pith. sign in

arxiv: 2603.12296 · v2 · pith:T2PYWGF3new · submitted 2026-03-11 · 💻 cs.LG · cs.AI· eess.SP

Synthetic Data Generation for Brain-Computer Interfaces: Overview, Benchmarking, and Future Directions

Ziwei Wang , Zhentao He , Xingyi He , Hongbin Wang , Tianwang Jia , Jingwei Luo , Siyang Li , Xiaoqing Chen
show 1 more author
Dongrui Wu
This is my paper

Pith reviewed 2026-05-21 11:21 UTC · model grok-4.3

classification 💻 cs.LG cs.AIeess.SP
keywords synthetic databrain-computer interfacesdata generationmotor imageryepileptic seizure detectionsteady-state visually evoked potentialsauditory attention detection
0
0 comments X

The pith

Synthetic brain signals can mitigate data scarcity and improve brain-computer interface models.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper reviews approaches to generating synthetic brain signals that are physiologically plausible. It categorizes them into signal-transformation, feature-based, model-based, and translation-based methods. The authors benchmark these on four BCI paradigms to show their value for training models with limited real data. This matters because real neural recordings are scarce, heterogeneous, and raise privacy issues, so synthetic data could enable more practical and generalizable BCIs.

Core claim

Generating synthetic yet physiologically plausible brain signals has emerged as a promising strategy to mitigate data scarcity, improve model generalization, and support data-efficient BCIs. The survey categorizes existing approaches into four types and provides benchmark experiments across motor imagery, epileptic seizure detection, steady-state visually evoked potentials, and auditory attention detection to compare their downstream utility.

What carries the argument

A taxonomy of four generation approaches—signal-transformation-based, feature-based, model-based, and translation-based—along with benchmarks on representative BCI tasks that demonstrate their relative strengths for producing useful synthetic data.

If this is right

  • Models trained with synthetic data achieve better generalization across users and sessions.
  • Synthetic data reduces the need for extensive real recordings, lowering costs and privacy risks.
  • Evaluation metrics combining signal realism, physiological plausibility, and task performance guide better method selection.
  • Future systems can integrate these generation techniques for more data-efficient BCI development.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Hybrid methods combining multiple generation types could yield even more effective synthetic datasets.
  • The benchmarking results suggest potential applications in other signal-processing domains facing similar data limitations.
  • Privacy-aware generation might become standard in medical AI to comply with regulations on neural data.

Load-bearing premise

The four chosen BCI paradigms and representative generation methods are sufficiently representative to draw general conclusions about the utility of synthetic data across the field.

What would settle it

Demonstrating that in a held-out BCI task, models using synthetic data show no performance gain or even degradation compared to using only the available real data would falsify the central claim.

Figures

Figures reproduced from arXiv: 2603.12296 by Dongrui Wu, Hongbin Wang, Jingwei Luo, Siyang Li, Tianwang Jia, Xiaoqing Chen, Xingyi He, Zhentao He, Ziwei Wang.

Figure 1
Figure 1. Figure 1: Data scarcity issue in BCIs, owing to the small data si [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Data generation driven machine learning pipeline fo [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Four types of data generation approaches for brain si [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Visualizations of brain signals before (blue lines) [PITH_FULL_IMAGE:figures/full_fig_p003_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Model-based generation approaches for brain signal [PITH_FULL_IMAGE:figures/full_fig_p004_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Two types of translation-based generation for brain [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Evaluation framework of generated data in BCIs, enco [PITH_FULL_IMAGE:figures/full_fig_p015_7.png] view at source ↗
read the original abstract

Deep learning has achieved transformative performance across diverse domains, largely driven by large-scale and high-quality training data. In contrast, the development of brain-computer interfaces (BCIs) is fundamentally constrained by limited, heterogeneous, and privacy-sensitive neural recordings. Generating synthetic yet physiologically plausible brain signals has therefore emerged as a promising strategy to mitigate data scarcity, improve model generalization, and support data-efficient BCIs. This survey provides a comprehensive review of synthetic brain data generation for BCIs, covering methodological taxonomies, benchmark experiments, evaluation metrics, key applications, and future directions. We systematically categorize existing generation approaches into four types: signal-transformation-based, feature-based, model-based, and translation-based generation, and discuss their characteristics, advantages, and limitations. Furthermore, we benchmark representative brain signal generation approaches across four BCI paradigms, including motor imagery, epileptic seizure detection, steady-state visually evoked potentials, and auditory attention detection, to provide an objective comparison of their downstream utility. We also summarize evaluation principles for generated brain signals from multiple perspectives, including signal realism, physiological plausibility, downstream utility, and privacy preservation. Finally, we discuss the potential and challenges of current generation approaches and outline future research directions toward accurate, data-efficient, generalizable, and privacy-aware BCI systems. The benchmark codebase is available at https://github.com/wzwvv/DG4BCI.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The paper surveys synthetic data generation for brain-computer interfaces, taxonomizing approaches into signal-transformation-based, feature-based, model-based, and translation-based categories. It benchmarks representative methods from these categories on four BCI paradigms (motor imagery, epileptic seizure detection, SSVEP, auditory attention detection) to compare downstream task performance, summarizes evaluation metrics across realism, plausibility, utility, and privacy, and outlines future directions, with an accompanying open-source benchmark codebase.

