pith. sign in

arxiv: 2511.23384 · v4 · submitted 2025-11-28 · 💻 cs.HC

Improving motor imagery decoding methods for an EEG-based mobile brain-computer interface in the context of the 2024 Cybathlon

Isabel Whiteley Tscherniak , Niels Christopher Thiemann , Ana McWhinnie-Fern\'andez , Iustin Curcean , Leon Jokinen , Sadat Hodzic , Thomas E. Huber , Daniel Pavlov
show 7 more authors
Manuel Methasani Pietro Marcolongo Glenn Viktor Krafczyk Oscar Osvaldo Soto Rivera Thien Le Flaminia Pallotti Enrico A. Fazzi
This is my paper

Pith reviewed 2026-05-17 04:20 UTC · model grok-4.3

classification 💻 cs.HC
keywords brain-computer interfacemotor imageryEEGS4Ddeep learningCybathlonreal-time controlportable BCI
0
0 comments X

The pith

An S4D-layer classifier decodes three motor imagery classes from EEG to drive a modular mobile brain-computer interface at up to 84% offline accuracy.

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

The paper develops a complete online EEG pipeline for the 2024 Cybathlon that lets a tetraplegic user control up to five signals through three mental and motor imagery tasks. Data move through acquisition, preprocessing, a deep classifier of three diagonalized structured state-space sequence layers, and a transfer function that turns class outputs into commands, all wrapped in a training game and live mobile feedback app. Offline tests reach 84% accuracy; after the competition the same pipeline yields 73% real-time success with the original pilot plus one new participant and beats reference machine-learning models. Readers care because the design deliberately uses low-cost hardware, user-centered feedback, and modular blocks to move brain-computer interfaces out of the lab and into daily settings for people with severe mobility loss.

Core claim

The authors establish that three diagonalized structured state-space sequence layers can serve as an effective real-time classifier for three classes of motor and mental imagery in EEG signals, delivering up to 84% offline accuracy, enabling a full modular pipeline that completed tasks during the Cybathlon and reached 73% success in post-competition real-time gameplay with two participants while outperforming reference models.

What carries the argument

Three diagonalized structured state-space sequence layers acting as the deep learning classifier that maps preprocessed EEG signals to the three imagery classes and onward to control dimensions.

If this is right

  • Three imagery classes can be mapped to as many as five independent control signals through the transfer function.
  • The modular structure supports online operation, live user feedback, and rapid integration of low-cost acquisition hardware.
  • The S4D classifier trains faster than the EEGEncoder baseline while exceeding the accuracy of conventional reference models.
  • The same pipeline bridges competition performance to post-event validation, indicating readiness for portable daily-life use.

Where Pith is reading between the lines

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

  • The emphasis on minimal recalibration suggests the pipeline could support unsupervised home rehabilitation if environmental noise remains manageable.
  • Adding simple adaptation layers for stress or background interference might raise the 73% real-time rate without changing the core classifier.
  • Longitudinal testing over weeks rather than single sessions would reveal whether the reported success rate persists under sustained daily use.

Load-bearing premise

Observed offline accuracy and post-competition real-time success rates will hold for new users and repeated daily operation without per-user recalibration or controlled environments.

What would settle it

Classification accuracy falling below 60% or real-time success dropping under 50% when the identical pipeline is tested on five new tetraplegic users across multiple sessions in ordinary home settings would show the reported performance does not generalize.

read the original abstract

Motivated by the Cybathlon 2024 competition, we developed a modular, online EEG-based brain-computer interface to address these challenges, increasing accessibility for individuals with severe mobility impairments. Our system uses three mental and motor imagery classes to control up to five control signals. The pipeline consists of four modules: data acquisition, preprocessing, classification, and the transfer function to map classification output to control dimensions. We use three diagonalized structured state-space sequence layers as a deep learning classifier. We developed a training game for our pilot where the mental tasks control the game during quick-time events. We implemented a mobile web application for live user feedback. The components were designed with a human-centred approach in collaboration with the tetraplegic user. We achieve up to 84% classification accuracy in offline analysis using an S4D-layer-based model. In a competition setting, our pilot successfully completed one task; we attribute the reduced performance in this context primarily to factors such as stress and the challenging competition environment. Following the Cybathlon, we further validated our pipeline with the original pilot and an additional participant, achieving a success rate of 73% in real-time gameplay. We also compare our model to the EEGEncoder, which is slower in training but has a higher performance. The S4D model outperforms the reference machine learning models. We provide insights into developing a framework for portable BCIs, bridging the gap between the laboratory and daily life. Specifically, our framework integrates modular design, real-time data processing, user-centred feedback, and low-cost hardware to deliver an accessible and adaptable BCI solution, addressing critical gaps in current BCI applications.

