Learning suite of kernel feature spaces enhances SMR-based EEG-BCI Classification

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Citations (Scopus)

Abstract

Brain-Computer Interface (BCI) research hopes to improve the quality of life for people with severe motor disabilities by providing a capability to control external devices using their thoughts. To control a device through BCI, neural signals of a user must be translated to meaningful control commands using various machine learning components, e.g. feature extraction, dimensionality reduction and classification, that should also be carefully designed for practical use. However, the noise and variability in the neural data pose one of the greatest challenges that in practice previously functioning BCI fails in the subsequent operation requiring re-Tuning/optimization. This paper presents an idea of defining multiple feature spaces and optimal decision boundaries therein to account for noise and variability in data and improve a generalization of a learning machine. The spaces are defined in the Reproducing Kernel Hilbert Spaces induced by a Radial Basis Gaussian function. Then the learning is done via L1-regularized Support Vector Machines. The central idea behind our approach is that a classifier predicts an unseen test examples by learning more rich feature spaces with a suite of optimal hyperparameters. Empirical evaluation have shown an improved generalization performance (range 79-90%) on two class motor imagery Electroencephalography (EEG) data, when compared with other conventional machine learning methods.

Original languageEnglish
Title of host publication5th International Winter Conference on Brain-Computer Interface, BCI 2017
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages55-59
Number of pages5
ISBN (Electronic)9781509050963
DOIs
Publication statusPublished - Feb 16 2017
Externally publishedYes
Event5th International Winter Conference on Brain-Computer Interface, BCI 2017 - Gangwon Province, Korea, Republic of
Duration: Jan 9 2017Jan 11 2017

Conference

Conference5th International Winter Conference on Brain-Computer Interface, BCI 2017
CountryKorea, Republic of
CityGangwon Province
Period1/9/171/11/17

Fingerprint

Brain computer interface
Electroencephalography
Learning systems
Hilbert spaces
Support vector machines
Feature extraction
Classifiers
Tuning

ASJC Scopus subject areas

  • Signal Processing
  • Human-Computer Interaction

Cite this

Abibullaev, B. (2017). Learning suite of kernel feature spaces enhances SMR-based EEG-BCI Classification. In 5th International Winter Conference on Brain-Computer Interface, BCI 2017 (pp. 55-59). [7858158] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/IWW-BCI.2017.7858158

Learning suite of kernel feature spaces enhances SMR-based EEG-BCI Classification. / Abibullaev, Berdakh.

5th International Winter Conference on Brain-Computer Interface, BCI 2017. Institute of Electrical and Electronics Engineers Inc., 2017. p. 55-59 7858158.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abibullaev, B 2017, Learning suite of kernel feature spaces enhances SMR-based EEG-BCI Classification. in 5th International Winter Conference on Brain-Computer Interface, BCI 2017., 7858158, Institute of Electrical and Electronics Engineers Inc., pp. 55-59, 5th International Winter Conference on Brain-Computer Interface, BCI 2017, Gangwon Province, Korea, Republic of, 1/9/17. https://doi.org/10.1109/IWW-BCI.2017.7858158
Abibullaev B. Learning suite of kernel feature spaces enhances SMR-based EEG-BCI Classification. In 5th International Winter Conference on Brain-Computer Interface, BCI 2017. Institute of Electrical and Electronics Engineers Inc. 2017. p. 55-59. 7858158 https://doi.org/10.1109/IWW-BCI.2017.7858158
Abibullaev, Berdakh. / Learning suite of kernel feature spaces enhances SMR-based EEG-BCI Classification. 5th International Winter Conference on Brain-Computer Interface, BCI 2017. Institute of Electrical and Electronics Engineers Inc., 2017. pp. 55-59
@inproceedings{65d98436a42b4c5290e4b21677cfa4c8,
title = "Learning suite of kernel feature spaces enhances SMR-based EEG-BCI Classification",
abstract = "Brain-Computer Interface (BCI) research hopes to improve the quality of life for people with severe motor disabilities by providing a capability to control external devices using their thoughts. To control a device through BCI, neural signals of a user must be translated to meaningful control commands using various machine learning components, e.g. feature extraction, dimensionality reduction and classification, that should also be carefully designed for practical use. However, the noise and variability in the neural data pose one of the greatest challenges that in practice previously functioning BCI fails in the subsequent operation requiring re-Tuning/optimization. This paper presents an idea of defining multiple feature spaces and optimal decision boundaries therein to account for noise and variability in data and improve a generalization of a learning machine. The spaces are defined in the Reproducing Kernel Hilbert Spaces induced by a Radial Basis Gaussian function. Then the learning is done via L1-regularized Support Vector Machines. The central idea behind our approach is that a classifier predicts an unseen test examples by learning more rich feature spaces with a suite of optimal hyperparameters. Empirical evaluation have shown an improved generalization performance (range 79-90{\%}) on two class motor imagery Electroencephalography (EEG) data, when compared with other conventional machine learning methods.",
author = "Berdakh Abibullaev",
year = "2017",
month = "2",
day = "16",
doi = "10.1109/IWW-BCI.2017.7858158",
language = "English",
pages = "55--59",
booktitle = "5th International Winter Conference on Brain-Computer Interface, BCI 2017",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
address = "United States",