Significance. If the benchmark results hold, the work offers a structured synthesis of the field and objective, reproducible comparisons of generation methods' utility for mitigating data scarcity in BCIs. The open codebase at https://github.com/wzwvv/DG4BCI is a clear strength for reproducibility and follow-on research. The taxonomy and multi-perspective evaluation framework could help standardize future studies, though the significance is tempered by the survey nature without new theoretical contributions.

major comments (1)
  1. [Benchmarking section] Benchmarking section: the central claim that synthetic generation is a promising strategy to mitigate data scarcity and improve generalization across BCIs rests on the downstream-utility comparisons, yet the manuscript provides no explicit justification, coverage analysis, or rationale for selecting these four paradigms (motor imagery, epileptic seizure detection, SSVEP, auditory attention detection) or the specific representative methods within each generation type. This selection is load-bearing for generalizing the benchmark conclusions, as the paper omits other common paradigms such as P300 or error-related potentials without discussing why the chosen set is representative or diverse enough to support broad utility claims.
minor comments (1)
  1. [Abstract] The abstract and introduction could more clearly distinguish between the survey synthesis and the novel benchmark contributions to help readers quickly identify the paper's incremental value.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive and detailed feedback. We address the major comment on the benchmarking section below. We agree that additional justification is warranted and will revise the manuscript to include it.

read point-by-point responses

  1. Referee: [Benchmarking section] Benchmarking section: the central claim that synthetic generation is a promising strategy to mitigate data scarcity and improve generalization across BCIs rests on the downstream-utility comparisons, yet the manuscript provides no explicit justification, coverage analysis, or rationale for selecting these four paradigms (motor imagery, epileptic seizure detection, SSVEP, auditory attention detection) or the specific representative methods within each generation type. This selection is load-bearing for generalizing the benchmark conclusions, as the paper omits other common paradigms such as P300 or error-related potentials without discussing why the chosen set is representative or diverse enough to support broad utility claims.

    Authors: We agree that the manuscript would benefit from an explicit rationale for the paradigm and method selections. These four paradigms were chosen to span diverse BCI signal characteristics and use cases: motor imagery (endogenous motor signals), SSVEP (exogenous visual evoked responses), auditory attention detection (auditory processing), and epileptic seizure detection (clinical pathological activity). This mix covers both healthy-subject and clinical applications while drawing on publicly available datasets that support reproducible benchmarking. For representative methods, we selected prominent, implementable approaches from each of the four taxonomy categories that had sufficient documentation or code to enable fair, controlled comparisons. In the revision we will add a new subsection (or expanded paragraph) in the Benchmarking section that states these criteria, notes the omission of P300 and error-related potentials (primarily due to dataset availability and scope constraints), and discusses the implications for generalizability. We will also qualify the utility claims more carefully as observations drawn from the selected set rather than universal assertions. revision: yes

Circularity Check

0 steps flagged

No significant circularity in survey and benchmarking structure

full rationale

This is a survey paper that taxonomizes prior generation methods into four categories, summarizes evaluation principles, and reports new benchmark results on downstream utility across four BCI paradigms. No equations, fitted parameters, or self-referential derivations appear in the abstract or described structure; the central claims rest on independent empirical benchmarking and synthesis of external literature rather than reducing to the paper's own inputs by construction. The selection of paradigms and methods is presented as representative without any load-bearing self-citation chain or uniqueness theorem imported from the authors' prior work.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

As a review and benchmarking study, the central claims rest on the completeness of the literature synthesis and the representativeness of the selected methods and tasks rather than on new axioms or free parameters.

pith-pipeline@v0.9.0 · 5808 in / 1045 out tokens · 48916 ms · 2026-05-21T11:21:36.977102+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

145 extracted references · 145 canonical work pages · 2 internal anchors

  1. [1]

    Brain leaks an d consumer neurotechnology,

    M. Ienca, P . Haselager, and E. J. Emanuel, “Brain leaks an d consumer neurotechnology,” Nature Biotechnology, vol. 36, no. 9, pp. 805–810, 2018

    work page 2018

  2. [2]

    Language models are few-shot learners,

    T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P . Dh ariwal et al. , “Language models are few-shot learners,” in Proc. Advances in Neural Information Processing Systems , V ancouver, Canada, Dec. 2020, pp. 1877–1901

    work page 2020

  3. [3]

    The removal o f EOG artifacts from EEG signals using independent component analysis and multivariate empirical mode decomposition,

    G. Wang, C. Teng, K. Li, Z. Zhang, and X. Y an, “The removal o f EOG artifacts from EEG signals using independent component analysis and multivariate empirical mode decomposition,” IEEE Journal of Biomedical and Health Informatics , vol. 20, no. 5, pp. 1301–1308, 2015

    work page 2015

  4. [4]

