CoChat

Transfer Learning Algorithm of P300-EEG Signal Based on XDAWN Spatial Filter and Riemannian Geometry Classifier

Open access
Li, Feng, Xia, Yi, Wang, Fei, Zhang, Dengyong, Li, Xiaoyu, He, Fan · 2020 · Applied Sciences

The DOI matches its Crossref record; 39 of the 39 references it cites were found in a registry or an index; 2 of the 15 sources its claims cite were checked for support.

Abstract

The electroencephalogram (EEG) signal in the brain–computer interface (BCI) has suffered great cross-subject variability. The BCI system needs to be retrained before each time it is used, which is a waste of resources and time. Thus, it is difficult to generalize a fixed classification method for all subjects. Therefore, the transfer learning method proposed in this article, which combines XDAWN spatial filter and Riemannian Geometry classifier (RGC), can achieve offline cross-subject transfer learning in the P300-speller paradigm. The XDAWN spatial filter is used to enhanced the P300 components in the raw signal as well as reduce its dimensions. Then, the Riemannian Geometry Mean (RGM) is used as the reference matrix to perform the affine transformation of the symmetric positive definite (SPD) covariance matrix calculated from the filtered signal, which makes the data from different subjects comparable. Finally, the RGC is used to obtain the result of transfer learning experiments. The proposed algorithm was evaluated on two datasets (Dataset I from real patients and Dataset II from the laboratory). By comparing with two state-of-the-art and classic algorithms in the current BCI field, Ensemble of Support Vector Machine (E-SVM) and Stepwise Linear Discriminant Analysis (SWLDA), the maximum averaged area under the receiver operating characteristic curve (AUC) score of our algorithm reached 0.836, proving the potential of our proposed algorithm.

Overview

CoChat Summary

TL;DR

The paper proposes an offline cross-subject transfer learning method for P300-EEG that combines an XDAWN spatial filter with a Riemannian Geometry classifier, using each subject’s Riemannian Geometry mean to affine-transform SPD covariance matrices so data from different subjects become comparable. Across two P300-speller datasets, the method outperformed E-SVM and SWLDA, reaching a maximum average AUC of 0.836.

Key findings

  • In one-to-one transfer learning, XDAWN + RGC achieved average AUC 0.776 on Dataset I (max 0.821) and 0.787 on Dataset II (max 0.813), compared with E-SVM averages of 0.547 and 0.542 and SWLDA averages of 0.548 and 0.552.
  • In all-to-one transfer learning, XDAWN + RGC achieved average AUC 0.836 on Dataset I (max 0.879) and 0.830 on Dataset II (max 0.865), compared with E-SVM averages of 0.636 and 0.650 and SWLDA averages of 0.655 and 0.660.
  • XDAWN + RGC also improved slightly over RGC alone in both experiments; for example, all-to-one Dataset I RGC averaged 0.800 versus 0.836 for XDAWN + RGC.
  • Visualization of two subjects per dataset showed that after affine transformation, SPD covariance matrices were more concentrated and consistent in spatial distribution.

Methods

  • Dataset I was a public P300-speller dataset from eight ALS subjects, with 420 samples per subject (70 P300, 350 non-P300), 8 channels, 256 Hz sampling, and 0.1–30 Hz bandpass filtering. Dataset II was collected from 10 healthy subjects (five men, five women), with 1200 samples per subject (30 P300, 1170 non-P300), 32 channels, 250 Hz sampling, and 0.1–60 Hz bandpass filtering.
  • Preprocessing selected 0–0.5 s after flashing, applied a fifth-order Butterworth 0.1–20 Hz bandpass filter, and downsampled to 34 Hz.
  • The algorithm applies XDAWN spatial filtering to enhance P300 components and reduce dimensions, computes SPD covariance matrices, affine-transforms them using each subject’s Riemannian Geometry mean as reference, maps them onto the Riemannian manifold, and classifies by minimum Riemannian distance to class means. Experiments used one-to-one leave-one-subject-out transfer and all-to-one leave-one-out transfer with Bootstrap Aggregating voting; performance was measured by AUC due to class imbalance.

Limitations

  • The paper does not present a dedicated limitations section; it states that future work will focus on a more robust BCI transfer learning algorithm that can be used online, indicating the current method is offline.
  • Evident from the text: evaluation is limited to offline cross-subject transfer within two P300-speller datasets, with no online or real-time validation reported.
  • Evident from the text: transfer learning is defined as training on Subject A and testing on Subject B from the same dataset, so cross-dataset generalization is not tested; sample sizes are also limited to 8 ALS subjects in Dataset I and 10 healthy subjects in Dataset II.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/app10051804

Cites

39 of 39 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

2 Valid · 1 Not valid · 9 Not checked · 27 Own results and interpretations

15 more claims

Create a free account to see all 39 claims and every source verdict.

See all claims — free account
  • It may state two separate claims as one. 12
  • Probably not a substantial claim. 9

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

3 of 3

Citations

Cites

Research works printed but not linked: 3

  1. Clerc, 2013, Brain Computer Interfaces, Principles and Practise, Biomed. Eng. Online, 12, 1
  2. Lang, S. (2012). Differential and Riemannian Manifolds, Springer.
  3. Blankertz, B., Dornhege, G., Müller, K.R., Schalk, G., Krusienski, D., Wolpaw, J.R., Schlogl, A., Graimann, B., Pfurtscheller, G., and Chiappa, S. (2005). Results of the BCI Competition III. BCI Meeting, Elsevier.

Cited by

See all 64

Metadata

Authors
Li, F., Xia, Y., Wang, F., Zhang, D., Li, X., & He, F.
Affiliations
  • Changsha University of Science and Technology (CN)
  • South China Normal University (CN)
Corresponding author
Wang, Fei
Venue
Applied Sciences
Publisher
MDPI AG
Published
2020-03-05
Volume
10
Issue
5
Pages
1804
ISSN
2076-3417 (ISSN-L)
OpenAlex
W3009134349
Cited by
64
References
38
Citations by year
2026: 5 · 2025: 8 · 2024: 11 · 2023: 15 · 2022: 13 · 2021: 7 · 2020: 5
FWCI
4.78
Type
journal article
Language
en
Open access
Locations
  • Applied Sciences (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • National Natural Science Foundation of China (61906019)
  • Natural Science Foundation of Hunan Province (2019JJ50649)
  • Scientific Research Fund of Hunan Provincial Education Department (18C0238 and 19B004)
  • "Double First-class" International Cooperation and Development Scientific Research Project of Changsha University of Science and Technology (2018IC25)
  • Young Teacher Growth Plan Project of Changsha University of Science and Technology (2019QJCZ076)
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › EEG and Brain-Computer Interfaces
  • Physical Sciences › Computer Science › Human-Computer Interaction › Gaze Tracking and Assistive Technology
  • Physical Sciences › Computer Science › Artificial Intelligence › Emotion and Mood Recognition
Keywords
P300, EEG, brain-computer interface, cross-subject transfer learning, P300 speller, cross-subject variability

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 2026-10-07, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-07 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

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.