2026.11585463
“1) Pre-training: We employed DeepConvLSTM architecture proposed by [33] to test our approach. have opted for this hybrid architecture, originally designed for HAR using wearable sensors, because it has been proven to combine spatial and features: the convolutional layers extract local spatial…”
Deep learning models for predictive maintenance in Industry 4.0: An analysis
Vyas, B., Pranati, P., & Pareek, G. (2026) · Concurrent Engineering · doi:10.1177/1063293x261444587
Person Identification Based on Accelerations on Drawing Figures with a Smartphone
Takahashi, Y., Sonoyama, A., Kamiyamaton, T., et al. (2022) · International Conference on Ubiquitous Information Management and Communication (IMCOM) · doi:10.1109/imcom53663.2022.9721744
Chandrasekaran, K., Buquicchio, L., Agu, E., & Rundensteiner, E. (2025) · IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI ...) · doi:10.1109/bhi67747.2025.11269539
Mekruksavanich, S., & Jitpattanakul, A. (2026) · Technologies · doi:10.3390/technologies14050272 · Cited by 1
Xie, D., Yan, P., Li, Y., et al. (2026) · Sensors · doi:10.3390/s26185967
“DeepConvLSTM [25] models recurrent temporal context well but confuses downstairs with level walking, while LIMUNet [28]’s aggressive compression removes part of the class-specific detail needed to separate similar stance–swing patterns.”
de Biaz, B. M., Lopes, T. T. E., Fortes, M. Z., Ferreira, V. H., & de Mello, F. L. (2020) · Editora Poisson eBooks · doi:10.36229/978-65-86127-20-1.cap.09
State Prediction Method for A-Class Insulation Board Production Line Based on Transfer Learning
Wang, Y., Wang, H., Guo, X., Liu, X., & Liu, X. (2022) · Mathematics · doi:10.3390/math10203906
Shift-Aware Threshold Adaptation for Cross-Dataset Wearable Anomaly Detection
Nithish kumar, V., & Nathiya, R. (2026) · doi:10.1109/icscst70031.2026.11709770
“The concept of CNN & LSTM for activity recognition within the wearable sensor community was presented for the first time by Ordonez and Roggen [9].”
Classifying falls using out-of-distribution detection in human activity recognition
Roy, D., Komini, V., & Girdzijauskas, Š. (2023) · AI Communications · doi:10.3233/aic-220205 · Cited by 2
“Activity recognition (AR) with wearable sensors [5,36] is popularly achieved using deep learning methods [11, 14,23,30,35].”
Using Wearable IoT Devices to Analyze Healthcare Data for Human Activity Recognition
Laxmi, G. A., & Reddy, S. P. K. (2021) · International Conference on Electronics, Communication and Aerospace Technology (ICECA) · doi:10.1109/iceca52323.2021.9676006
Mohapatra, S., Singh, J. G., Samantaray, S. P., & Mishra, M. (2026) · Algorithms · doi:10.3390/a19050390 · Cited by 2
Huang, T., Lin, F., Ling, S., Liu, J., & Li, S. (2026) · Neurocomputing · doi:10.1016/j.neucom.2026.135022
Evaluation of Inertial and Physiological Data for Human Activity Recognition
Alcaraz-Galindo, I., Chávez-Campos, G. M., Morales-Cervantes, A., Fraga-Aguilar, M., & Masselli, Y. M. C. (2026) · Communications in computer and information science · doi:10.1007/978-3-032-37444-8_14
Li, X., Huang, Y., Su, L., Yao, S., Li, Y., & Zhao, J. (2026) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas66217.2026.11562792
“…DownSample Encode * Weight stride = 2, dilation = 1 learning enables end-to-end spatio–temporal feature extraction: CNNs capture local motion cues [7], LSTMs model long-term temporal dependencies [8], and hybrid CNN–LSTM architectures achieve 90–95% accuracy on public datasets such as SisFall [9].”
Xie, D., Yan, P., Li, Y., Wang, Q., & Liu, J. (2026) · Sensors · doi:10.3390/s26185879
“In particular, recent studies have indicated that increasing model size is a key factor in improving recognition performance, which inevitably leads to a significant rise in computational cost [13,14].”
Construction Crane Hand Signal Recognition Using Wearable EMG-IMU Sensor Fusion
Kim, T., & Yang, K. (2026) · IEEE Access · doi:10.1109/access.2026.3721095
“Because these modalities also differ in sampling rate, bandwidth, and morphology, this study adopted a late-fusion architecture in which each is processed by a dedicated 1D convolutional branch before integration at the feature level [28], [29].”
Andriansyah, G., Suratman, F. Y., & Istiqomah (2026) · IEEE Access · doi:10.1109/access.2026.3696665
“Although promising, most radar-based HAR and fall detection studies primarily emphasize training performance rather than practical real-time evaluation [8], [9], [10], [11], [12].”