    V ME- DWT: An efficient algorithm for detection and elimination of eye blink from short segments of single EEG channel,

    M. Shahbakhti, M. Beiramvand, M. Nazari, A. Broniec-W´ o jcik, P . Au- gustyniak, A. S. Rodrigues, M. Wierzchon, and V . Marozas, “V ME- DWT: An efficient algorithm for detection and elimination of eye blink from short segments of single EEG channel,” IEEE Trans. on Neural Systems and Rehabilitation Engineering , vol. 29, pp. 408–417, 2021

    work page 2021

  5. [5]

    Artifact and recording concepts in EEG,

    W. O. Tatum, B. A. Dworetzky, and D. L. Schomer, “Artifact and recording concepts in EEG,” Journal of Clinical Neurophysiology , vol. 28, no. 3, pp. 252–263, 2011

    work page 2011

  6. [6]

    Self-su pervised EEG denoising via dual-branch consistency learning with ma sked reconstruction,

    W. Chen, C. Y u, Z. Zhao, N. Zheng, H. Li, and Y . Li, “Self-su pervised EEG denoising via dual-branch consistency learning with ma sked reconstruction,” Knowledge-Based Systems, vol. 330, p. 114703, 2025

    work page 2025

  7. [7]

    Data augmenta tion for enhancing EEG-based emotion recognition with deep gene rative models,

    Y . Luo, L.-Z. Zhu, Z.-Y . Wan, and B.-L. Lu, “Data augmenta tion for enhancing EEG-based emotion recognition with deep gene rative models,” Journal of Neural Engineering , vol. 17, no. 5, p. 056021, 2020

    work page 2020

  8. [8]

    Data augmenta- tion for learning predictive models on EEG: A systematic com parison,

    C. Rommel, J. Paillard, T. Moreau, and A. Gramfort, “Data augmenta- tion for learning predictive models on EEG: A systematic com parison,” Journal of Neural Engineering , vol. 19, no. 6, p. 066020, 2022

    work page 2022

  9. [9]

    Time-frequency transf orm based EEG data augmentation for brain-computer interfaces,

    Z. Wang, S. Li, X. Chen, and D. Wu, “Time-frequency transf orm based EEG data augmentation for brain-computer interfaces,” Knowledge- Based Systems , vol. 311, p. 113074, 2025

    work page 2025

  10. [10]

    Data augmentation for EEG-based emotion recognition with deep convolutional neural net- works,

    F. Wang, S.-h. Zhong, J. Peng, J. Jiang, and Y . Liu, “Data augmentation for EEG-based emotion recognition with deep convolutional neural net- works,” in Proc. Int’l Conf. Multimedia Modeling , Bangkok, Thailand, Feb. 2018, pp. 82–93

    work page 2018

  11. [11]

    Contrastive repre- sentation learning for electroencephalogram classificati on,

    M. N. Mohsenvand, M. R. Izadi, and P . Maes, “Contrastive repre- sentation learning for electroencephalogram classificati on,” in Proc. Advances in Neural Information Processing Systems Machine Learning for Health W orkshops, V ancouver, Canada, Dec. 2020, pp. 238–253

    work page 2020

  12. [12]

    Multi-source decentraliz ed transfer for privacy-preserving BCIs,

    W. Zhang, Z. Wang, and D. Wu, “Multi-source decentraliz ed transfer for privacy-preserving BCIs,” IEEE Trans. Neural Systems and Reha- bilitation Engineering , vol. 30, pp. 2710–2720, 2022

    work page 2022

  13. [13]

    Addressing Class Imbalance in Classification Problems of Noisy Signals by using Fourier Transform Surrogates

    J. T. Schwabedal, J. C. Snyder, A. Cakmak, S. Nemati, and G. D. Clifford, “Addressing class imbalance in classification pr oblems of noisy signals by using fourier transform surrogates,” arXiv preprint arXiv:1806.08675, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  14. [14]

    Seizu re onset zone classification based on imbalanced iEEG with data augme ntation,

    X. Zhao, J. Sole Casals, H. Sugano, and T. Tanaka, “Seizu re onset zone classification based on imbalanced iEEG with data augme ntation,” Journal of Neural Engineering , vol. 19, no. 6, p. 065001, 2022

    work page 2022

  15. [15]

    Channel reflecti on: Knowledge-driven data augmentation for EEG-based brain-c omputer interfaces,

    Z. Wang, S. Li, J. Luo, J. Liu, and D. Wu, “Channel reflecti on: Knowledge-driven data augmentation for EEG-based brain-c omputer interfaces,” Neural Networks , vol. 176, p. 106351, 2024

    work page 2024

  16. [16]

    Rotational data augmentation for electroencephalographic data,

    M. M. Krell and S. K. Kim, “Rotational data augmentation for electroencephalographic data,” in Int’l Conf. of the IEEE Engineering in Medicine and Biology Society , Jeju Island, Korea, July 2017, pp. 471–474

    work page 2017

  17. [17]

    Data augmentation: Using channel-level recombina tion to improve classification performance for motor imagery EEG,

