Pith. sign in

REVIEW 3 major objections 6 minor 43 references

TFOC-Net: A Short-time Fourier Transform-based Deep Learning Approach for Enhancing Cross-Subject Motor Imagery Classification

T0 review · 3 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read TFOC-Net claims benchmark-level cross-subject motor-imagery accuracy by classifying raw STFT spectra directly, raising STFT overlap to 50%, and balancing training batches across subjects.

desk verdict A useful but overclaimed empirical study: the method is simple and the four-dataset evaluation is valuable, yet the headline accuracies are likely inflated because key hyperparameters were selected on the LOSO test folds. read the letter →

arxiv 2507.02510 v1 pith:L6VVC3S4 submitted 2025-07-03 cs.LG cs.HCcs.NE

Ahmed G. Habashi , Ahmed M. Azab , Seif Eldawlatly , Gamal M. Aly This is my paper · ORCID
classification cs.LGcs.HCcs.NE
keywords EEGmotorimagerycross-subjectclassificationshort-timeFouriertransformconvolutionalneuralnetworkcalibration-freeBCIbalancedbatchingVGG
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that a convolutional neural network fed directly with the time-frequency representation of EEG signals can classify motor imagery in people never seen during training, reaching mean accuracies of 67.60% on BCI Competition IV Dataset 1, 65.96% on Dataset 2A, and 80.22% on Dataset 2B. The authors attribute the gains to three pipeline choices: using the raw short-time Fourier transform (STFT) magnitude instead of converting it to a small image, increasing the STFT window overlap to 50%, and forcing every training batch to contain trials from all training subjects. If these results hold, they strengthen the case for calibration-free brain-computer interfaces that work on a new user immediately. The paper also reports that shorter motor-imagery windows, especially the first three seconds or the middle two seconds, often approach the accuracy of the full four-second window, which could make BCIs faster and less tiring to use.

What carries the argument

The central objects are the STFT magnitude of each EEG trial, treated as a single-channel image whose height is the frequency bins and whose width is the time points concatenated across the three channels, and a VGG-based CNN with stacked 3x3 convolutions, dropout, and MaxNorm weight constraints. Three choices carry the argument: direct classification of the raw STFT rather than a down-sampled 32x32 image; increasing the STFT overlap from minimal to 50%, which densifies the time-frequency grid and acts as a regularizer; and balanced batching, defined as each training batch containing data from all training subjects. The mechanism is that these choices preserve temporal detail, smooth the input representation, and force the network to see inter-subject variation at every optimization step, improving generalization to held-out subjects.

What would settle it

Run the identical TFOC-Net pipeline with nested leave-one-out evaluation, selecting the STFT overlap, batching mode, and early-stopping epoch on an inner validation fold exclusive of the test subject; if the mean accuracies fall materially below 80.22% on IV-2B, 67.60% on IV-1, and 65.96% on IV-2A, the reported benchmark depends on test-set information. An independent reproduction that fixes overlap at 50% and balanced batching a priori can check whether per-subject accuracies in Figures 3–5 are replicated.

Watch

Extended reading notes

Core claim

The central discovery is that a VGG-style CNN trained on the magnitude of the STFT of three EEG channels (C3, Cz, C4) with 50% window overlap outperforms previously published cross-subject motor-imagery classifiers on two of the three benchmark datasets and does not significantly lose on the third. On BCI Competition IV Dataset 2B, mean accuracy rises from 70.55% with the authors' previous WGAN-based pipeline to 80.22%; on IV-1 it rises from 63.7% to 67.60%; on IV-2A it rises modestly from 65.27% to 65.96%, where one published method (67.42%) is not significantly beaten. The approach is validated on a newly recorded five-subject dataset, achieving 63.54% mean accuracy, with each modification—direct STFT classification, increased overlap, and balanced batching—contributing incrementally. Evaluation is leave-one-subject-out, and the authors state that no test-subject data were used during training.

Load-bearing premise

The reported accuracies are trustworthy only if the preprocessing and training choices—particularly the 50% STFT overlap, the balanced-batching scheme, and early stopping—were fixed without reference to the labels of the held-out subjects; the paper does not describe a separate validation set used to select them.

