The DOI matches its Crossref record; 50 of the 52 references it cites were found in a registry or an index; 9 of the 30 sources its claims cite were checked for support.
Transfer Learning for Brain–Computer Interfaces: A Euclidean Space Data Alignment Approach
Open accessAbstract
OBJECTIVE: This paper targets a major challenge in developing practical electroencephalogram (EEG)-based brain-computer interfaces (BCIs): how to cope with individual differences so that better learning performance can be obtained for a new subject, with minimum or even no subject-specific data? METHODS: We propose a novel approach to align EEG trials from different subjects in the Euclidean space to make them more similar, and hence improve the learning performance for a new subject. Our approach has three desirable properties: first, it aligns the EEG trials directly in the Euclidean space, and any signal processing, feature extraction, and machine learning algorithms can then be applied to the aligned trials; second, its computational cost is very low; and third, it is unsupervised and does not need any label information from the new subject. RESULTS: Both offline and simulated online experiments on motor imagery classification and event-related potential classification verified that our proposed approach outperformed a state-of-the-art Riemannian space data alignment approach, and several approaches without data alignment. CONCLUSION: The proposed Euclidean space EEG data alignment approach can greatly facilitate transfer learning in BCIs. SIGNIFICANCE: Our proposed approach is effective, efficient, and easy to implement. It could be an essential pre-processing step for EEG-based BCIs.
Overview
CoChat Summary
TL;DR
The paper proposes Euclidean-space alignment (EA), an unsupervised transfer-learning preprocessing step that aligns EEG trials from different subjects in Euclidean space to reduce individual differences for BCI classification. Across offline and simulated online experiments on motor imagery and ERP datasets, EA generally outperformed Riemannian-space alignment and no-alignment baselines while being much faster.
Key findings
- In offline unsupervised MI classification, EA-CSP-LDA achieved average accuracies of 79.79% on MI Dataset 1 and 73.53% on MI Dataset 2a, versus RA-MDRM at 68.07% and 70.91%, CSP-LDA at 59.71% and 67.75%, and MDRM at 54.36% and 57.18%.
- In offline ERP classification, EA-xDAWN-SVM achieved an average balanced classification accuracy (BCA) of 68.80%, compared with EA-SVM 67.85%, SVM 64.64%, and xDAWN-SVM 64.60%; EA-SVM significantly outperformed SVM (p=0.0386) and EA-xDAWN-SVM significantly outperformed xDAWN-SVM (p=0.0449).
- In simulated online MI classification, EA-CSP-LDA outperformed RA-MDRM on 12 out of 16 subjects; in simulated online ERP classification, EA-xDAWN-SVM outperformed RA-MDRM on all 11 subjects.
- EA-CSP-LDA was 3.6–19.5 times faster than RA-MDRM; for example, on MI Dataset 1 it took 0.3864 s versus 7.5326 s, and on MI Dataset 2a it took 0.2405 s versus 0.8766 s.
Methods
- The proposed EA aligns time-domain EEG trials in Euclidean space by using the arithmetic mean of all covariance matrices from a subject as the reference matrix, then transforming trials so that the mean covariance matrix of each subject becomes the identity matrix; it requires no labels from the new subject.
- Experiments used two MI datasets from BCI Competition IV (Dataset 1: 7 subjects, 59-channel EEG, 100 trials per class in calibration; Dataset 2a: 9 subjects, 22-channel EEG plus 3-channel EOG, 72 trials per class for left- and right-hand MI) and one ERP RSVP dataset from PhysioNet (11 subjects, 8 channels, 5 Hz Session a, target-to-non-target ratio around 1:9).
- Evaluation included offline unsupervised leave-one-subject-out cross-validation and simulated online supervised classification, with CSP-LDA, MDRM, SVM, and xDAWN-SVM pipelines; paired-sample t-tests and area under the curve (AUC) were used for statistical comparisons.