    Y . Pei, Z. Luo, Y . Y an, H. Y an, J. Jiang, W. Li, L. Xie, and E. Yin, “Data augmentation: Using channel-level recombina tion to improve classification performance for motor imagery EEG,” Frontiers in Human Neuroscience , vol. 15, p. 645952, 2021

    work page 2021

  18. [18]

    Unsupervise d domain adaptation for cross-patient seizure classification,

    Z. Wang, W. Zhang, S. Li, X. Chen, and D. Wu, “Unsupervise d domain adaptation for cross-patient seizure classification,” Journal of Neural Engineering, vol. 20, no. 6, p. 066002, 2023

    work page 2023

  19. [19]

    Accurate mental stress detection using sequential backwa rd selection and adaptive synthetic methods,

    H.-C. Tseng, K.-Y . Tai, Y .-Z. Ma, L.-D. V an, L.-W. Ko, and T.-P . Jung, “Accurate mental stress detection using sequential backwa rd selection and adaptive synthetic methods,” IEEE Trans. Neural Systems and Rehabilitation Engineering , vol. 32, pp. 3095–3103, 2024

    work page 2024

  20. [20]

    Mix up: Beyond empirical risk minimization,

    H. Zhang, M. Cisse, Y . N. Dauphin, and D. Lopez-Paz, “Mix up: Beyond empirical risk minimization,” in Proc. Int’l Conf. on Learning Representations, V ancouver, Canada, Apr. 2018, pp. 1–13

    work page 2018

  21. [21]

    Data augmentation for self-pa ced motor imagery classification with C-LSTM,

    D. Freer and G.-Z. Y ang, “Data augmentation for self-pa ced motor imagery classification with C-LSTM,” Journal of Neural Engineering , vol. 17, no. 1, p. 016041, 2020. IEEE TRANSACTIONS ON ... 18

    work page 2020

  22. [22]

    Manifold Mixup: Better representation s by interpolating hidden states,

    V . V erma, A. Lamb, C. Beckham, A. Najafi, I. Mitliagkas, D . Lopez- Paz, and Y . Bengio, “Manifold Mixup: Better representation s by interpolating hidden states,” in Proc. Int’l Conf. on Machine Learning , Long Beach, CA, USA, Jun. 2019, pp. 6438–6447

    work page 2019

  23. [23]

    Domain-adversarial tr aining of neural networks,

    Y . Ganin, E. Ustinova, H. Ajakan, P . Germain, H. Laroche lle, F. Lavi- olette, M. March, and V . Lempitsky, “Domain-adversarial tr aining of neural networks,” Journal of Machine Learning Research , vol. 17, no. 59, pp. 1–35, 2016

    work page 2016

  24. [24]

    Auto-encoding variationa l bayes,

    D. P . Kingma and M. Welling, “Auto-encoding variationa l bayes,” in Int’l Conf. on Learning Representations , Banff, Canada, Apr. 2014, pp. 1–14

    work page 2014

  25. [25]

    Xlnet: Generalized autoregressive pretraining f or language understanding,

    Z. Y ang, Z. Dai, Y . Y ang, J. Carbonell, R. R. Salakhutdin ov, and Q. V . Le, “Xlnet: Generalized autoregressive pretraining f or language understanding,” in Proc. Advances in Neural Information Processing Systems, V ancouver, BC, Canada, Dec. 2019

    work page 2019

  26. [26]

    Denoising diffusion prob abilistic models,

    J. Ho, A. Jain, and P . Abbeel, “Denoising diffusion prob abilistic models,” in Proc. Advances in Neural Information Processing Systems , V ancouver, Canada, Dec. 2020, pp. 6840–6851

    work page 2020

  27. [27]

    Rapid adaptation of brain–computer interfaces to new neuronal ensembles or participants via generative modelling,

    S. Wen, A. Yin, T. Furlanello, M. G. Perich, L. E. Miller, and L. Itti, “Rapid adaptation of brain–computer interfaces to new neuronal ensembles or participants via generative modelling,” Nature Biomedical Engineering, vol. 7, no. 4, pp. 546–558, 2023

    work page 2023

  28. [28]

    Generative adversarial networks-based data augmentatio n for brain– computer interface,

    F. Fahimi, S. Dosen, K. K. Ang, N. Mrachacz-Kersting, an d C. Guan, “Generative adversarial networks-based data augmentatio n for brain– computer interface,” IEEE Trans. Neural Networks and Learning Systems, vol. 32, no. 9, pp. 4039–4051, 2020

    work page 2020

  29. [29]

    Semi-supervis ed generative and discriminative adversarial learning for motor imagery -based brain– computer interface,

    W. Ko, E. Jeon, J. S. Y oon, and H.-I. Suk, “Semi-supervis ed generative and discriminative adversarial learning for motor imagery -based brain– computer interface,” Scientific Reports , vol. 12, no. 1, p. 4587, 2022

    work page 2022

  30. [30]

    Optimizing SSVEP-based BCI training through adv ersarial generative neural networks,