Editorial extensions

If this is right

  • A calibration-free motor-imagery BCI becomes practical at usable accuracies: a new user can be classified immediately without a recording session for personal calibration.
  • Because only three channels (C3, Cz, C4) are used, the approach can run on low-density, low-cost EEG headsets rather than full-cap arrays.
  • The time-window analysis indicates that trials can be shortened to about three seconds, or to the middle two seconds, with modest accuracy loss, enabling more responsive BCIs.
  • Increasing STFT overlap and balancing batches across subjects are simple, model-agnostic changes that improve cross-subject generalization and could transfer to other EEG classification pipelines.
  • The public release of the code and the newly recorded five-subject EEG dataset gives the community a directly comparable resource for future cross-subject work.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the absence of test-subject information in hyperparameter selection is confirmed through a nested validation design, the results suggest that simple preprocessing and batching choices can close much of the gap to more complex transfer-learning and few-shot methods, so future comparisons should control for this factor.
  • The segment analysis implies an accuracy-versus-trial-length trade-off that could be used to optimize BCI communication rate; a natural next experiment is to combine the best single segment with the same pipeline and measure throughput.
  • Because the method uses only three motor-cortex channels, it is a candidate for consumer-grade wearable EEG; a testable extension is to evaluate the same pipeline on dry-electrode recordings to quantify the penalty of noisier signals.
  • Balanced batching's consistent benefit points to inter-subject variance as a key optimization obstacle, suggesting that other variance-reduction techniques, such as stratified sampling or contrastive objectives, may yield similar or larger gains.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper proposes TFOC-Net, a VGG-based CNN that classifies STFT-transformed EEG directly (rather than converting STFT outputs to images), using increased STFT overlap and a balanced batching strategy for calibration-free cross-subject motor imagery. It reports mean leave-one-subject-out accuracies of 67.60% on BCI Competition IV Dataset 1, 65.96% on Dataset 2A, and 80.22% on Dataset 2B, claims state-of-the-art performance, and additionally contributes a newly recorded EEG dataset and a time-segment analysis of the MI period.

Significance. The paper is a clearly written engineering study with a plausible architecture and an unusually broad evaluation across three public benchmarks plus a new dataset. The explicit calibration-free evaluation protocol and the systematic time-segment analysis are useful contributions. If the reported accuracies are unbiased, TFOC-Net would be a competitive calibration-free baseline. However, the central evaluation protocol currently permits hyperparameter choices to be made on the same test folds used for reporting, and at least one key statistical statement is incorrect. These issues must be resolved before the performance claims can be accepted as state-of-the-art.

