Limits...
Online artifact removal for brain-computer interfaces using support vector machines and blind source separation.

Halder S, Bensch M, Mellinger J, Bogdan M, Kübler A, Birbaumer N, Rosenstiel W - Comput Intell Neurosci (2007)

Bottom Line: In order to select a suitable BSS/ICA method, three ICA algorithms (JADE, Infomax, and FastICA) and one BSS algorithm (AMUSE) are evaluated to determine their ability to isolate electromyographic (EMG) and electrooculographic (EOG) artifacts into individual components.An implementation of the selected BSS/ICA method with SVMs trained to classify EMG and EOG artifacts, which enables the usage of the method as a filter in measurements with online feedback, is described.This filter is evaluated on three BCI datasets as a proof-of-concept of the method.

View Article: PubMed Central - PubMed

Affiliation: Institute of Medical Psychology and Behavioral Neurobiology, University of Tübingen, Gartenstr. 29, 72074 Tübingen, Germany. halder@informatik.uni-tuebingen.de

ABSTRACT
We propose a combination of blind source separation (BSS) and independent component analysis (ICA) (signal decomposition into artifacts and nonartifacts) with support vector machines (SVMs) (automatic classification) that are designed for online usage. In order to select a suitable BSS/ICA method, three ICA algorithms (JADE, Infomax, and FastICA) and one BSS algorithm (AMUSE) are evaluated to determine their ability to isolate electromyographic (EMG) and electrooculographic (EOG) artifacts into individual components. An implementation of the selected BSS/ICA method with SVMs trained to classify EMG and EOG artifacts, which enables the usage of the method as a filter in measurements with online feedback, is described. This filter is evaluated on three BCI datasets as a proof-of-concept of the method.

No MeSH data available.


Related in: MedlinePlus

Topographic plots illustrating the differences in the features usedfor classification. The topographies of two ICs containing eye blinks (left) andeye movement (right) are shown.
© Copyright Policy - open-access
Related In: Results  -  Collection


getmorefigures.php?uid=PMC2234090&req=5

fig2: Topographic plots illustrating the differences in the features usedfor classification. The topographies of two ICs containing eye blinks (left) andeye movement (right) are shown.

Mentions: EEG data recorded while the subject was performing no particular task was cleaned of blinks and other obvious artifacts by removing the corresponding sections using EEGLAB and then used as background . To obtain the first artifact source, an EMG recording was made on the forearm of a subject. This ensures that no CNS signals are contained in this artifact component. We assume that the spectral properties of an EMG signal generated at the forearm are comparable to those generated by muscles located on the head, for example, jaw muscles. This assumption seems to be a feasible tradeoff considering that it ensures that no CNS signals are contained in the EMG signal. The mixing matrix is constructed from jaw muscle recordings so that the spatial pattern is also as similar as possible to a real-jaw muscle recording. To ensure the availability of an EOG artifact component free of CNS signals, 20 blinks from channel Fp1 (see Figure 2 for electrode location) of an artifact recording were extracted, averaged, and then added to a zero baseline signal with varying gaps and a random multiplication factor ranging from 0.5 to 1.5.


Online artifact removal for brain-computer interfaces using support vector machines and blind source separation.

Halder S, Bensch M, Mellinger J, Bogdan M, Kübler A, Birbaumer N, Rosenstiel W - Comput Intell Neurosci (2007)

Topographic plots illustrating the differences in the features usedfor classification. The topographies of two ICs containing eye blinks (left) andeye movement (right) are shown.
© Copyright Policy - open-access
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC2234090&req=5

fig2: Topographic plots illustrating the differences in the features usedfor classification. The topographies of two ICs containing eye blinks (left) andeye movement (right) are shown.
Mentions: EEG data recorded while the subject was performing no particular task was cleaned of blinks and other obvious artifacts by removing the corresponding sections using EEGLAB and then used as background . To obtain the first artifact source, an EMG recording was made on the forearm of a subject. This ensures that no CNS signals are contained in this artifact component. We assume that the spectral properties of an EMG signal generated at the forearm are comparable to those generated by muscles located on the head, for example, jaw muscles. This assumption seems to be a feasible tradeoff considering that it ensures that no CNS signals are contained in the EMG signal. The mixing matrix is constructed from jaw muscle recordings so that the spatial pattern is also as similar as possible to a real-jaw muscle recording. To ensure the availability of an EOG artifact component free of CNS signals, 20 blinks from channel Fp1 (see Figure 2 for electrode location) of an artifact recording were extracted, averaged, and then added to a zero baseline signal with varying gaps and a random multiplication factor ranging from 0.5 to 1.5.

Bottom Line: In order to select a suitable BSS/ICA method, three ICA algorithms (JADE, Infomax, and FastICA) and one BSS algorithm (AMUSE) are evaluated to determine their ability to isolate electromyographic (EMG) and electrooculographic (EOG) artifacts into individual components.An implementation of the selected BSS/ICA method with SVMs trained to classify EMG and EOG artifacts, which enables the usage of the method as a filter in measurements with online feedback, is described.This filter is evaluated on three BCI datasets as a proof-of-concept of the method.

View Article: PubMed Central - PubMed

Affiliation: Institute of Medical Psychology and Behavioral Neurobiology, University of Tübingen, Gartenstr. 29, 72074 Tübingen, Germany. halder@informatik.uni-tuebingen.de

ABSTRACT
We propose a combination of blind source separation (BSS) and independent component analysis (ICA) (signal decomposition into artifacts and nonartifacts) with support vector machines (SVMs) (automatic classification) that are designed for online usage. In order to select a suitable BSS/ICA method, three ICA algorithms (JADE, Infomax, and FastICA) and one BSS algorithm (AMUSE) are evaluated to determine their ability to isolate electromyographic (EMG) and electrooculographic (EOG) artifacts into individual components. An implementation of the selected BSS/ICA method with SVMs trained to classify EMG and EOG artifacts, which enables the usage of the method as a filter in measurements with online feedback, is described. This filter is evaluated on three BCI datasets as a proof-of-concept of the method.

No MeSH data available.


Related in: MedlinePlus