    G. Figueiredo, S. N. Carvalho, G. V argas, V . Barbosa, C. Peixoto, and H. Leite, “Optimizing SSVEP-based BCI training through adv ersarial generative neural networks,” Int’l Journal of Electrical and Computer Engineering Research, vol. 3, no. 4, pp. 8–14, 2023

    work page 2023

  31. [31]

    Task-oriente d EEG denoising generative adversarial network for enhancing SS VEP-BCI performance,

    P . Zeng, L. Fan, Y . Luo, H. Shen, and D. Hu, “Task-oriente d EEG denoising generative adversarial network for enhancing SS VEP-BCI performance,” Journal of Neural Engineering, vol. 21, no. 6, p. 066003, 2024

    work page 2024

  32. [32]

    Leveraging synthetic subject invariant eeg signals for zero calibration BCI,

    N. K. N. Aznan, A. Atapour-Abarghouei, S. Bonner, J. D. C onnolly, and T. P . Breckon, “Leveraging synthetic subject invariant eeg signals for zero calibration BCI,” in Int’l Conf. on Pattern Recognition , Virtual, Mar. 2021, pp. 10 418–10 425

    work page 2021

  33. [33]

    Brain-supervised conditional g enerative mod- eling,

    J. Ma and T. Ruotsalo, “Brain-supervised conditional g enerative mod- eling,” IEEE Trans. on Human-Machine Systems , vol. 55, no. 3, pp. 383–393, 2025

    work page 2025

  34. [34]

    E nhanced generative adversarial networks for unseen word generatio n from EEG signals,

    Y .-E. Lee, S.-H. Lee, S. Kim, J.-S. Lee, and D.-S. Kim, “E nhanced generative adversarial networks for unseen word generatio n from EEG signals,” in Int’l Winter Conf. on Brain-Computer Interface , Gangwon, South Korea, Feb. 2024, pp. 1–4

    work page 2024

  35. [35]

    Neural decoding of ima gined speech and visual imagery as intuitive paradigms for BCI com munica- tion,

    S.-H. Lee, M. Lee, and S.-W. Lee, “Neural decoding of ima gined speech and visual imagery as intuitive paradigms for BCI com munica- tion,” IEEE Trans. on Neural Systems and Rehabilitation Engineeri ng, vol. 28, no. 12, pp. 2647–2659, 2020

    work page 2020

  36. [36]

    Speech synthesis from brain signals based on generative mo del,

    Y .-E. Lee, S.-H. Kim, S.-H. Lee, J.-S. Lee, S. Kim, and S. -W. Lee, “Speech synthesis from brain signals based on generative mo del,” in Int’l Winter Conf. on Brain-Computer Interface , Gangwon, South Korea, Feb. 2023, pp. 1–4

    work page 2023

  37. [37]

    Common spatial gene rative adversarial networks based EEG data augmentation for cross -subject brain-computer interface,

    Y . Song, L. Y ang, X. Jia, and L. Xie, “Common spatial gene rative adversarial networks based EEG data augmentation for cross -subject brain-computer interface,” arXiv preprint arXiv:2102.04456 , 2021

    work page arXiv 2021

  38. [38]

    EEG-GAN: Generative adversarial networks for electroencephalograhic (EEG) brain signals

    K. G. Hartmann, R. T. Schirrmeister, and T. Ball, “EEG-G AN: Gen- erative adversarial networks for electroencephalograhic (EEG) brain signals,” arXiv preprint arXiv:1806.01875 , 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  39. [39]

    Generative adversarial networks conditioned by brain sig nals,

    S. Palazzo, C. Spampinato, I. Kavasidis, D. Giordano, a nd M. Shah, “Generative adversarial networks conditioned by brain sig nals,” in Proc. IEEE Int’l Conf. on Computer Vision , V enice, Italy, Oct. 2017, pp. 3410–3418

    work page 2017

  40. [40]

    SleepEGAN: A GAN- enhanced ensemble deep learning model for imbalanced class ification of sleep stages,

    X. Cheng, K. Huang, Y . Zou, and S. Ma, “SleepEGAN: A GAN- enhanced ensemble deep learning model for imbalanced class ification of sleep stages,” Biomedical Signal Processing and Control , vol. 92, p. 106020, 2024

    work page 2024

  41. [41]

    SynSigGAN: Generative advers arial net- works for synthetic biomedical signal generation,

    D. Hazra and Y .-C. Byun, “SynSigGAN: Generative advers arial net- works for synthetic biomedical signal generation,” Biology, vol. 9, no. 12, p. 441, 2020

    work page 2020

  42. [42]

    Boosting EEG and ECG clas- sification with synthetic biophysical data generated via ge nerative adversarial networks,

    A. V enugopal and D. Resende Faria, “Boosting EEG and ECG clas- sification with synthetic biophysical data generated via ge nerative adversarial networks,” Applied Sciences , vol. 14, no. 23, p. 10818, 2024

    work page 2024

  43. [43]

    BWG AN- GP: An EEG data generation method for class imbalance proble m in RSVP tasks,