major comments (3)
  1. [§2.5, §2.6, §3.1–3.2, Figures 3–5] The final pipeline—direct STFT classification, optimized (50%) overlap, and balanced batching—is selected by comparing mean LOSO accuracies across variants on the three benchmark datasets (Figures 3–5). No validation split or nested selection procedure is described: Section 2.6 states only that test-subject data were excluded from training, and Section 2.5 says the best checkpoint is saved by 'validation accuracy' without defining a validation set. This means the reported headline numbers (67.60%, 65.96%, and 80.22%) may be maxima over variants evaluated on the test folds, which is an optimistic bias, especially with only five or nine subjects. Please either (a) describe a validation-based selection protocol performed inside the training folds, (b) show that the overlap and batching choices were fixed a priori before any test-fold accuracy was inspected, or (c) report selection-bias-corrected accuracies (e.g., nested LOSO or a held-out test split). Without this, the per-modification gains and the state-of-the-art comparison are not reliable.
  2. [§3.6] The text describes the IV-1 improvement over WGAN-CNN as 'reaching a significance enhancement (p = 0.062)'. Since p = 0.062 > 0.05, this is not statistically significant at the conventional 5% level, and the wording must be corrected. Moreover, with n = 5 subjects, the Wilcoxon signed-rank test has very limited power; please report the exact test version, effect sizes, and per-subject differences, and temper the conclusion that this improvement 'reinforces the robustness and generalizability' of the method on IV-1.
  3. [§3.6, Table 5] The blanket claim of outperforming state-of-the-art is not supported for IV-2A: TFOC-Net's mean accuracy (65.96%) is below Liang et al. (67.42%), and the argument that the difference is not significant and that TFOC-Net wins in 5 of 9 subjects does not establish superiority. In addition, the only IV-1 comparator is the authors' own WGAN-CNN, so the 'state-of-the-art' label for IV-1 rests on a single self-comparison. Please qualify the state-of-the-art claims per dataset, include independent baselines for IV-1, and clarify how the p-values in Table 5 were computed when the comparison methods do not report per-subject accuracies.
minor comments (6)
  1. [§2.4, §4] The STFT window size and overlap parameters are not fully specified in the methodology; the overlap is first described as 'minimal (1 sample)' versus 50% only in Section 4. Please provide exact preprocessing parameters (window, hop, FFT points, channel handling) so that the experiments are reproducible.
  2. [§2.6, §3.1] Section 2.6 says a Wilcoxon rank-sum test will be used, while Section 3.1 reports Wilcoxon signed-rank tests for paired LOSO comparisons; please clarify which test was applied and why.
  3. [§3.1–3.6] Figures 3–5 are tables rather than figures, the section numbering jumps from 3.2 to 3.4 (Section 3.3 is missing), and the standard-deviation formatting is inconsistent across Tables 2–5; please fix the layout and numbering.
  4. [Abstract, §2.3.4, §5] The abstract and introduction promise public release of code, trained models, and a new dataset, but no repository URL or data availability statement appears in the manuscript; please add these or explicitly state the current availability status.
  5. [§3.5] The recorded-dataset results in Figure 6 are reported without standard deviations or per-condition significance tests despite large trial-count differences across subjects (40–150 per session); please report these details.
  6. [§4] The claim that this is 'the first and only study' validated on four datasets is not substantiated by the cited literature and should be rephrased.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: TFOC-Net's evaluations are empirical LOSO accuracies on independent public benchmarks, and no claimed prediction is defined in terms of its inputs.

full rationale

TFOC-Net's derivation chain is empirical rather than algebraic: direct STFT classification, increased overlap, and balanced batching are each tested by comparing LOSO accuracies on public benchmark datasets, with no equation that defines the reported accuracy in terms of a fitted parameter. The model architecture is adapted from the authors' prior VGG/WGAN work [18], but that prior work is used as a baseline and design starting point, not as an unverified theorem that forces the current result. The only same-author entry in the comparison tables is WGAN-CNN [18], and it is the sole IV-1 comparator in Table 5; this is a baseline-selection weakness for the state-of-the-art claim, but the reported 67.60% is an independent measurement on held-out subjects of a public dataset, so it is not circular. The possible tuning of STFT overlap and balanced batching on the same LOSO test folds (Sections 3.1-3.2) and the undefined validation set for early stopping (Section 2.5) are optimistic-bias or reproducibility concerns, not circularity. The promise of a code repository in Section 1 without a working link is a reproducibility gap, not a circular step. The manuscript also concedes that Liang et al. achieves a higher IV-2A mean (67.42% vs 65.96%), which further shows the SOTA claim is a comparative judgment rather than a tautology. Thus no circular step meets the required 'equivalent by construction' standard.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

There is no mathematical derivation and no invented physical entity. The central claim is empirical, and it rests on several manually chosen signal-processing and network hyperparameters; the most consequential is the STFT overlap, which appears to be tuned on the test folds.