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

2 major / 3 minor

Summary. The manuscript describes the development of a modular, online EEG-based brain-computer interface (BCI) for motor imagery control, motivated by the 2024 Cybathlon. The system employs three mental and motor imagery classes to generate up to five control signals via a pipeline of data acquisition, preprocessing, classification using three diagonalized structured state-space sequence (S4D) layers, and a transfer function. Offline analysis reaches up to 84% classification accuracy; post-competition real-time validation with the original pilot and one additional participant achieves a 73% success rate in gameplay. The S4D model is compared to EEGEncoder and reference machine learning models, with claims of superior performance, and the work emphasizes human-centered design, low-cost hardware, and bridging laboratory results to daily-life accessibility for tetraplegic users.

Significance. If the empirical results hold, the work contributes to the field by demonstrating a practical, portable BCI framework that integrates real-time processing, user feedback via a mobile web application, and collaboration with the end user. The concrete offline and real-time performance numbers, together with the modular architecture, offer a template for accessible assistive technologies that could reduce the gap between controlled lab settings and everyday use.

major comments (2)
  1. [Results (post-Cybathlon validation)] The real-time validation reports a 73% success rate based on the original pilot plus one additional participant, yet provides no details on trial counts, data exclusion criteria, error bars, or cross-validation procedures. Given the well-documented high inter-subject variability in EEG motor imagery signals, this N=2 sample is insufficient to support the central claim that the pipeline delivers usable performance for broader accessibility and daily-life use without per-user calibration.
  2. [Model comparison section] The claim that the S4D model outperforms reference machine learning models and EEGEncoder is stated without accompanying quantitative tables, statistical tests, or effect sizes. This weakens the ability to evaluate whether the reported advantage is robust or merely descriptive.
minor comments (3)
  1. [Abstract] The abstract states 'up to 84%' offline accuracy without specifying the number of runs, sessions, or conditions under which this peak value was obtained.
  2. [Methods (classification and transfer function)] Clarify the precise hyperparameters of the S4D layers and the exact form of the transfer function mapping classification outputs to control dimensions.
  3. [Discussion] The attribution of competition performance drop-off primarily to stress lacks quantitative controls or physiological measures to distinguish it from other factors such as electrode placement or environmental noise.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major comment below, clarifying the scope of our claims and outlining planned revisions to improve transparency and rigor.

read point-by-point responses

  1. Referee: [Results (post-Cybathlon validation)] The real-time validation reports a 73% success rate based on the original pilot plus one additional participant, yet provides no details on trial counts, data exclusion criteria, error bars, or cross-validation procedures. Given the well-documented high inter-subject variability in EEG motor imagery signals, this N=2 sample is insufficient to support the central claim that the pipeline delivers usable performance for broader accessibility and daily-life use without per-user calibration.

    Authors: We acknowledge that the post-competition validation uses a small sample (N=2) and that the manuscript currently provides limited details on trial counts, exclusion criteria, error bars, or cross-validation. This validation was performed to demonstrate real-time feasibility with the original pilot and one additional participant after the Cybathlon, rather than to support broad statistical claims across populations. The manuscript emphasizes human-centered design with the specific tetraplegic user and does not assert that the system is ready for daily-life use without per-user calibration. In the revised version, we will expand the methods and results sections to include available details on the number of gameplay trials, success criteria, and any exclusion rules. We will also add an explicit discussion of inter-subject variability and the limitations of the small sample, thereby clarifying that the 73% rate serves as a proof-of-concept rather than evidence of general accessibility. revision: yes

  2. Referee: [Model comparison section] The claim that the S4D model outperforms reference machine learning models and EEGEncoder is stated without accompanying quantitative tables, statistical tests, or effect sizes. This weakens the ability to evaluate whether the reported advantage is robust or merely descriptive.