}

TY - GEN

T1 - Learning suite of kernel feature spaces enhances SMR-based EEG-BCI Classification

AU - Abibullaev, Berdakh

PY - 2017/2/16

Y1 - 2017/2/16

N2 - Brain-Computer Interface (BCI) research hopes to improve the quality of life for people with severe motor disabilities by providing a capability to control external devices using their thoughts. To control a device through BCI, neural signals of a user must be translated to meaningful control commands using various machine learning components, e.g. feature extraction, dimensionality reduction and classification, that should also be carefully designed for practical use. However, the noise and variability in the neural data pose one of the greatest challenges that in practice previously functioning BCI fails in the subsequent operation requiring re-Tuning/optimization. This paper presents an idea of defining multiple feature spaces and optimal decision boundaries therein to account for noise and variability in data and improve a generalization of a learning machine. The spaces are defined in the Reproducing Kernel Hilbert Spaces induced by a Radial Basis Gaussian function. Then the learning is done via L1-regularized Support Vector Machines. The central idea behind our approach is that a classifier predicts an unseen test examples by learning more rich feature spaces with a suite of optimal hyperparameters. Empirical evaluation have shown an improved generalization performance (range 79-90%) on two class motor imagery Electroencephalography (EEG) data, when compared with other conventional machine learning methods.

AB - Brain-Computer Interface (BCI) research hopes to improve the quality of life for people with severe motor disabilities by providing a capability to control external devices using their thoughts. To control a device through BCI, neural signals of a user must be translated to meaningful control commands using various machine learning components, e.g. feature extraction, dimensionality reduction and classification, that should also be carefully designed for practical use. However, the noise and variability in the neural data pose one of the greatest challenges that in practice previously functioning BCI fails in the subsequent operation requiring re-Tuning/optimization. This paper presents an idea of defining multiple feature spaces and optimal decision boundaries therein to account for noise and variability in data and improve a generalization of a learning machine. The spaces are defined in the Reproducing Kernel Hilbert Spaces induced by a Radial Basis Gaussian function. Then the learning is done via L1-regularized Support Vector Machines. The central idea behind our approach is that a classifier predicts an unseen test examples by learning more rich feature spaces with a suite of optimal hyperparameters. Empirical evaluation have shown an improved generalization performance (range 79-90%) on two class motor imagery Electroencephalography (EEG) data, when compared with other conventional machine learning methods.

UR - http://www.scopus.com/inward/record.url?scp=85016000035&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85016000035&partnerID=8YFLogxK

U2 - 10.1109/IWW-BCI.2017.7858158

DO - 10.1109/IWW-BCI.2017.7858158

M3 - Conference contribution

SP - 55

EP - 59

BT - 5th International Winter Conference on Brain-Computer Interface, BCI 2017

PB - Institute of Electrical and Electronics Engineers Inc.

ER -