Linear and nonlinear methods for brain-computer interfaces

Müller, Klaus‐Robert; Anderson, Charles W.; Birch, Gary E.

doi:10.1109/tnsre.2003.814484

Cited by 331 publications

(170 citation statements)

References 22 publications

Supporting

Mentioning

165

Contrasting

Unclassified

Order By: Relevance

“…One explanation for this decrease is that bad model selection strategies have resulted in overly complex classification models that overfit the EEG data [23]. The current work has clearly shown that an alternative reason for failure should also be considered: non-stationarities in the EEG statistics.…”

Section: Resultsmentioning

confidence: 90%

“…In this setup, we use linear discriminant analysis (LDA) to separate data points with high accuracy into classes in the low-dimensional feature space. Note that more elaborate paradigms or other feature extraction techniques may require the use of non-linear classifiers (cf [21,23,25,31]). Previous work [3,5,8] has reported on the efficacy of our classification scheme.…”

Section: The Berlin Bcimentioning

confidence: 99%

See 1 more Smart Citation

Towards adaptive classification for BCI

Shenoy

Krauledat

Blankertz³

et al. 2006

J. Neural Eng.

390

330

View full text Add to dashboard Cite

Non-stationarities are ubiquitous in EEG signals. They are especially apparent in the use of EEG-based brain-computer interfaces (BCIs): (a) in the differences between the initial calibration measurement and the online operation of a BCI, or (b) caused by changes in the subject's brain processes during an experiment (e.g. due to fatigue, change of task involvement, etc). In this paper, we quantify for the first time such systematic evidence of statistical differences in data recorded during offline and online sessions. Furthermore, we propose novel techniques of investigating and visualizing data distributions, which are particularly useful for the analysis of (non-)stationarities. Our study shows that the brain signals used for control can change substantially from the offline calibration sessions to online control, and also within a single session. In addition to this general characterization of the signals, we propose several adaptive classification schemes and study their performance on data recorded during online experiments. An encouraging result of our study is that surprisingly simple adaptive methods in combination with an offline feature selection scheme can significantly increase BCI performance.

show abstract

Section: Resultsmentioning

confidence: 90%

Section: The Berlin Bcimentioning

confidence: 99%

Towards adaptive classification for BCI

Shenoy

Krauledat

Blankertz³

et al. 2006

J. Neural Eng.

390

330

View full text Add to dashboard Cite

show abstract

“…That is, we select the trial that has all the remaining trials forming a matrix with the largest 1-norm for its singular values (similar to the strategy used with compressive sensing (Candè & Wakin, 2008 (Lotte, Congedo, Lécuyer, Lamarche, & Arnaldi, 2007), (Müller, Anderson, & Birch, 2003), (Croux, Filzmoser, & Joossens, 2008 (Müller et al, 2003), (Croux et al, 2008). However, these techniques are black box models which are difficult to understand and analyze by a normal clinician.…”

Section: Singular Value Decomposition (Svd) Methodsmentioning

confidence: 99%

A Survey on the Project in title "Unsupervised Adaptive P300 BCI in the Framework of Chaotic Theory and Stochastic Theory"

Alhaddad¹

2016

IJAERS

View full text Add to dashboard Cite

Abstract-In this paper we present a survey of work that has been done in the project "Unsupervised Adaptive P300 BCI in the framework of chaotic theory and stochastic theory"we summarised the following papers, (Mohammed J Alhaddad & 2011), (Mohammed J. Alhaddad & Kamel M, 2012), (Mohammed J Alhaddad, Kamel, & Al-Otaibi, 2013), (Mohammed J Alhaddad, Kamel, & Bakheet, 2013), (Mohammed J Alhaddad, Kamel, & Al-Otaibi, 2014), (Mohammed J Alhaddad, Kamel, & Bakheet, 2014), (Mohammed J Alhaddad, Kamel, & Kadah, 2014), (Mohammed J Alhaddad, Kamel, Makary, Hargas, & Kadah, 2014), (Mohammed J Alhaddad, Mohammed, Kamel, & Hagras, 2015

show abstract

“…The second advantage is that one may use non-linear similarity measures to construct K, which is equivalent to performing linear classification on data points that have been mapped into a higher-dimensional feature space and which can consequently yield a more powerful classifier, without the requirement that the feature-space mapping be known explicitly (the socalled kernel trick ). However, it has generally been observed in BCI classification applications (for example, see Müller et al, 2003) that, given a well-chosen sequence of preprocessing steps (an explicit feature mapping), a further implicit mapping is usually unnecessary: thus a linear classifier, in which K ij is equal to the dot-product between the feature representations of data points i and j, performs about as well as any non-linear classifier one might attempt. This is often the case in situations in which the number of data points is low, and indeed we find it to be the case in the current application.…”

Section: Support Vector Machinesmentioning

confidence: 99%

Classifying Event-Related Desynchronization in EEG, ECoG and MEG Signals

Hill

Lal

Schröder³

et al. 2006

Lecture Notes in Computer Science

View full text Add to dashboard Cite

We present the results from three motor-imagery-based Brain-Computer Interface experiments. Brain signals were recorded from 8 untrained subjects using EEG, 4 using ECoG and 10 using MEG. In all cases, we aim to develop a system that could be used for fast, reliable preliminary screening in the clinical application of a BCI, so we aim to obtain the best possible classification performance in a short time. Accordingly, the burden of adaptation is on the side of the computer rather than the user, so we must adopt a machine-learning approach to the analysis. We introduce the required machine-learning vocabulary and concepts, and then present quantitative results that focus on two main issues. The first is the effect of the number of trials-how long does the recording session need to be? We find that good performance could be achieved, on average, after the first 200 trials in EEG, 75-100 trials in MEG, or 25-50 trials in ECoG. The second issue is the effect of spatial filtering-we compare the performance of the original sensor signals with that of the outputs of Independent Component Analysis and the Common Spatial Pattern algorithm, in each of the three sensor types. We find that spatial filtering does not help in MEG, helps a little in ECoG, and improves performance a great deal in EEG. The unsupervised ICA algorithm performed at least as well as the supervised CSP algorithm in all cases-the latter suffered from poor generalization performance due to overfitting in ECoG and MEG, although this could be alleviated by reducing the number of sensors used as input to the algorithm.

show abstract

Linear and nonlinear methods for brain-computer interfaces

Cited by 331 publications

References 22 publications

Towards adaptive classification for BCI

Towards adaptive classification for BCI

A Survey on the Project in title "Unsupervised Adaptive P300 BCI in the Framework of Chaotic Theory and Stochastic Theory"

Classifying Event-Related Desynchronization in EEG, ECoG and MEG Signals

Contact Info

Product

Resources

About