    Authors: The manuscript states that the S4D model outperforms the reference machine learning models while noting that EEGEncoder achieves higher performance at the expense of longer training time. We agree that the current text lacks quantitative tables, statistical tests, or effect sizes to support these comparisons. In the revised manuscript, we will insert a comparison table reporting classification accuracies, training times, and other metrics for the S4D model, EEGEncoder, and the reference models. Where the data allow, we will include statistical tests or effect sizes to make the performance differences more rigorous and evaluable. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper reports empirical results from an applied BCI engineering effort: an S4D-based classifier trained and tested on motor imagery EEG data from a tetraplegic pilot and one additional participant, yielding measured offline accuracy up to 84% and real-time success of 73%. No derivation chain, equations, or first-principles claims exist that could reduce to fitted inputs or self-citations by construction. Performance numbers are direct experimental outcomes, externally falsifiable by replication on new subjects or hardware, and compared against reference models without invoking author-specific uniqueness theorems or ansatzes. The analysis is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The pipeline relies on standard EEG preprocessing assumptions and the existence of distinguishable motor imagery signals in tetraplegic users; the S4D model introduces multiple hyperparameters whose values are not enumerated.

free parameters (2)
  • S4D layer hyperparameters
    Number of layers, state dimension, and training hyperparameters for the three diagonalized structured state-space sequence layers are chosen to achieve the reported accuracies.
  • Transfer function mapping
    Parameters that map the three-class output to up to five control signals are not specified and must be tuned per application.
axioms (2)
  • domain assumption Motor imagery signals remain detectable and classifiable in tetraplegic users after spinal cord injury
    The entire system presupposes that the pilot can reliably produce distinguishable EEG patterns for the three mental tasks.
  • standard math Standard EEG preprocessing steps preserve task-relevant information
    The pipeline assumes conventional filtering and artifact removal do not remove the signals needed for 84% accuracy.

pith-pipeline@v0.9.0 · 5683 in / 1555 out tokens · 55460 ms · 2026-05-17T04:20:03.259028+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

62 extracted references · 62 canonical work pages · 1 internal anchor

  1. [1]

    A brain–computer interface working definition

    Slutzky MW, Vansteensel MJ, Herff C, Gaunt RA. A brain–computer interface working definition. Nature Biomedical Engineering. 2025;9(6):792. https://doi.org/10.1038/s41551-025-01414-8

    work page doi:10.1038/s41551-025-01414-8 2025

  2. [2]

    A review of critical challenges in MI-BCI: From conventional to deep learning methods

    Khademi Z, Ebrahimi F, Kordy HM. A review of critical challenges in MI-BCI: From conventional to deep learning methods. Journal of Neuroscience Methods. 2023;383:109736. https://doi.org/10. 1016/j.jneumeth.2022.109736

    work page arXiv 2023

  3. [3]

    Covariate shift estimation based adaptive ensemble learning for handling non-stationarity in motor imagery related EEG-based brain- computer interface

    Raza H, Rathee D, Zhou SM, Cecotti H, Prasad G. Covariate shift estimation based adaptive ensemble learning for handling non-stationarity in motor imagery related EEG-based brain- computer interface. Neurocomputing. 2019;343:154–166. https://doi.org/10.1016/j.neucom.2018. 04.087

    work page doi:10.1016/j.neucom.2018 2019

  4. [4]

    The User-Centered Design as Novel Perspective for Evaluating the Usability of BCI-Controlled Applications

    Kübler A, Holz EM, Riccio A, Zickler C, Kaufmann T, Kleih SC, et al. The User-Centered Design as Novel Perspective for Evaluating the Usability of BCI-Controlled Applications. PLOS ONE. 2014;9(12):e112392. https://doi.org/10.1371/journal.pone.0112392

    work page doi:10.1371/journal.pone.0112392 2014

  5. [5]

    Why user-centered design is relevant for brain–computer interfacing and how it can be implemented in study protocols

    Kleih SC, Kübler A. Why user-centered design is relevant for brain–computer interfacing and how it can be implemented in study protocols. In: Berger K, Chen L, Müller KR, editors. Brain– Computer Interfaces Handbook. Boca Raton, FL: CRC Press, Taylor & Francis Group; 2018. p. 557–566

    work page 2018

  6. [6]

    How the CYBATHLON Competition Has Advanced Assistive Technologies

    Jaeger L, de Souza Baptista R, Basla C, Capsi-Morales P, Kim YK, Nakajima S, et al. How the CYBATHLON Competition Has Advanced Assistive Technologies. Annual Review of Control, Robotics, and Autonomous Systems. 2023;6:447–476. https://doi.org/10.1146/ annurev-control-071822-095355

    work page 2023

  7. [7]

