Online and Offline Domain Adaptation for Reducing BCI Calibration Effort

Wu, Dongrui

doi:10.1109/thms.2016.2608931

Cited by 105 publications

(85 citation statements)

References 47 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For the two MI datasets, six CSP variances [see (4)] were used as the features. For the two ERP datasets, after spatial filtering by xDAWN, we assembled each EEG trail into a vector, performed principal component analysis on all vectors…”

Section: B Eeg Data Preprocessingmentioning

confidence: 99%

See 1 more Smart Citation

Manifold Embedded Knowledge Transfer for Brain-Computer Interfaces

Zhang

2020

IEEE Trans. Neural Syst. Rehabil. Eng.

Self Cite

130

101

View full text Add to dashboard Cite

Transfer learning makes use of data or knowledge in one problem to help solve a different, yet related, problem. It is particularly useful in brain-computer interfaces (BCIs), for coping with variations among different subjects and/or tasks. This paper considers offline unsupervised cross-subject electroencephalogram (EEG) classification, i.e., we have labeled EEG trials from one or more source subjects, but only unlabeled EEG trials from the target subject. We propose a novel manifold embedded knowledge transfer (MEKT) approach, which first aligns the covariance matrices of the EEG trials in the Riemannian manifold, extracts features in the tangent space, and then performs domain adaptation by minimizing the joint probability distribution shift between the source and the target domains, while preserving their geometric structures. MEKT can cope with one or multiple source domains, and can be computed efficiently. We also propose a domain transferability estimation (DTE) approach to identify the most beneficial source domains, in case there are a large number of source domains. Experiments on four EEG datasets from two different BCI paradigms demonstrated that MEKT outperformed several stateof-the-art transfer learning approaches, and DTE can reduce more than half of the computational cost when the number of source subjects is large, with little sacrifice of classification accuracy.

show abstract

Section: B Eeg Data Preprocessingmentioning

confidence: 99%

“…Electroencephalogram (EEG), a multi-channel time-series, is the most frequently used BCI input signal. There are three common paradigms in EEG-based BCIs: motor imagery (MI) [3], event-related potentials (ERPs) [4], and steady-state visual evoked potentials [2]. The first two are the focus of this paper.…”

Section: Introductionmentioning

confidence: 99%

Manifold Embedded Knowledge Transfer for Brain-Computer Interfaces

Zhang

2020

IEEE Trans. Neural Syst. Rehabil. Eng.

Self Cite

130

101

View full text Add to dashboard Cite

show abstract

“…Increasingly, BCI (brain-computer interface) applications use data obtained from previous sessions or from other subjects to reduce the amount of calibration data required to train EEG classifiers. Most underlying classifiers assume that training and test signals come from the same distributions or use domain adaption to transform the data to make the distributions similar (Wu et al, 2014) (Wu, 2016). Simple scaling improves the similarity of these distributions.…”

Section: Channel Analysismentioning

confidence: 99%

Automated EEG mega-analysis I: Spectral and amplitude characteristics across studies

Bigdely-Shamlo¹,

Touryan

Ojeda³

et al. 2018

Preprint

View full text Add to dashboard Cite

Significant achievements have been made in the fMRI field by pooling statistical results from multiple studies (meta-analysis). More recently, fMRI standardization efforts have focused on enabling the joint analysis of raw fMRI data across studies (mega-analysis), with the hope of achieving more detailed insights. However, it has not been clear if such analyses in the EEG field are possible or equally fruitful. Here we present the results of a large-scale EEG mega-analysis using 18 studies from six sites representing several different experimental paradigms. We demonstrate that when meta-data are consistent across studies, both channel-level and sourcelevel EEG mega-analysis are possible and can provide insights unavailable in single studies. The analysis uses a fully-automated processing pipeline to reduce line noise, interpolate noisy channels, perform robust referencing, remove eye-activity, and further identify outlier signals. We define several robust measures based on channel amplitude and dispersion to assess the comparability of data across studies and observe the effect of various processing steps on these measures. Using ICA-based dipolar sources, we also observe consistent differences in overall frequency baseline amplitudes across brain areas. For example, we observe higher alpha in posterior vs anterior regions and higher beta in temporal regions. We also detect consistent differences in the slope of the aperiodic portion of the EEG spectrum across brain areas. In a companion paper, we apply mega-analysis to assess commonalities in event-related EEG features across studies. The continuous raw and preprocessed data used in this analysis are available through the DataCatalog at https://cancta.net.

show abstract

“…Electroencephalogram (EEG), which measures the brain signal from the scalp, is the most widely used input signal in BCIs, due to its simplicity and low cost [25]. There are different paradigms in using EEG signals in BCIs, e.g., P300 evoked potentials [10], [33], [40], [43], motor imagery (MI) [29], steady-state visual evoked potential (SSVEP) [47], drowsiness/reaction time estimation [41], [42], [44], etc.…”

Section: Introductionmentioning

confidence: 99%

On the Vulnerability of CNN Classifiers in EEG-Based BCIs

Zhang

2019

IEEE Trans. Neural Syst. Rehabil. Eng.

Self Cite

View full text Add to dashboard Cite

Deep learning has been successfully used in numerous applications because of its outstanding performance and the ability to avoid manual feature engineering. One such application is electroencephalogram (EEG) based brain-computer interface (BCI), where multiple convolutional neural network (CNN) models have been proposed for EEG classification. However, it has been found that deep learning models can be easily fooled with adversarial examples, which are normal examples with small deliberate perturbations. This paper proposes an unsupervised fast gradient sign method (UFGSM) to attack three popular CNN classifiers in BCIs, and demonstrates its effectiveness. We also verify the transferability of adversarial examples in BCIs, which means we can perform attacks even without knowing the architecture and parameters of the target models, or the datasets they were trained on. To our knowledge, this is the first study on the vulnerability of CNN classifiers in EEG-based BCIs, and hopefully will trigger more attention on the security of BCI systems.

show abstract

Online and Offline Domain Adaptation for Reducing BCI Calibration Effort

Cited by 105 publications

References 47 publications

Manifold Embedded Knowledge Transfer for Brain-Computer Interfaces

Manifold Embedded Knowledge Transfer for Brain-Computer Interfaces

Automated EEG mega-analysis I: Spectral and amplitude characteristics across studies

On the Vulnerability of CNN Classifiers in EEG-Based BCIs

Contact Info

Product

Resources

About