free parameters (4)
  • STFT overlap ratio = 50% (vs. minimum 1-sample overlap in baseline)
    Selected after observing LOSO test accuracy gains in Section 3.2; no held-out validation or nested model selection is described.
  • STFT window length = 256 samples for IV-2A, IV-2B, and recorded dataset; 128 samples for IV-1
    Taken from prior literature rather than optimized here, but it defines the input dimensions and time-frequency resolution.
  • Butterworth bandpass = 6th order, 8-30 Hz
    A hand-chosen frequency band applied before STFT; standard for mu/beta rhythms but not optimized or justified from this data.
  • CNN training configuration = 200 epochs, batch size = (number of training subjects) x 4, dropout 0.2/0.4, filters 32/64/128
    Hand-chosen capacity and regularization settings; no search or sensitivity analysis is reported.
assumptions (4)
  • domain assumption Relevant MI information is contained in C3, Cz, C4 and the 8-30 Hz band.
    The pipeline discards all other channels and frequencies before classification (Sections 2.2 and 2.4).
  • domain assumption STFT magnitude, with phase discarded, is sufficient for cross-subject MI classification.
    Section 2.4 uses the STFT magnitude as the 2D input; phase information is never modeled.
  • ad hoc to paper Leave-one-subject-out folds isolate test subjects, and hyperparameters were selected without test labels.
    Section 2.6 asserts calibration-free evaluation, but Sections 3.1-3.2 show overlap and batching choices were evaluated on the same LOSO test accuracies, so this premise is not auditable.
  • domain assumption The three benchmark datasets can be treated as comparable after selecting the same classes and channels.
    IV-1 uses 100 Hz sampling and three monopolar channels; IV-2A uses 22 channels but only C3/Cz/C4 are kept; IV-2B uses bipolar channels; the paper assumes a single architecture works across these formats.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TFOC-Net: A Short-time Fourier Transform-based Deep Learning Approach for Enhancing Cross-Subject Motor Imagery Classification." pith.science (2026). https://pith.science/paper/L6VVC3S4

@misc{pith2026250702510,
  author       = {Pith},
  title        = {Pith review of: TFOC-Net: A Short-time Fourier Transform-based Deep Learning Approach for Enhancing Cross-Subject Motor Imagery Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L6VVC3S4}},
  note         = {Machine review of arXiv:2507.02510}
}
read the original abstract

Cross-subject motor imagery (CS-MI) classification in brain-computer interfaces (BCIs) is a challenging task due to the significant variability in Electroencephalography (EEG) patterns across different individuals. This variability often results in lower classification accuracy compared to subject-specific models, presenting a major barrier to developing calibration-free BCIs suitable for real-world applications. In this paper, we introduce a novel approach that significantly enhances cross-subject MI classification performance through optimized preprocessing and deep learning techniques. Our approach involves direct classification of Short-Time Fourier Transform (STFT)-transformed EEG data, optimized STFT parameters, and a balanced batching strategy during training of a Convolutional Neural Network (CNN). This approach is uniquely validated across four different datasets, including three widely-used benchmark datasets leading to substantial improvements in cross-subject classification, achieving 67.60% on the BCI Competition IV Dataset 1 (IV-1), 65.96% on Dataset 2A (IV-2A), and 80.22% on Dataset 2B (IV-2B), outperforming state-of-the-art techniques. Additionally, we systematically investigate the classification performance using MI windows ranging from the full 4-second window to 1-second windows. These results establish a new benchmark for generalizable, calibration-free MI classification in addition to contributing a robust open-access dataset to advance research in this domain.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 26 canonical work pages

  1. [1]

    Brain computer interfaces, a review,

    L. F. Nicolas -Alonso and J. Gomez -Gil, “Brain computer interfaces, a review,” Feb

    work page

  2. [2]

    Brain – Computer Interface Spellers: A Review,

    A. Rezeika, M. Benda, P. Stawicki, F. Gembler, A. Saboor, and I. Volosyak, “Brain – Computer Interface Spellers: A Review,” Brain Sciences 2018, Vol. 8, Page 57 , vol. 8, no. 4, p. 57, Mar. 2018, doi: 10.3390/BRAINSCI8040057

    doi:10.3390/brainsci8040057 2018

  3. [3]

    Combining generative adversarial networks and multi-output CNN for motor imagery classification,

    J. Xie, S. Chen, Y. Zhang, D. Gao, and T. Liu, “Combining generative adversarial networks and multi-output CNN for motor imagery classification,” J Neural Eng, vol. 18, no. 4, Aug. 2021, doi: 10.1088/1741-2552/abecc5

    work page doi:10.1088/1741-2552/abecc5 2021

  4. [4]