    Review on motor imagery based BCI systems for upper-limb post-stroke neurorehabilitation: From designing to application

    Khan MA, Das R, Iversen HK, Puthusserypady S. Review on motor imagery based BCI systems for upper-limb post-stroke neurorehabilitation: From designing to application. Computers in Biology and Medicine. 2020;123:103843. https://doi.org/10.1016/j.compbiomed.2020.103843. 21

    work page doi:10.1016/j.compbiomed.2020.103843 2020

  8. [8]

    Mental imagery in the motor context

    Jeannerod M. Mental imagery in the motor context. Neuropsychologia. 1995;33(11):1419–1432. https://doi.org/10.1016/0028-3932(95)00073-C

    work page doi:10.1016/0028-3932(95)00073-c 1995

  9. [9]

    Motor activation prior to observation of a predictedmovement

    Kilner JM, Vargas C, Duval S, Blakemore SJ, Sirigu A. Motor activation prior to observation of a predictedmovement. NatureNeuroscience.2004;7(12):1299–1301. https://doi.org/10.1038/nn1355

    work page doi:10.1038/nn1355 2004

  10. [10]

    Motor imagery

    Lotze M, Halsband U. Motor imagery. Journal of Physiology - Paris. 2006;99(4–6):386–395. https: //doi.org/10.1016/j.jphysparis.2006.03.012

    work page doi:10.1016/j.jphysparis.2006.03.012 2006

  11. [11]

    Human-Centered Design and Development in Digital Health: Approaches, Challenges, and Emerging Trends

    Tzimourta KD. Human-Centered Design and Development in Digital Health: Approaches, Challenges, and Emerging Trends. Cureus. 2025;17(6):e85897. https://doi.org/10.7759/cureus. 85897

    work page doi:10.7759/cureus 2025

  12. [12]

    Brain-computer interfaces and human factors: the role of language and cultural differences – Still a missing gap? Frontiers in Human Neuroscience

    Herbert C. Brain-computer interfaces and human factors: the role of language and cultural differences – Still a missing gap? Frontiers in Human Neuroscience. 2024;18:1305445. https: //doi.org/10.3389/fnhum.2024.1305445

    work page doi:10.3389/fnhum.2024.1305445 2024

  13. [13]

    Review of public motor imagery and execution datasets in brain–computer interfaces

    Gwon D, Won K, Song M, Nam CS, Jun SC, Ahn M. Review of public motor imagery and execution datasets in brain–computer interfaces. Frontiers in Human Neuroscience. 2023;17:1134869. https: //doi.org/10.3389/fnhum.2023.1134869

    work page doi:10.3389/fnhum.2023.1134869 2023

  14. [14]

    Cybathlon experi- ences of the Graz BCI racing team MIRAGE91 in the brain-computer interface discipline

    Statthaler K, Schwarz A, Steyrl D, Kobler R, Höller MK, Brandstetter J, et al. Cybathlon experi- ences of the Graz BCI racing team MIRAGE91 in the brain-computer interface discipline. Journal of NeuroEngineering and Rehabilitation. 2017;14:129. https://doi.org/10.1186/s12984-017-0344-9

    work page doi:10.1186/s12984-017-0344-9 2017

  15. [15]

    Online EEG artifact removal for BCI applications by adaptive spatial filtering

    Guarnieri R, Marino M, Barban F, Ganzetti M, Mantini D. Online EEG artifact removal for BCI applications by adaptive spatial filtering. Journal of Neural Engineering. 2018;15(5):056005. https://doi.org/10.1088/1741-2552/aacfdf

    work page doi:10.1088/1741-2552/aacfdf 2018

  16. [16]

    A comparison of classification performance among the var- ious combinations of motor imagery tasks for brain–computer interface

    Kim HS, Chang MH, Lee HJ, Park KS. A comparison of classification performance among the var- ious combinations of motor imagery tasks for brain–computer interface. In: 2013 6th International IEEE/EMBS Conference on Neural Engineering (NER 2013); 2013. p. 435–438

    work page 2013

  17. [17]

    Event-related dynamics of cortical rhythms: Frequency-specific features and functional correlates

    Neuper C, Pfurtscheller G. Event-related dynamics of cortical rhythms: Frequency-specific features and functional correlates. International Journal of Psychophysiology. 2001;43(1):41–58. https: //doi.org/10.1016/S0167-8760(01)00178-7

    work page doi:10.1016/s0167-8760(01)00178-7 2001

  18. [18]