    M. Xu, Y . Chen, Y . Wang, D. Wang, Z. Liu, and L. Zhang, “BWG AN- GP: An EEG data generation method for class imbalance proble m in RSVP tasks,” IEEE Trans. on Neural Systems and Rehabilitation Engineering, vol. 30, pp. 251–263, 2022

    work page 2022

  44. [44]

    A study on mental state classification using EEG-based brain- machine interface,

    J. J. Bird, L. J. Manso, E. P . Ribeiro, A. Ekart, and D. R. F aria, “A study on mental state classification using EEG-based brain- machine interface,” in Int’l Conf. on Intelligent Systems , Funchal, Madeira, Portugal, Sep. 2018, pp. 795–800

    work page 2018

  45. [45]

    Brain2image: Converting brain signals into images,

    I. Kavasidis, S. Palazzo, C. Spampinato, D. Giordano, a nd M. Shah, “Brain2image: Converting brain signals into images,” in Proc. of ACM Int’l Conf. on Multimedia , Mountain View, CA USA, Oct. 2017, pp. 1809–1817

    work page 2017

  46. [46]

    A generative model to synthesize EEG data for epileptic seizu re predic- tion,

    K. Rasheed, J. Qadir, T. J. O’Brien, L. Kuhlmann, and A. R azi, “A generative model to synthesize EEG data for epileptic seizu re predic- tion,” IEEE Trans. on Neural Systems and Rehabilitation Engineeri ng, vol. 29, pp. 2322–2332, 2021

    work page 2021

  47. [47]

    Dual- encoder V AE-GAN with spatiotemporal features for emotiona l EEG data augmentation,

    C. Tian, Y . Ma, J. Cammon, F. Fang, Y . Zhang, and M. Meng, “ Dual- encoder V AE-GAN with spatiotemporal features for emotiona l EEG data augmentation,” IEEE Trans.on Neural Systems and Rehabilitation Engineering, vol. 31, pp. 2018–2027, 2023

    work page 2018

  48. [48]

    EEG2V ec: Learning affective EE G representations via variational autoencoders,

    D. Bethge, P . Hallgarten, T. Grosse-Puppendahl, M. Kar i, L. L. Chuang, O. ¨Ozdenizci, and A. Schmidt, “EEG2V ec: Learning affective EE G representations via variational autoencoders,” in IEEE Int’l Conf. on Systems, Man, and Cybernetics , Prague, Czech Republic, Oct. 2022, pp. 3150–3157

    work page 2022

  49. [49]

    Causal recurrent variati onal autoencoder for medical time series generation,

    H. Li, S. Y u, and J. Principe, “Causal recurrent variati onal autoencoder for medical time series generation,” in Proc. AAAI Conf. on Artificial Intelligence, Washington, DC, USA, Feb. 2023, pp. 8562–8570

    work page 2023

  50. [50]

    Emergent network topology at seizure onset in humans,

    M. A. Kramer, E. D. Kolaczyk, and H. E. Kirsch, “Emergent network topology at seizure onset in humans,” Epilepsy Research , vol. 79, no. 2-3, pp. 173–186, 2008

    work page 2008

  51. [51]

    V AEEG: V ariational auto-encoder for extracting EEG representation,

    T. Zhao, Y . Cui, T. Ji, J. Luo, W. Li, J. Jiang, Z. Gao, W. Hu , Y . Y an, Y . Jiang et al. , “V AEEG: V ariational auto-encoder for extracting EEG representation,” NeuroImage, vol. 304, p. 120946, 2024

    work page 2024

  52. [52]

    Conditional V AE for personalized neurofeed back in cognitive training,

    I. E. Tibermacine, S. Russo, G. Scarano, G. Tedesco, A. R abehi, A. A. Alhussan, D. S. Khafaga, M. M. Eid, E.-S. M. El-Kenawy, and C. Napoli, “Conditional V AE for personalized neurofeed back in cognitive training,” PLoS One , vol. 20, no. 10, p. e0335364, 2025

    work page 2025

  53. [53]

    Unsupervised seizure identification on EEG,

    ˙ I. Yıldız, R. Garner, M. Lai, and D. Duncan, “Unsupervised seizure identification on EEG,” Computer Methods and Programs in Biomedicine, vol. 215, p. 106604, 2022

    work page 2022

  54. [54]

    Neuro-GPT: Towards a foundation model for EEG,

    W. Cui, W. Jeong, P . Th¨ olke, T. Medani, K. Jerbi, A. A. Jo shi, and R. M. Leahy, “Neuro-GPT: Towards a foundation model for EEG, ” in IEEE Int’l Symposium on Biomedical Imaging , Athens, Greece, May 2024, pp. 1–5

    work page 2024

  55. [55]

    Synthetic biological signals machine-generated by GPT-2 improve the classification of EEG and EMG through data augmentation,

    J. J. Bird, M. Pritchard, A. Fratini, A. Ek´ art, and D. R. Faria, “Synthetic biological signals machine-generated by GPT-2 improve the classification of EEG and EMG through data augmentation, ” IEEE Robotics and Automation Letters , vol. 6, no. 2, pp. 3498–3504, 2021

    work page 2021

  56. [56]