Limitations
- The paper states that EA only compensates for covariate shift and ignores prior probability shift and concept shift; it may even increase concept shift, and bad trials or outliers can corrupt the reference matrix. It also states that the simulated online experiments are not identical to real online experiments.
- The text makes evident that EA degraded accuracy for some subjects (e.g., Subjects 4 and 9 in MI Dataset 2a, and some ERP subjects), and that RA-MDRM could not be applied to offline unsupervised ERP classification because it requires labeled non-target and target trials from the new subject.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
4 Valid · 8 Not valid · 15 Not checked · 89 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“So Barachant and Congedo [2] proposed a novel approach to augment the ERP trials so that their covariance matrices can also encode the temporal information.”
What the sources say
- Supportedcites [2] [2] A. Barachant and M. Congedo, 'A plug & play P300 BCI using information geometry,' arXiv: 1409.0107 , 2014. - source
Section B (pp. 3–4) describes a prototyped ERP response obtained by averaging single-trial responses, then concatenating it with each trial to form a super trial. The resulting covariance matrices include a cross-covariance block that carries information about the ERP waveform’s temporal structure. (LLM, conf=0.94)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“However, the main discriminative information of ERP trials is carried temporally rather than spatially.”
What the sources say
- Supportedcites [2] [2] A. Barachant and M. Congedo, 'A plug & play P300 BCI using information geometry,' arXiv: 1409.0107 , 2014. - source
Section B (p. 3) says ERP discriminative information lies in the potential’s waveform rather than the spatial distribution of its variance. It also says ignoring temporal structure loses information needed for classification, supporting the paraphrase. (LLM, conf=0.93)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Our proposed EA-CSP-LDA was 3.6-19.5 times faster than RA-MDRM, and also it had much smaller standard deviation.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“On average RI-MDRM outperformed RR-MDRM, and EI-CSP-LDA outperformed ER-CSP-LDA, on both datasets, suggesting that estimating the reference matrix from all imagery trials would be better than using all resting trials.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Our proposed EA-CSP-LDA was 3.6-19.5 times faster than RA-MDRM, and also it had much smaller standard deviation.”
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
47 more claims
Create a free account to see all 116 claims and to verify the 4 references still open.
- It may state two separate claims as one. 47
- Probably not a substantial claim. 30
- The quote only partly states this claim. 1
Citations
Cites
LIBSVM: A library for support vector machines
Chang, C.-C., & Lin, C.-J. (2011) · ACM Transactions on Intelligent Systems and Technology · doi:10.1145/1961189.1961199 · Cited by 40,918 · cited 1×
“LibSVM [6] with a linear kernel was used for classification.”
Method(2008) · Journal of Machine Learning Research · Cited by 33,051
Delorme, A., & Makeig, S. (2004) · Journal of Neuroscience Methods · doi:10.1016/j.jneumeth.2003.10.009 · Cited by 24,371
Pan, S. J., & Yang, Q. (2010) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2009.191 · Cited by 23,387 · cited 2×
“This idea has been widely used in TL [24], [31], [40].”
BackgroundMethodBook Review: Pattern Recognition and Machine Learning
Nasrabadi, N. M. (2007) · Journal of Electronic Imaging · doi:10.1117/1.2819119 · Cited by 22,066
A Simple Sequentially Rejective Multiple Test Procedure
(1979) · Scandinavian Journal of Statistics · doi:10.2307/4615733 · Cited by 21,526 · cited 1×
“Hence, we performed Holm-Bonferroni corrections [12] on the three p-values of each dataset.”
MethodGoldberger, A. L., Amaral, L. A. N., Glass, L., Hausdorff, J. M., Ivanov, P. C., Mark, R. G., Mietus, J. E., Moody, G. B., Peng, C.-K., & Stanley, H. E. (2000) · Circulation · doi:10.1161/01.cir.101.23.e215 · Cited by 14,809
Pattern Recognition and Machine Learning
(2006) · doi:10.1007/978-0-387-45528-0 · Cited by 9,824 · cited 1×
“Usually the EEG signals are first band-pass filtered and spatially filtered to increase the signal-to-noise ratio, and then discriminative features are extracted, which are next fed into machine learning algorithms such as Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM) [3] for classification.”