    Learning with self-supervision on EEG data

    Gramfort A, Banville H, Chehab O, Hyvärinen A, Engemann D. Learning with self-supervision on EEG data. In: 2021 9th IEEE International Winter Conference on Brain-Computer Interface 22 (BCI). United States: IEEE; 2021. p. 28–29

    work page 2021

  19. [19]

    Protocol for semi-automatic EEG preprocessing incorporating independent com- ponent analysis and principal component analysis

    Ouyang G, Li Y. Protocol for semi-automatic EEG preprocessing incorporating independent com- ponent analysis and principal component analysis. STAR Protocols. 2025;6(1):103682. https: //doi.org/10.1016/j.xpro.2025.103682

    work page doi:10.1016/j.xpro.2025.103682 2025

  20. [20]

    EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis

    Delorme A, Makeig S. EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. Journal of Neuroscience Methods. 2004;134(1):9–21. https://doi.org/10.1016/j.jneumeth.2003.10.009

    work page doi:10.1016/j.jneumeth.2003.10.009 2004

  21. [21]

    Digital filter design for electrophysiological data – a practical approach

    Widmann A, Schröger E, Maess B. Digital filter design for electrophysiological data – a practical approach. Journal of Neuroscience Methods. 2015;250:34–46. https://doi.org/10.1016/j.jneumeth. 2014.08.002

    work page doi:10.1016/j.jneumeth 2015

  22. [22]

    EMG and EOG artifacts in brain computer interface systems: A survey

    Fatourechi M, Bashashati A, Ward RK, Birch GE. EMG and EOG artifacts in brain computer interface systems: A survey. Clinical Neurophysiology. 2007;118(3):480–494. https://doi.org/10. 1016/j.clinph.2006.10.009

    work page 2007

  23. [23]

    Real-time modeling and 3D visualization of source dynamics and connectivity using wearable EEG

    Mullen T, Kothe C, Chi YM, Ojeda A, Kerth T, Makeig S, et al. Real-time modeling and 3D visualization of source dynamics and connectivity using wearable EEG. In: Proceedings of the 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC). IEEE; 2013. p. 2184–2187

    work page 2013

  24. [24]

    Using a common average reference to improve cortical neuron recordings from microelectrode arrays

    Ludwig KA, Miriani RM, Langhals NB, Joseph MD, Anderson DJ, Kipke DR. Using a common average reference to improve cortical neuron recordings from microelectrode arrays. Journal of Neurophysiology. 2009;101(3):1679–1689. https://doi.org/10.1152/jn.91095.2008

    work page doi:10.1152/jn.91095.2008 2009

  25. [25]

    Artifact removal techniques with signal reconstruction

    Kothe CAE, Jung TP, inventors; San Diego The Regents of the University of California, assignee. Artifact removal techniques with signal reconstruction. ; 2016. US Patent Application No. 14/895,440, Publication No. US20160113587A1

    work page 2016

  26. [26]

    Oscillatoryγ-Band (30–70 Hz) Activity Induced by a Visual Search Task in Humans

    Tallon-Baudry C, Bertrand O, Delpuech C, Pernier J. Oscillatoryγ-Band (30–70 Hz) Activity Induced by a Visual Search Task in Humans. Journal of Neuroscience. 1997;17(2):722–734. https: //doi.org/10.1523/JNEUROSCI.17-02-00722.1997

    work page doi:10.1523/jneurosci.17-02-00722.1997 1997

  27. [27]

    The quantitative extraction and topographic mapping of the abnormal components in the clinical EEG

    Koles ZJ. The quantitative extraction and topographic mapping of the abnormal components in the clinical EEG. Electroencephalography and Clinical Neurophysiology. 1991;79(6):440–447. https://doi.org/10.1016/0013-4694(91)90163-X. 23

    work page doi:10.1016/0013-4694(91)90163-x 1991

  28. [28]

    Designing optimal spatial filters for single-trial EEG classification in a movement task

    Müller-Gerking J, Pfurtscheller G, Flyvbjerg H. Designing optimal spatial filters for single-trial EEG classification in a movement task. Clinical Neurophysiology. 1999;110(5):787–798. https: //doi.org/10.1016/S1388-2457(98)00038-8

    work page doi:10.1016/s1388-2457(98)00038-8 1999

  29. [29]