    Electromyog- raphy signal-based gesture recognition for human-machine interaction in real-time through model calibration,

    C. Dolopikos, M. Pritchard, J. J. Bird, and D. R. Faria, “ Electromyog- raphy signal-based gesture recognition for human-machine interaction in real-time through model calibration,” in Future of Information and Communication Conf. , Virtual, Apr. 2021, pp. 898–914

    work page 2021

  57. [57]

    Epileptic se izure prediction by synthesizing EEG signals through GPT,

    R. Niu, Y . Wang, H. Xi, Y . Hao, and M. Zhang, “Epileptic se izure prediction by synthesizing EEG signals through GPT,” in Proc. Int’l Conf. on Artificial Intelligence and Pattern Recognition , Yibin, China, Aug. 2021, pp. 419–423

    work page 2021

  58. [58]

    MEG-GPT: A transformer-based foundation model for magnetoencephal ography data,

    R. Huang, S. Cho, C. Gohil, O. P . Jones, and M. Woolrich, “ MEG-GPT: A transformer-based foundation model for magnetoencephal ography data,” arXiv preprint arXiv:2510.18080 , 2025

    work page arXiv 2025

  59. [59]

    Fo undational GPT model for MEG,

    R. Csaky, M. W. J. van Es, O. P . Jones, and M. Woolrich, “Fo undational GPT model for MEG,” arXiv preprint arXiv:2404.09256 , 2024

    work page arXiv 2024

  60. [60]

    Com- parison of deep neural networks to spatio-temporal cortica l dynamics of human visual object recognition reveals hierarchical corr espondence,

    R. M. Cichy, A. Khosla, D. Pantazis, A. Torralba, and A. O liva, “Com- parison of deep neural networks to spatio-temporal cortica l dynamics of human visual object recognition reveals hierarchical corr espondence,” Scientific Reports , vol. 6, no. 1, p. 27755, 2016

    work page 2016

  61. [61]

    Thought2Text: Text generation from EEG signal using large language models,

    A. Mishra, S. Shukla, J. Torres, J. Gwizdka, and S. Roych owdhury, “Thought2Text: Text generation from EEG signal using large language models,” in Proc. Annu. Conf. of the Nations of the Americas Chapter of the Assoc. for Comput. Linguist. , Albuquerque, New Mexico, Apr. 2025, pp. 3747–3759

    work page 2025

  62. [62]

    GET: A generative EEG transformer for continu ous context-based neural signals,

    O. Ali, M. Saif-ur Rehman, M. Metzler, T. Glasmachers, I . Iossifidis, and C. Klaes, “GET: A generative EEG transformer for continu ous context-based neural signals,” arXiv preprint arXiv:2406.03115 , 2024

    work page arXiv 2024

  63. [63]

    Multivariate autoregressive model estimatio n for high- IEEE TRANSACTIONS ON ... 19 dimensional intracranial electrophysiological data,

    C. M. Endemann, B. M. Krause, K. V . Nourski, M. I. Banks, a nd B. V an V een, “Multivariate autoregressive model estimatio n for high- IEEE TRANSACTIONS ON ... 19 dimensional intracranial electrophysiological data,” NeuroImage, vol. 254, p. 119057, 2022

    work page 2022

  64. [64]

    Enhanced BCI performance using diffusion model for EEG generation,

    Y . Zhong, L. Y ao, and Y . Wang, “Enhanced BCI performance using diffusion model for EEG generation,” in Int’l Conf. of the IEEE Engineering in Medicine and Biology Society , Houston, TX, USA, Dec. 2024, pp. 1–5

    work page 2024

  65. [65]

    Generative modeling and aug mentation of EEG signals using improved diffusion probabilistic models ,

    S. Torma and L. Szegletes, “Generative modeling and aug mentation of EEG signals using improved diffusion probabilistic models ,” Journal of Neural Engineering , vol. 22, no. 1, p. 016001, 2025

    work page 2025

  66. [66]

    Generativ e adver- sarial network with adaptive synthesis for brain-computer interfaces in motor imagery classification,

    Q. Zhou, X. Y e, S. Li, Y . Zhao, Q. Du, and L. Ke, “Generativ e adver- sarial network with adaptive synthesis for brain-computer interfaces in motor imagery classification,” in IEEE Int’l Conf. on Acoustics, Speech and Signal Processing . Hyderabad, India: IEEE, Apr. 2025, pp. 1–5

    work page 2025

  67. [67]

    Motor-intent decoding from synthetic EEG data using denoising diffusion probabilisti c models,

    M. D. Dere, J.-H. Jo, and B. Lee, “Motor-intent decoding from synthetic EEG data using denoising diffusion probabilisti c models,” Expert Systems with Applications , p. 130134, 2025

    work page 2025

  68. [68]