Neuro-Mirror AI-Driven Wearable Robotic Glove for Adaptive Hand Rehabilitation Using Mirror Therapy
Al-Jada, R., Isleem, L., Alahmadl, A. A., & Aljarrah, A. (2026) · IEEE Jordan Conference on Applied Electrical Engineering and Computing Technologies · doi:10.1109/aeect69724.2026.11657857
Rivas-Caicedo, J. L., Saldaña-Aristizábal, L., Niño-Tejada, K., et al. (2025) · Lecture notes in networks and systems · doi:10.1007/978-3-032-09694-4_52
Mobile data analysis : robust alignment and flexible clustering methods
Yan, Y. (2024) · theses.fr (ABES) · doi:10.70675/22a43d74z31abz42e9za12ez21695cba9c63
Dankovich, L., Vaughn-Cooke, M., & Bergbreiter, S. (2022) · Sensors · doi:10.3390/s22197512
高, 婧. (2026) · Artificial Intelligence and Robotics Research · doi:10.12677/airr.2026.152063
Behavioral model inference of black-box software using deep neural networks
Mashhadi, M. J., Ataiefard, F., Hemmati, H., & Walkinshaw, N. (2021) · White Rose Research Online (University of Leeds, The University of Sheffield, University of York)
Revisiting Time-Domain Interpretability for Self-Supervised IMU Sensing Models
Wang, Y., Zhao, T., & Wang, X. (2026) · IEEE/ACM Conference on Connected Health: Applications, Systems and Engineering Technologies · doi:10.1109/chase69719.2026.00039
“CNNs, RNNs, and Transformers can produce highly accurate predictions, yet the sensor patterns and temporal cues driving those predictions are often unclear [5].”
End-to-End Human Activity Recognition using Convolutional Neural Network and Center Loss
Nguyen, T. (2020) · International Journal of Advanced Trends in Computer Science and Engineering · doi:10.30534/ijatcse/2020/182942020
“[10] used the LSTM to extract the changes of each activity flow time to recognize human activities.”
MethodLabel Leakage in Federated Inertial-based Human Activity Recognition
Bock, M., Hopp, M., Laerhoven, K. V., & Moeller, M. (2025) · ACM International Symposium on Wearable Computers (ISWC) · doi:10.1145/3715071.3750413
Amjad, H., Asif, N., Khan, U. S., et al. (2024) · International Conference on Robotics and Automation in Industry (ICRAI) · doi:10.1109/icrai62391.2024.10894302
“However, when examining the action sequences over several frames, 3D-based techniques consider both temporal and spatial features which improve the identification accuracy and capture the dynamic character of actions more [16].”
Nagar, M., Sikarwar, H., & Das, D. (2026) · IEEE Sensors Journal · doi:10.1109/jsen.2026.3694802
“The comparison includes DeepConvLSTM [16], self-attention HAR [32], METIER [18], adversarial CNN [34], TASKED [19], ConvNet-Boosting [35], GT-WHAR [36], and Cross-Att [22].”
Wang, X., Zheng, L., & Li, Z. (2026) · Frontiers in Medicine · doi:10.3389/fmed.2026.1759763
On the Limits of Lossy Compression for Human Activity Recognition in Sensor Networks
Schulz, J., Radak, H., Nguyen, P. T., Nguyen, G. T. K., & Fitzek, F. H. P. (2023) · Conference on Local Computer Networks · doi:10.1109/lcn58197.2023.10223374
Efficient cheap all-layer aggregation network for time-sensitive time series classification
Meng, H., Khairulov, T., Kim, D., & Lee, J. (2026) · Knowledge-Based Systems · doi:10.1016/j.knosys.2026.116016
Ariza-Colpas, P., Piñeres-Melo, M. A., Carrascal, A. I. O., & Díaz-Jiménez, D. (2026) · Sensors · doi:10.3390/s26123751
“In the field of HAR, transfer learning has gained increasing relevance due to the high cost and complexity of collecting large-scale, labeled datasets across varying contexts, such as different environments, sensor placements, or user populations [41,42].”
Human Activity Recognition Using a Convolutional Neural Network with Self-Attention Mechanism
Raj, A., Kumari, K. N., & Shreya, S. (2025) · International Conference on Research in Intelligent Computing in Engineering (RICE) · doi:10.1109/rice67503.2025.11439937
Selected Papers from IEEE ICKII 2018
Meen, T.-., & Zhao, W. (2019) · MDPI eBooks · doi:10.3390/books978-3-03921-274-3
Burchard, R., Ali, H., & Laerhoven, K. V. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-13312-0_18 · Cited by 2
Simulation-Driven Design of Smart Gloves for Gesture RecognitionPreprint
Leite, C. F. S., Byvshev, P., Mauranen, H., & Xiao, Y. (2022) · SSRN Electronic Journal · doi:10.2139/ssrn.4195252
EgoInertia-MI: A Multimodal Egocentric Vision and IMU Benchmark for Motor Impairment Assessment
Alhamdoosh, F., Pala, P., Mohamed, A., & Arvind, D. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-39379-1_13
Unsupervised Sensor-Based Human Activity Recognition via Contrastive Learning of Representation
Wang, K., Li, Z., Zhang, L., Gao, Q., & Wang, Z. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2026.3719110
“Ord´o˜nez and Roggen [33] introduced a deep learning framework integrating convolutional layers with LSTM units, allowing the model to simultaneously learn local feature representations and long-range temporal dependencies from multimodal sensor inputs.”
Physics-informed bio-gated fusion for robust multimodal human activity recognition
Liu, Y., Wang, Y., Qian, G., Chen, Y., & Yu, W. (2026) · Applied Soft Computing · doi:10.1016/j.asoc.2026.116595
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.