Restricted Boltzmann Machines in Sensory Motor Rhythm Brain-Computer Interfacing: A Study on Inter-Subject Transfer and Co-Adaptation

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Abstract

Naive users often perceive calibration of a Sensory Motor Rhythm (SMR) based Brain-Computer Interfaces (BCI) as tedious and lengthy. The lack of feedback during training is assumed to be a major cause. I.e. if one had already a reasonable model to start with, feedback training could be started immediately. One concept to address this issue is learning a general model and adapting it to new observations. In this study we applied this concept by utilizing a generative model entitled Restricted Boltzmann Machine (RBM). We investigated its feature extraction capabilities by fitting a RBM to recordings of 9 subjects. Generalization was assessed in an online coadaptive study, covering 12 volunteers (10 naive). An overall median accuracy of 88.9% (83.5% naive) with a standard-error of 6.5% (6.6% naive) was achieved for a classical hand versus feet motor imagery task. The online co-adaptive training itself lasted approximately 25 minutes. Feedback was already presented after a one minute setup run, whose purpose was to estimate initial statistics and to train an online artifact detection system.
Original languageEnglish
Title of host publicationSystems, Man, and Cybernetics (SMC), 2016 IEEE International Conference on
Pages 000469 - 000474
ISBN (Electronic)978-1-5090-1897-0
DOIs
Publication statusPublished - 2016
Event2016 IEEE International Conference on Systems, Man, and Cybernetics - Budapest, Hungary
Duration: 9 Oct 201612 Oct 2016

Conference

Conference2016 IEEE International Conference on Systems, Man, and Cybernetics
CountryHungary
CityBudapest
Period9/10/1612/10/16

Fingerprint

Brain
Brain-Computer Interfaces
Imagery (Psychotherapy)
Artifacts
Calibration
Foot
Volunteers
Hand
Learning

Fields of Expertise

  • Human- & Biotechnology

Cite this

Restricted Boltzmann Machines in Sensory Motor Rhythm Brain-Computer Interfacing: A Study on Inter-Subject Transfer and Co-Adaptation. / Kobler, Reinmar; Scherer, Reinhold.

Systems, Man, and Cybernetics (SMC), 2016 IEEE International Conference on . 2016. p. 000469 - 000474.

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Kobler, R & Scherer, R 2016, Restricted Boltzmann Machines in Sensory Motor Rhythm Brain-Computer Interfacing: A Study on Inter-Subject Transfer and Co-Adaptation. in Systems, Man, and Cybernetics (SMC), 2016 IEEE International Conference on . pp. 000469 - 000474, 2016 IEEE International Conference on Systems, Man, and Cybernetics, Budapest, Hungary, 9/10/16. https://doi.org/10.1109/SMC.2016.7844284
@inproceedings{39cfd653c37e4b2589d1c18a83d98d06,
title = "Restricted Boltzmann Machines in Sensory Motor Rhythm Brain-Computer Interfacing: A Study on Inter-Subject Transfer and Co-Adaptation",
abstract = "Naive users often perceive calibration of a Sensory Motor Rhythm (SMR) based Brain-Computer Interfaces (BCI) as tedious and lengthy. The lack of feedback during training is assumed to be a major cause. I.e. if one had already a reasonable model to start with, feedback training could be started immediately. One concept to address this issue is learning a general model and adapting it to new observations. In this study we applied this concept by utilizing a generative model entitled Restricted Boltzmann Machine (RBM). We investigated its feature extraction capabilities by fitting a RBM to recordings of 9 subjects. Generalization was assessed in an online coadaptive study, covering 12 volunteers (10 naive). An overall median accuracy of 88.9{\%} (83.5{\%} naive) with a standard-error of 6.5{\%} (6.6{\%} naive) was achieved for a classical hand versus feet motor imagery task. The online co-adaptive training itself lasted approximately 25 minutes. Feedback was already presented after a one minute setup run, whose purpose was to estimate initial statistics and to train an online artifact detection system.",
author = "Reinmar Kobler and Reinhold Scherer",
year = "2016",
doi = "10.1109/SMC.2016.7844284",
language = "English",
pages = "000469 -- 000474",
booktitle = "Systems, Man, and Cybernetics (SMC), 2016 IEEE International Conference on",

}

TY - GEN

T1 - Restricted Boltzmann Machines in Sensory Motor Rhythm Brain-Computer Interfacing: A Study on Inter-Subject Transfer and Co-Adaptation

AU - Kobler, Reinmar

AU - Scherer, Reinhold

PY - 2016

Y1 - 2016

N2 - Naive users often perceive calibration of a Sensory Motor Rhythm (SMR) based Brain-Computer Interfaces (BCI) as tedious and lengthy. The lack of feedback during training is assumed to be a major cause. I.e. if one had already a reasonable model to start with, feedback training could be started immediately. One concept to address this issue is learning a general model and adapting it to new observations. In this study we applied this concept by utilizing a generative model entitled Restricted Boltzmann Machine (RBM). We investigated its feature extraction capabilities by fitting a RBM to recordings of 9 subjects. Generalization was assessed in an online coadaptive study, covering 12 volunteers (10 naive). An overall median accuracy of 88.9% (83.5% naive) with a standard-error of 6.5% (6.6% naive) was achieved for a classical hand versus feet motor imagery task. The online co-adaptive training itself lasted approximately 25 minutes. Feedback was already presented after a one minute setup run, whose purpose was to estimate initial statistics and to train an online artifact detection system.

AB - Naive users often perceive calibration of a Sensory Motor Rhythm (SMR) based Brain-Computer Interfaces (BCI) as tedious and lengthy. The lack of feedback during training is assumed to be a major cause. I.e. if one had already a reasonable model to start with, feedback training could be started immediately. One concept to address this issue is learning a general model and adapting it to new observations. In this study we applied this concept by utilizing a generative model entitled Restricted Boltzmann Machine (RBM). We investigated its feature extraction capabilities by fitting a RBM to recordings of 9 subjects. Generalization was assessed in an online coadaptive study, covering 12 volunteers (10 naive). An overall median accuracy of 88.9% (83.5% naive) with a standard-error of 6.5% (6.6% naive) was achieved for a classical hand versus feet motor imagery task. The online co-adaptive training itself lasted approximately 25 minutes. Feedback was already presented after a one minute setup run, whose purpose was to estimate initial statistics and to train an online artifact detection system.

U2 - 10.1109/SMC.2016.7844284

DO - 10.1109/SMC.2016.7844284

M3 - Conference contribution

SP - 469

EP - 474

BT - Systems, Man, and Cybernetics (SMC), 2016 IEEE International Conference on

ER -