MethodBrain–computer interfaces for communication and control
Wolpaw, J. R., Birbaumer, N., McFarland, D. J., Pfurtscheller, G., & Vaughan, T. M. (2002) · Clinical Neurophysiology · doi:10.1016/s1388-2457(02)00057-3 · Cited by 7,927
On the Kolmogorov-Smirnov Test for Normality with Mean and Variance Unknown
Lilliefors, H. W. (1967) · Journal of the American Statistical Association · doi:10.2307/2283970 · Cited by 4,530
Cited by
Artificial intelligence for the metaverse: A survey
Huynh-The, T., Pham, Q.-V., Pham, X.-Q., Nguyen, T. T., Han, Z., & Kim, D.-S. (2023) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2022.105581 · Cited by 710
“With electroencephalogram (EEG) signal as one of the most popular inputs of BCI systems, the work in [128] studied a”
MethodA review on transfer learning in EEG signal analysis
Wan, Z., Yang, R., Huang, M., Zeng, N., & Liu, X. (2021) · Neurocomputing · doi:10.1016/j.neucom.2020.09.017 · Cited by 468
Zhang, W., Deng, L., Zhang, L., & Wu, D. (2023) · IEEE/CAA Journal of Automatica Sinica · doi:10.1109/jas.2022.106004 · Cited by 427
Self-citationTransfer Learning for EEG-Based Brain–Computer Interfaces: A Review of Progress Made Since 2016
Wu, D., Xu, Y., & Lu, B.-L. (2022) · IEEE Transactions on Cognitive and Developmental Systems · doi:10.1109/tcds.2020.3007453 · Cited by 434
“For example, EA and LA data alignment schemes have been combined with TL classifiers [50], [60], and CSDF-ReliefF and CSDF-mRMR feature selection approaches have also been integrated with TL classifiers [114].”
Self-citationProgress in Brain Computer Interface: Challenges and Opportunities
Saha, S., Mamun, K. A., Ahmed, K., Mostafa, R., Naik, G. R., Darvishi, S., Khandoker, A. H., & Baumert, M. (2021) · Frontiers in Systems Neuroscience · doi:10.3389/fnsys.2021.578875 · Cited by 365
Metadata
- Authors
- He, H., & Wu, D.
- Affiliations
- Huazhong University of Science and Technology (CN)
- ORCID
- Venue
- IEEE Transactions on Biomedical Engineering
- Published
- 2020-02
- Volume
- 67
- Issue
- 2
- Pages
- 399-410
- ISSN
- 0018-9294 (ISSN-L)
- 1558-2531
- OpenAlex
- W2963727766
- Cited by
- 569
- References
- 52
- Citations by year
- 2026: 135 · 2025: 117 · 2024: 101 · 2023: 69 · 2022: 70 · 2021: 39 · 2020: 26
- FWCI
- 19.17
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- IEEE Transactions on Biomedical Engineering (publishedVersion)
- PubMed (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Life Sciences › Neuroscience › Cognitive Neuroscience › EEG and Brain-Computer Interfaces
- Physical Sciences › Computer Science › Artificial Intelligence › Emotion and Mood Recognition
- Physical Sciences › Computer Science › Human-Computer Interaction › Gaze Tracking and Assistive Technology
- Keywords
- brain-computer interface, EEG, transfer learning, data alignment, interindividual variability, motor imagery classification, unsupervised alignment, EEG preprocessing
- MeSH terms
- Machine Learning, Algorithms, Computer Simulation, Electroencephalography, Evoked Potentials, Humans, Imagination, Signal Processing, Computer-Assisted, Databases, Factual, Brain-Computer Interfaces
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.