    Autoreject: Automated artifact rejection for MEG and EEG data,

    M. Jas, D. A. Engemann, Y. Bekhti, F. Raimondo, and A. Gramfort, “Autoreject: Automated artifact rejection for MEG and EEG data,” Neuroimage, vol. 159, pp. 417 –429, Oct. 2017, doi: 10.1016/J.NEUROIMAGE.2017.06.030

    doi:10.1016/j.neuroimage.2017.06.030 2017

  5. [5]

    Generative adversarial networks in EEG analysis: an overview,

    A. G. Habashi, A. M. Azab, S. Eldawlatly, and G. M. Aly, “Generative adversarial networks in EEG analysis: an overview,” J Neuroeng Rehabil, vol. 20, no. 1, pp. 1 –24, Dec. 2023, doi: 10.1186/S12984-023-01169-W/TABLES/5

    work page doi:10.1186/s12984-023-01169-w/tables/5 2023

  6. [6]

    An Ensemble CNN for Subject-Independent Classification of Motor Imagery-based EEG,

    I. Dolzhikova, B. Abibullaev, R. Sameni, and A. Zollanvari, “An Ensemble CNN for Subject-Independent Classification of Motor Imagery-based EEG,” Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS , pp. 319–324, 2021, doi: 10.1109/EMBC46164.2021.9630419

    arXiv 2021

  7. [7]

    Separated channel convolutional neural network to realize the training free motor imagery BCI systems,

    X. Zhu, P. Li, C. Li, D. Yao, R. Zhang, and P. Xu, “Separated channel convolutional neural network to realize the training free motor imagery BCI systems,” Biomed Signal Process Control, vol. 49, pp. 396–403, Mar. 2019, doi: 10.1016/J.BSPC.2018.12.027

    work page doi:10.1016/j.bspc.2018.12.027 2019

  8. [8]

    Cross-Subject EEG Signal Classification with Deep Neural Networks Applied to Motor Imagery,

    M. Riyad, M. Khalil, and A. Adib, “Cross-Subject EEG Signal Classification with Deep Neural Networks Applied to Motor Imagery,” Lecture Notes in Computer Science (including 29 subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) , vol. 11557 LNCS, pp. 124–139, 2019, doi: 10.1007/978-3-030-22885-9_12/FIGURES/7

    work page doi:10.1007/978-3-030-22885-9_12/figures/7 2019

Show all 43 references
  1. [9]

    Fusion Convolutional Neural Network for Cross-Subject EEG Motor Imagery Classification,

    K. Roots, Y. Muhammad, and N. Muhammad, “Fusion Convolutional Neural Network for Cross-Subject EEG Motor Imagery Classification,” Computers 2020, Vol. 9, Page 72, vol. 9, no. 3, p. 72, Sep. 2020, doi: 10.3390/COMPUTERS9030072

    2020 doi

  2. [10]

    Deep Learning Based Inter - subject Continuous Decoding of Motor Imagery for Practical Brain -Computer Interfaces,

    S. Roy, A. Chowdhury, K. McCreadie, and G. Prasad, “Deep Learning Based Inter - subject Continuous Decoding of Motor Imagery for Practical Brain -Computer Interfaces,” Front Neurosci, vol. 14, p. 563817, Sep. 2020, doi: 10.3389/FNINS.2020.00918/BIBTEX

    2020

  3. [11]

    CNN -based Approaches For Cross -Subject Classification in Motor Imagery: From the state -of-the-art to dynamicnet,

    A. Zancanaro, G. Cisotto, J. R. Paulo, G. Pires, and U. J. Nunes, “CNN -based Approaches For Cross -Subject Classification in Motor Imagery: From the state -of-the-art to dynamicnet,” 2021 IEEE Conference on Computational Intelligence in Bioinformatics and Computational Biolog...