    Brain-driven representation learning base d on diffusion model,

    S. Kim, S.-H. Lee, Y .-E. Lee, J.-W. Lee, J.-H. Park, P . Ka zanzides, and S.-W. Lee, “Brain-driven representation learning base d on diffusion model,” in Int’l Winter Conf. on Brain-Computer Interface , Gangwon, South Korea, Feb. 2024, pp. 1–4

    work page 2024

  69. [69]

    Diff-E: Diffus ion-based learning for decoding imagined speech EEG,

    S. Kim, Y . E. Lee, S. H. Lee, and S. W. Lee, “Diff-E: Diffus ion-based learning for decoding imagined speech EEG,” in Proc. of the Annual Conf. of the Int’l Speech Communication Association , Dublin, Ireland, Aug. 2023, pp. 1159–1163

    work page 2023

  70. [70]

    Data augmentation for seizure prediction with generative diffu sion model,

    K. Shu, L. Wu, Y . Zhao, A. Liu, R. Qian, and X. Chen, “Data augmentation for seizure prediction with generative diffu sion model,” IEEE Trans. on Cognitive and Developmental Systems , vol. 17, no. 3, pp. 577–591, 2024

    work page 2024

  71. [71]

    Synthe tic sleep EEG signal generation using latent diffusion models,

    B. Aristimunha, R. Y . de Camargo, S. Chevallier, O. Luce na, A. G. Thomas, M. J. Cardoso, W. H. L. Pinaya, and J. Dafflon, “Synthe tic sleep EEG signal generation using latent diffusion models, ” in Proc. Advances in Neural Information Processing Systems DGM4H W o rk- shop, New Orleans, LA, USA, Dec. 2023

    work page 2023

  72. [72]

    Diffusion model-based multi-channel EEG representation and forecasting for early epileptic seizur e warning,

    Z. Jiang, W. Dai, Q. Wei, Z. Qin, R. Wei, M. Li, X. Chen, Y . H uo, J. Liu, K. Li et al. , “Diffusion model-based multi-channel EEG representation and forecasting for early epileptic seizur e warning,” Interdisciplinary Sciences: Computational Life Sciences , pp. 1–12, 2025

    work page 2025

  73. [73]

    Synthesizing EEG signals from event-related potential paradigms with co nditional diffusion models,

    G. Klein, P . Guetschel, G. Silvestri, and M. Tangermann , “Synthesizing EEG signals from event-related potential paradigms with co nditional diffusion models,” arXiv preprint arXiv:2403.18486 , 2024

    work page arXiv 2024

  74. [74]

    A diffusion-based f eature enhancement approach for driving behavior classification w ith EEG data,

    T. Liu, Y . Qin, S. Zhang, and X. Tao, “A diffusion-based f eature enhancement approach for driving behavior classification w ith EEG data,” Advanced Engineering Informatics , vol. 65, p. 103279, 2025

    work page 2025

  75. [75]

    Uncon ditional EEG synthesis based on diffusion models for sound generation,

    E. I. Chetkin, B. L. Kozyrsky, and S. L. Shishkin, “Uncon ditional EEG synthesis based on diffusion models for sound generation,” in Int’l Multi-Conference on Engineering, Computer and Informatio n Sciences. Akademgorodok, Russia: IEEE, Sep. 2024, pp. 416–420

    work page 2024

  76. [76]

    Naturalistic music decoding from ee g data via latent diffusion models,

    E. Postolache, N. Polouliakh, H. Kitano, A. Connelly, E . Rodol` a, L. Cosmo, and T. Akama, “Naturalistic music decoding from ee g data via latent diffusion models,” in IEEE Int’l Conf. on Acoustics, Speech and Signal Processing , Hyderabad, India, Apr. 2025, pp. 1–5

    work page 2025

  77. [77]

    Dr eamDif- fusion: Generating high-quality images from brain EEG sign als,

    Y . Bai, X. Wang, Y .-p. Cao, Y . Ge, C. Y uan, and Y . Shan, “Dr eamDif- fusion: Generating high-quality images from brain EEG sign als,” arXiv preprint arXiv:2306.16934, 2023

    work page arXiv 2023

  78. [78]

    Generative adversar ial nets,

    I. J. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. W arde-Farley, S. Ozair, A. Courville, and Y . Bengio, “Generative adversar ial nets,” in Proc. Advances in Neural Information Processing Systems , vol. 27, Montreal, Canada, Dec. 2014

    work page 2014

  79. [79]

    Least squares generative adversarial networks,

    X. Mao, Q. Li, H. Xie, R. Y . K. Lau, Z. Wang, and S. Paul Smol ley, “Least squares generative adversarial networks,” in Proc. of the IEEE Int’l Conf. on Computer Vision , V enice, Italy, Oct. 2017, pp. 2794– 2802

    work page 2017

  80. [80]

    Adversarial a udio syn- thesis,

    C. Donahue, J. McAuley, and M. Puckette, “Adversarial a udio syn- thesis,” in Proc. Int’l Conf. on Learning Representations , V ancouver, Canada, Apr. 2018, pp. 1–16

    work page 2018

Showing first 80 references.