Calibration-Free Error-Related Potential Decoding With Adaptive Subject-Independent Models: A Comparative Study
Abstract
Error-related potentials (ErrPs) provide an elegant method to improve human-machine interaction by detecting incorrect system behavior from the electroencephalogram of a human operator in real time. In this paper, we focus on adaptive subject-independent decoding models particularly suitable for ErrP classification. As individualized decoding models require a time-consuming calibration phase, such models provide a promising alternative. Based on an investigation of the characteristics of inter-subject variations in the signal and feature space, we evaluate the performance of a decoding model solely trained on prior data and the effectiveness of adapting this model to a new subject in a comparative study. Our results show that such a generalized model can decode ErrPs with an acceptable average accuracy of 72.7 ± 9.66% and that supervised adaptation can significantly improve the accuracy of the generalized model after adaptation with 85 trials by on average +3.8 ± 5.1%. We show that adaptation of subject-independent decoding models is superior to the traditional calibration procedure. Unsupervised adaptation only proved effective for some subjects and requires further attention to be practical for a broader range of subjects. Consequently, our work contributes to the development of calibration-free ErrP decoding in the broader scope of enhancing usability of ErrPs for human-machine interaction.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
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?
Calibration-free cross-subject ErrP classification has been demonstrated with generalized LDA, an xDAWN+RG pipeline with a support vector machine classifier, ensemble classifiers, and online PCA-based pipelines, though without inherent interpretability.
“Calibration-free cross-subject ErrP classification has been demonstrated with generalized LDA [16], an xDAWN+RG pipeline with a support vector machine classifier [17], ensemble classifiers [18], and online PCA-based pipelines [19], though without inherent interpretability.”
From ERP-XTTN: interpretable prototype-guided cross-attention for cross-subject ERP classification (2026) · cites this paper as [16]
On HRI, Sch ̈ onleitner et al. [16] reported 72.7% balanced accuracy with generalized LDA using 27 channels; ERP-XTTN achieved 76.6% with three midline channels.
“On HRI, Sch¨ onleitner et al. [16] reported 72.7% balanced accuracy with generalized LDA using 27 channels; ERP-XTTN achieved 76.6% with three midline channels.”
From ERP-XTTN: interpretable prototype-guided cross-attention for cross-subject ERP classification (2026) · cites this paper as [16]
All results reported here use causal IIR filtering under calibration-free LOSO evaluation; many published LOSO benchmarks use acausal filtering instead, so their reported performance is not reproducible in a forward-processing pipeline.
“All results reported here use causal IIR filtering under calibration-free LOSO evaluation; many published LOSO benchmarks use acausal filtering instead [9], [16], so their reported performance is not reproducible in a forward-processing pipeline.”
From ERP-XTTN: interpretable prototype-guided cross-attention for cross-subject ERP classification (2026) · cites this paper as [16]
Cites
A review of classification algorithms for EEG-based brain–computer interfaces: A 10 year update
Lotte, F., Bougrain, L., Cichocki, A., Clerc, M., Congedo, M., Rakotomamonjy, A., & Yger, F. (2018) · Journal of Neural Engineering · doi:10.1088/1741-2552/aab2f2 · Cited by 2,239 · see citing works
ERP components on reaction errors and their functional significance: A tutorial
Falkenstein, M., Hoormann, J., Christ, S., & Hohnsbein, J. (2000) · Biological Psychology · doi:10.1016/s0301-0511(99)00031-9 · Cited by 1,442 · cited 1× · see citing works
“As error perception and evaluation are essential abilities of all humans, characteristic changes in the neurophysiological activity are measurable in the prefrontal cortex directly after observing wrong responses of an interaction partner [3], [4].”
BackgroundUnsupervised Brain Computer Interface Based on Intersubject Information and Online Adaptation
Lu, S., Guan, C., & Zhang, H. (2009) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2009.2015197 · Cited by 122 · see citing works
A Generic Transferable EEG Decoder for Online Detection of Error Potential in Target Selection
Bhattacharyya, S., Konar, A., Tibarewala, D. N., & Hayashibe, M. (2017) · Frontiers in Neuroscience · doi:10.3389/fnins.2017.00226 · Cited by 43 · cited 1× · see citing works
“3) Generalized Model (GM): A generalized model [35]– [38] can be built without any information of the current subject by training a LDA on a balanced training set which contains all prior data from other subjects.”
Background
Flagged first, then by citations
Cited by
Sorted by citations
ERP-XTTN: interpretable prototype-guided cross-attention for cross-subject ERP classification
Genevier Wyman, C., & Hirshfield, L. (2026) · Journal of Neural Engineering · doi:10.1088/1741-2552/ae9f95
“All results reported here use causal IIR filtering under calibration-free LOSO evaluation; many published LOSO benchmarks use acausal filtering instead [9, 16], so their reported performance is not reproducible in a forward-processing pipeline.”
Influential
- Authors
- Schönleitner, F. M., Otter, L., Ehrlich, S. K., & Cheng, G.
- Affiliations
- Technical University of Munich (DE)
- Venue
- IEEE Transactions on Medical Robotics and Bionics
- Published
- 2020-08
- Volume
- 2
- Issue
- 3
- Pages
- 399-409
- ISSN
- 2576-3202 (ISSN-L)
- OpenAlex
- W3046817909
- Cited by
- 12
- References
- 4
- Citations by year
- 2026: 1 · 2024: 3 · 2022: 2 · 2021: 3 · 2020: 3
- FWCI
- 1.08
- Type
- journal article
- Language
- en
- Open access
- closed
- Licence
- Locations
- IEEE Transactions on Medical Robotics and Bionics (publishedVersion)
- Funders
- Elite Master Program in Neuroengineering at the Technische Universität München
- Elite Network Bavaria
- Topics
- Life Sciences › Neuroscience › Cognitive Neuroscience › EEG and Brain-Computer Interfaces
- Social Sciences › Psychology › Experimental and Cognitive Psychology › Emotion and Mood Recognition
- Physical Sciences › Computer Science › Human-Computer Interaction › Gaze Tracking and Assistive Technology
- Keywords
- error-related potentials, adaptive decoding, supervised adaptation, unsupervised adaptation, interindividual variability, EEG, human-machine interaction, brain-computer interface, error detection
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.
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