    2021

  4. [12]

    Dual Attention Relation Network With Fine-Tuning for Few -Shot EEG Motor Imagery Classification,

    S. An, S. Kim, P. Chikontwe, and S. H. Park, “Dual Attention Relation Network With Fine-Tuning for Few -Shot EEG Motor Imagery Classification,” IEEE Trans Neural Netw Learn Syst, pp. 1–15, Jun. 2023, doi: 10.1109/TNNLS.2023.3287181

    2023

  5. [13]

    Excellent fine-tuning: From specific-subject classification to cross-task classification for motor imagery,

    X. Jia, Y. Song, and L. Xie, “Excellent fine-tuning: From specific-subject classification to cross-task classification for motor imagery,” Biomed Signal Process Control , vol. 79, p. 104051, Jan. 2023, doi: 10.1016/J.BSPC.2022.104051

    2023

  6. [14]

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

    Y. R. Tabar and U. Halici, “A novel deep learning approach for classification of EEG motor imagery signals,” J Neural Eng, vol. 14, no. 1, p. 016003, Nov. 2016, doi: 10.1088/1741- 2560/14/1/016003

    2016 doi

  7. [15]

    Data augmentation for motor imagery signal classification based on a hybrid neural network,

    K. Zhang et al., “Data augmentation for motor imagery signal classification based on a hybrid neural network,” Sensors (Switzerland) , vol. 20, no. 16, pp. 1 –20, Aug. 2020, doi: 10.3390/s20164485. 30

    2020 doi

  8. [16]

    4 -class mi-eeg signal generation and recognition with cvae-gan,

    J. Yang, H. Yu, T. Shen, Y. Song, and Z. Chen, “4 -class mi-eeg signal generation and recognition with cvae-gan,” Applied Sciences (Switzerland), vol. 11, no. 4, pp. 1–14, Feb. 2021, doi: 10.3390/app11041798

    2021 doi

  9. [17]

    Motor Imagery Classification Enhancement using Generative Adversarial Networks for EEG Spectrum Image Generation,

    A. G. Habashi, A. M. Azab, S. Eldawlatly, and G. M. Aly, “Motor Imagery Classification Enhancement using Generative Adversarial Networks for EEG Spectrum Image Generation,” 2023 IEEE 36th International Symposium on Computer -Based Medical Systems (CBMS), pp. 354–359, Jun. 2023...

    2023

  10. [18]

    Toward calibration -free motor imagery brain –computer interfaces: a VGG -based convolutional neural network and WGAN approach,

    A. G. Habashi, A. M. Azab, S. Eldawlatly, and G. M. Aly, “Toward calibration -free motor imagery brain –computer interfaces: a VGG -based convolutional neural network and WGAN approach,” J Neural Eng , vol. 21, no. 4, p. 046032, Jul. 2024, doi: 10.1088/1741 - 2552/AD6598

    2024 doi

  11. [19]

    CNN based classification of motor imaginary using variational mode decomposed EEG -spectrum image,

    K. Keerthi Krishnan and K. P. Soman, “CNN based classification of motor imaginary using variational mode decomposed EEG -spectrum image,” Biomed Eng Lett, vol. 11, no. 3, pp. 235–247, Aug. 2021, doi: 10.1007/S13534-021-00190-Z/TABLES/8

    2021 doi

  12. [20]

    EEG-Based Classification of Lower Limb Motor Imagery with STFT and CNN,

    B. Lu, S. Ge, and H. Wang, “EEG-Based Classification of Lower Limb Motor Imagery with STFT and CNN,” Communications in Computer and Information Science , vol. 1517 CCIS, pp. 397–404, 2021, doi: 10.1007/978-3-030-92310-5_46/FIGURES/5

    2021 doi

  13. [21]

    Short -time Fourier transform and embedding method for recurrence quantification analysis of EEG time series,

    Ł. Furman, W. Duch, L. Minati, and K. Tołpa, “Short -time Fourier transform and embedding method for recurrence quantification analysis of EEG time series,” The European Physical Journal Special Topics 2022 232:1 , vol. 232, no. 1, pp. 135 –149, Oct. 2022, doi: 10.1140/EPJS/S1...