    MEG and EEG data analysis with MNE -Python,

    Gramfort A, Luessi M, Larson E, Engemann DA, Strohmeier D, Brodbeck C, et al. MEG and EEG Data Analysis with MNE-Python. Frontiers in Neuroscience. 2013;7(267):1–13. https://doi. org/10.3389/fnins.2013.00267

    work page doi:10.3389/fnins.2013.00267 2013

  30. [30]

    MIMO Radar with Widely Separated Antennas,

    Blankertz B, Tomioka R, Lemm S, Kawanabe M, Müller KR. Optimizing spatial filters for robust EEG single-trial analysis. IEEE Signal Processing Magazine. 2008;25(1):41–56. https://doi.org/ 10.1109/MSP.2008.4408441

    work page doi:10.1109/msp.2008.4408441 2008

  31. [31]

    Pedregosa, G

    Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, et al. Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research. 2011;12:2825–2830. https://doi.org/ 10.5555/1953048.2078195

    work page doi:10.5555/1953048.2078195 2011

  32. [32]

    Cortes, V

    Cortes C, Vapnik V. Support-vector networks. Machine Learning. 1995;20:273–297. https://doi. org/10.1007/BF00994018

    work page doi:10.1007/bf00994018 1995

  33. [33]

    On the Parameterization and Initialization of Diagonal State Space Models

    Gu A, Goel K, Gupta A, Ré C. On the Parameterization and Initialization of Diagonal State Space Models. In: Koyejo S, Mohamed S, Agarwal A, Belgrave D, Cho K, Oh A, edi- tors. Advances in Neural Information Processing Systems. vol. 35. Curran Associates, Inc.; 2022.p.35971–35983. Availablefrom:https://proceedings.neurips.cc/paper_files/paper/2022/file/ e9...

    work page 2022

  34. [34]

    HiPPO: Recurrent Memory with Optimal Polynomial Projections

    Gu A, Dao T, Ermon S, Rudra A, Ré C. HiPPO: Recurrent Memory with Optimal Polynomial Projections. In: Larochelle H, Ranzato M, Hadsell R, Balcan MF, Lin H, edi- tors. Advances in Neural Information Processing Systems. vol. 33. Curran Associates, Inc

    work page

  35. [35]

    1474–1487

    p. 1474–1487. Available from: https://proceedings.neurips.cc/paper_files/paper/2020/file/ 102f0bb6efb3a6128a3c750dd16729be-Paper.pdf

    work page 2020

  36. [36]

    Limit results for distribu ted estimation of invariant subspaces in multiple networks inference and pca

    Gu A, Johnson I, Timalsina A, Rudra A, Ré C. How to Train Your HiPPO: State Space Models with Generalized Orthogonal Basis Projections. arXiv. 2022;https://doi.org/10.48550/arXiv.2206. 12037

    work page doi:10.48550/arxiv.2206 2022

  37. [37]

    Efficiently Modeling Long Sequences with Structured State Spaces

    Gu A, Goel K, Ré C. Efficiently Modeling Long Sequences with Structured State Spaces. arXiv. 2022;https://doi.org/https://doi.org/10.48550/arXiv.2111.00396. 24

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2111.00396 2022

  38. [38]

    Dropout as a Bayesian approximation: Representing model uncertainty in deep learning

    Gal Y, Ghahramani Z. Dropout as a Bayesian approximation: Representing model uncertainty in deep learning. In: Proceedings of the 33rd International Conference on Machine Learning (ICML). JMLR; 2016. p. 1050–1059

    work page 2016

  39. [39]

    EEGEncoder: Advancing BCI with transformer-based motor imagery classification

    Liao W, Wang W. EEGEncoder: Advancing BCI with transformer-based motor imagery classification. Scientific Reports. 2025;https://doi.org/10.1038/s41598-025-06364-4

    work page doi:10.1038/s41598-025-06364-4 2025

  40. [40]

    2018 , month =

    Lawhern VJ, Solon AJ, Waytowich NR, Gordon SM, Hung CP, Lance BJ. EEGNet: A com- pact convolutional neural network for EEG-based brain-computer interfaces. Journal of Neural Engineering. 2018;15(5):056013. https://doi.org/10.1088/1741-2552/aace8c

    work page doi:10.1088/1741-2552/aace8c 2018

  41. [41]

    CTNet: a convolutional transformer network for EEG-based motor imagery classification