    2022 doi

  14. [22]

    Signal Estimation from Modified Short -Time Fourier Transform,

    D. W. Griffin and J. S. Lim, “Signal Estimation from Modified Short -Time Fourier Transform,” IEEE Trans Acoust , vol. 32, no. 2, pp. 236 –243, 1984, doi: 10.1109/TASSP.1984.1164317. 31

    1984

  15. [23]

    BCI Competition IV

    “BCI Competition IV.” Accessed: Mar. 18, 2023. [Online]. Available: https://www.bbci.de/competition/iv/#dataset1

    2023

  16. [24]

    Wavelet Transform Time-Frequency Image and Convolutional Network- Based Motor Imagery EEG Classification,

    B. Xu et al., “Wavelet Transform Time-Frequency Image and Convolutional Network- Based Motor Imagery EEG Classification,” IEEE Access, vol. 7, pp. 6084 –6093, 2019, doi: 10.1109/ACCESS.2018.2889093

    2019

  17. [25]

    BCI Competition 2008-Graz data set A Experimental paradigm,

    C. Brunner, R. Leeb, G. R. Müller -Putz, A. Schlögl, and G. Pfurtscheller, “BCI Competition 2008-Graz data set A Experimental paradigm,” Institute for Knowledge Discovery (Laboratory of Brain-Computer Interfaces), Graz University of Technology , 2008, Accessed: Nov. 27, 2021. ...

    2008

  18. [26]

    ‘BCI competition 2008 –Graz data set B’ ,

    C. B. G. M. -P. A. S. and G. P. R. Leeb, “‘BCI competition 2008 –Graz data set B’ ,” Graz Univ. Technol., Graz, Austria, Tech. Rep., , pp. 1–6, 2008

    2008

  19. [27]

    Validation of the g.tec Unicorn Hybrid Black wireless EEG system,

    M. B. Pontifex and C. A. Coffman, “Validation of the g.tec Unicorn Hybrid Black wireless EEG system,” Psychophysiology, vol. 60, no. 9, p. e14320, Sep. 2023, doi: 10.1111/PSYP.14320

    2023 doi

  20. [28]

    Generalized digital butterworth filter design,

    I. W. Selesnick and C. Sidney Burrus, “Generalized digital butterworth filter design,” IEEE Transactions on Signal Processing , vol. 46, no. 6, pp. 1688 –1694, 1998, doi: 10.1109/78.678493

    1998 doi

  21. [29]

    A compact multi -branch 1D convolutional neural network for EEG-based motor imagery classification,

    X. Liu, S. Xiong, X. Wang, T. Liang, H. Wang, and X. Liu, “A compact multi -branch 1D convolutional neural network for EEG-based motor imagery classification,” Biomed Signal Process Control, vol. 81, Mar. 2023, doi: 10.1016/j.bspc.2022.104456

    2023

  22. [30]

    Very Deep Convolutional Networks for Large-Scale Image Recognition,

    K. Simonyan and A. Zisserman, “Very Deep Convolutional Networks for Large-Scale Image Recognition,” 3rd International Conference on Learning Representations, ICLR 2015 - Conference Track Proceedings , Sep. 2014, Accessed: Aug. 22, 2024. [Online]. Available: https://arxiv.org/a...

    2015 arXiv

  23. [31]

    P-TELU: Parametric Tan Hyperbolic Linear Unit Activation for Deep Neural Networks,

    R. Duggal and A. Gupta, “P-TELU: Parametric Tan Hyperbolic Linear Unit Activation for Deep Neural Networks,” 2017. Accessed: Oct. 21, 2024. [Online]. Available: http://sbilab.iiitd.edu.in/

    2017

  24. [32]

    Improving the efficiency of RMSProp optimizer by utilizing Nestrove in deep learning,

    R. Elshamy, O. Abu-Elnasr, M. Elhoseny, and S. Elmougy, “Improving the efficiency of RMSProp optimizer by utilizing Nestrove in deep learning,” 123AD, doi: 10.1038/s41598- 023-35663-x

    doi

  25. [33]