    Zhao W, Jiang X, Zhang B, Xiao S, Weng S. CTNet: a convolutional transformer network for EEG-based motor imagery classification. Scientific Reports. 2024;14:20237. https://doi.org/10. 1038/s41598-024-71118-7

    work page 2024

  42. [42]

    Imaging Neuroscience3, IMAG.a.136 (2025)

    Kothe C, Shirazi SY, Stenner T, Medine D, Boulay C, Grivich MI, et al. The lab streaming layer for synchronized multimodal recording. Imaging Neuroscience. 2025 09;3:IMAG.a.136. https: //doi.org/10.1162/IMAG.a.136

    work page doi:10.1162/imag.a.136 2025

  43. [43]

    Jeunet C. Understanding & Improving Mental-Imagery Based Brain-Computer Interface (MiBCI) User-Training: Towards a New Generation of Reliable, Efficient & Accessible Brain-Computer Interfaces. PhD thesis. Psychology. Université de Bordeaux. Bordeaux, France; 2016. NNT: 2016BORD0221. tel-01417606

    work page 2016

  44. [44]

    Flaws in current human training protocols for spontaneous Brain- Computer Interfaces: lessons learned from instructional design

    Lotte F, Larrue F, Mühl C. Flaws in current human training protocols for spontaneous Brain- Computer Interfaces: lessons learned from instructional design. Frontiers in Human Neuroscience. 2013;7:568. https://doi.org/10.3389/fnhum.2013.00568

    work page doi:10.3389/fnhum.2013.00568 2013

  45. [45]

    Is It Significant? Guidelines forReportingBCIPerformance

    Billinger M, Daly I, Kaiser V, Jin J, Allison BZ, Müller-Putz GR, et al. Is It Significant? Guidelines forReportingBCIPerformance. In:AllisonBZ,DunneS,LeebR,DelRMillánJ,NijholtA,editors. Towards Practical Brain-Computer Interfaces: Bridging the Gap from Research to Real-World Applications. Berlin, Heidelberg: Springer Berlin Heidelberg; 2013. p. 333–354

    work page 2013

  46. [46]

    Accurate estimation of information transfer rate based on symbol occur- rence probability in brain-computer interfaces

    Sadeghi S, Maleki A. Accurate estimation of information transfer rate based on symbol occur- rence probability in brain-computer interfaces. Biomedical Signal Processing and Control. 2019;54:101607. https://doi.org/10.1016/j.bspc.2019.101607

    work page doi:10.1016/j.bspc.2019.101607 2019

  47. [47]

    How Many People Could Use an SSVEP BCI? Frontiers in Neuroscience

    Guger C, Allison BZ, Grosswindhager B, Prückl R, Hintermüller C, Kapeller C, et al. How Many People Could Use an SSVEP BCI? Frontiers in Neuroscience. 2012;6:169. https://doi.org/10.3389/ 25 fnins.2012.00169

    work page arXiv 2012

  48. [48]

    Review of Brain–Computer Interface Based on Steady-State Visual Evoked Potential (SSVEP)

    Liu S. Review of Brain–Computer Interface Based on Steady-State Visual Evoked Potential (SSVEP). Brain–Stimulation Advances. 2022;2(3):100035. https://doi.org/10.26599/BSA.2022. 9050022

    work page doi:10.26599/bsa.2022 2022

  49. [49]

    Riemannian geometry-based transfer learning for reducing training time in c-VEP BCIs

    Ying J, Wei Q, Zhou X. Riemannian geometry-based transfer learning for reducing training time in c-VEP BCIs. Scientific Reports. 2022;12:9818. https://doi.org/10.1038/s41598-022-14026-y

    work page doi:10.1038/s41598-022-14026-y 2022

  50. [50]

    Reducing execution time for real-time motor imagery based BCI systems

    Selim S, Tantawi M, Shedeed H, Badr A. Reducing execution time for real-time motor imagery based BCI systems. In: Proceedings of the International Conference on Advanced Intelligent Systems and Informatics 2016; 2016. p. 555–565

    work page 2016

  51. [51]

    Enhanced Low-Latency Detection of Motor Intention From EEG for Closed-Loop Brain-Computer Interface Applications

    Xu R, Jiang N, Lin C, Mrachacz-Kersting N, Dremstrup K, Farina D. Enhanced Low-Latency Detection of Motor Intention From EEG for Closed-Loop Brain-Computer Interface Applications. IEEE Transactions on Bio-Medical Engineering. 2014 02;61:288–296. https://doi.org/10.1109/ TBME.2013.2294203

    work page arXiv 2014

  52. [52]