    Cross -Entropy Loss Functions: Theoretical Analysis and Applications,

    A. Mao, M. Mohri, and Y. Zhong, “Cross -Entropy Loss Functions: Theoretical Analysis and Applications,” International Conference on Machine Learning , 2023, doi: 10.48550/ARXIV.2304.07288

    2023 doi arXiv

  26. [34]

    Understanding and Improving Early Stopping for Learning with Noisy Labels,

    Y. Bai et al., “Understanding and Improving Early Stopping for Learning with Noisy Labels,” Adv Neural Inf Process Syst, vol. 29, pp. 24392–24403, Jun. 2021, Accessed: Oct. 11,

    2021

  27. [35]

    MSATNet: multi-scale adaptive transformer network for motor imagery classification,

    L. Hu, W. Hong, and L. Liu, “MSATNet: multi-scale adaptive transformer network for motor imagery classification,” Front Neurosci , vol. 17, p. 1173778, Jun. 2023, doi: 10.3389/FNINS.2023.1173778/BIBTEX

    2023

  28. [36]

    Analysis of small sample size studies using nonparametric bootstrap test with pooled resampling method,

    A. K. Dwivedi, I. Mallawaarachchi, and L. A. Alvarado, “Analysis of small sample size studies using nonparametric bootstrap test with pooled resampling method,” Stat Med, vol. 36, no. 14, pp. 2187–2205, Jun. 2017, doi: 10.1002/SIM.7263

    2017 doi

  29. [37]

    Deep Learning for EEG motor imagery classification based on multi -layer CNNs feature fusion,

    S. U. Amin, M. Alsulaiman, G. Muhammad, M. A. Mekhtiche, and M. Shamim Hossain, “Deep Learning for EEG motor imagery classification based on multi -layer CNNs feature fusion,” Future Generation Computer Systems, vol. 101, pp. 542–554, Dec. 2019, doi: 10.1016/J.FUTURE.2019.06.027

    2019 doi

  30. [38]

    EISATC -Fusion: Inception Self - Attention Temporal Convolutional Network Fusion for Motor Imagery EEG Decoding,

    G. Liang, D. Cao, J. Wang, Z. Zhang, and Y. Wu, “EISATC -Fusion: Inception Self - Attention Temporal Convolutional Network Fusion for Motor Imagery EEG Decoding,” IEEE 33 Transactions on Neural Systems and Rehabilitation Engineering , vol. 32, pp. 1535 –1545, 2024, doi: 10.110...

    2024

  31. [39]

    Seizure Prediction in EEG Signals Using STFT and Domain Adaptation,

    P. Peng, Y. Song, L. Yang, and H. Wei, “Seizure Prediction in EEG Signals Using STFT and Domain Adaptation,” Front Neurosci , vol. 15, p. 825434, Jan. 2022, doi: 10.3389/FNINS.2021.825434/BIBTEX

    2022

  32. [40]

    EEG classification across sessions and across subjects through transfer learning in motor imagery -based brain-machine interface system,

    M. Zheng, B. Yang, and Y. Xie, “EEG classification across sessions and across subjects through transfer learning in motor imagery -based brain-machine interface system,” Med Biol Eng Comput , vol. 58, no. 7, pp. 1515 –1528, Jul. 2020, doi: 10.1007/S11517 -020-02176- Y/TABLES/4

    2020 doi

  33. [41]

    Improving cross -subject classification performance of motor imagery signals: a data augmentation -focused deep learning framework,

    E. Ozelbas, E. E. Tülay, and S. Ozekes, “Improving cross -subject classification performance of motor imagery signals: a data augmentation -focused deep learning framework,” Mach Learn Sci Technol, vol. 5, no. 1, p. 015021, Feb. 2024, doi: 10.1088/2632- 2153/AD200C

    2024 doi

  34. [2012]

    doi: 10.3390/s120201211

    doi

  35. [2024]

    Available: https://arxiv.org/abs/2106.15853v2

    [Online]. Available: https://arxiv.org/abs/2106.15853v2

    arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.