    Factors of Influence on the Per- formance of a Short-Latency Non-Invasive Brain Switch: Evidence in Healthy Individuals and Implication for Motor Function Rehabilitation

    Xu R, Jiang N, Mrachacz-Kersting N, Dremstrup K, Farina D. Factors of Influence on the Per- formance of a Short-Latency Non-Invasive Brain Switch: Evidence in Healthy Individuals and Implication for Motor Function Rehabilitation. Frontiers in Neuroscience. 2016;Volume 9 - 2015. https://doi.org/10.3389/fnins.2015.00527

    work page doi:10.3389/fnins.2015.00527 2016

  53. [53]

    A frequency-domain approach with learnable filters for image classification

    Stuchi JA, Canto NG, de Faissol Attux RR, Boccato L. A frequency-domain approach with learnable filters for image classification. Applied Soft Computing. 2024;155:111443. https://doi. org/10.1016/j.asoc.2024.111443

    work page doi:10.1016/j.asoc.2024.111443 2024

  54. [54]

    Common Techniques in Data Classification

    AggarwalCC. AnIntroductiontoDataClassification. In:AggarwalCC,editor.DataClassification: Algorithms and Applications. Boca Raton, FL: CRC Press; 2015. p. 4–15. Section 1.2 "Common Techniques in Data Classification"

    work page 2015

  55. [55]

    Multi-scale convolutional transformer network for motor imagery brain-computer interface

    Zhao W, Zhang B, Zhou H, Wei D, Huang C, Lan Q. Multi-scale convolutional transformer network for motor imagery brain-computer interface. Scientific Reports. 2025;15:12935. https: //doi.org/10.1038/s41598-025-96611-5

    work page doi:10.1038/s41598-025-96611-5 2025

  56. [56]

    A novel deep learning approach for classification of EEG motor imagery signals

    Tabar YR, Halici F. A novel deep learning approach for classification of EEG motor imagery signals. Journal of Neural Engineering. 2016;14(1):016003. https://doi.org/10.1088/1741-2560/14/ 1/016003. 26

    work page doi:10.1088/1741-2560/14/ 2016

  57. [57]

    Classification of motor imagery EEG using deep learning increases performance in inefficient BCI users

    Tibrewal N, Leeuwis N, Alimardani M. Classification of motor imagery EEG using deep learning increases performance in inefficient BCI users. PLOS ONE. 2022 07;17(7):1–18. https://doi.org/ 10.1371/journal.pone.0268880

    work page doi:10.1371/journal.pone.0268880 2022

  58. [58]

    Decoding EEG rhythms offline and online during motor imagery for standing and sitting based on a brain- computer interface

    Triana-Guzman N, Orjuela-Cañon AD, Jutinico AL, Mendoza-Montoya O, Antelis JM. Decoding EEG rhythms offline and online during motor imagery for standing and sitting based on a brain- computer interface. Frontiers in Neuroinformatics. 2022;16. https://doi.org/10.3389/fninf.2022. 961089

    work page doi:10.3389/fninf.2022 2022

  59. [59]

    A Review of Online Classification Performance in Motor Imagery-Based Brain–Computer Interfaces for Stroke Neurorehabilitation

    Vavoulis A, Figueiredo P, Vourvopoulos A. A Review of Online Classification Performance in Motor Imagery-Based Brain–Computer Interfaces for Stroke Neurorehabilitation. Signals. 2023;4(1):73–

    work page 2023

  60. [60]

    https://doi.org/10.3390/signals4010004

    work page doi:10.3390/signals4010004

  61. [61]

    Real-time multiclass motor imagery brain-computer interface by modified common spatial patterns and adaptive neuro-fuzzy classifier

    Jafarifarmand A, Badamchizadeh MA. Real-time multiclass motor imagery brain-computer interface by modified common spatial patterns and adaptive neuro-fuzzy classifier. Biomedical Sig- nal Processing and Control. 2020;57:101749. https://doi.org/https://doi.org/10.1016/j.bspc.2019. 101749

    work page doi:10.1016/j.bspc.2019 2020

  62. [62]

    neuroTUM e.V

    neuroTUM e V.: neuroTUM-BCI: Cybathlon Dataset. neuroTUM e.V. Available from: https: //doi.org/10.5281/zenodo.18087806. 27

    work page doi:10.5281/zenodo.18087806