Mean shrinkage improves the classification of ERP signals by exploiting additional label information

Johannes Hohne, Benjamin Blankertz, Klaus Muller, Daniel Bartz

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

3 Citations (Scopus)

Abstract

Linear discriminant analysis (LDA) is the most commonly used classification method for single trial data in a brain-computer interface (BCI) framework. The popularity of LDA arises from its robustness, simplicity and high accuracy. However, the standard LDA approach is not capable to exploit sublabel information (such as stimulus identity), which is accessible in data from event related potentials (ERPs): it assumes that the evoked potentials are independent of the stimulus identity and dependent only on the users' attentional state. We question this assumption and investigate several methods which extract subclass-specific features from ERP data. Moreover, we propose a novel classification approach which exploits subclass-specific features using mean shrinkage. Based on a reanalysis of two BCI data sets, we show that our novel approach outperforms the standard LDA approach, while being computationally highly efficient.

Original languageEnglish
Title of host publicationProceedings - 2014 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2014
PublisherIEEE Computer Society
ISBN (Print)9781479941506
DOIs
Publication statusPublished - 2014 Jan 1
Event4th International Workshop on Pattern Recognition in Neuroimaging, PRNI 2014 - Tubingen, Germany
Duration: 2014 Jun 42014 Jun 6

Other

Other4th International Workshop on Pattern Recognition in Neuroimaging, PRNI 2014
CountryGermany
CityTubingen
Period14/6/414/6/6

Fingerprint

Discriminant analysis
Labels
Brain computer interface
Bioelectric potentials

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition
  • Biomedical Engineering

Cite this

Hohne, J., Blankertz, B., Muller, K., & Bartz, D. (2014). Mean shrinkage improves the classification of ERP signals by exploiting additional label information. In Proceedings - 2014 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2014 [6858523] IEEE Computer Society. https://doi.org/10.1109/PRNI.2014.6858523

Mean shrinkage improves the classification of ERP signals by exploiting additional label information. / Hohne, Johannes; Blankertz, Benjamin; Muller, Klaus; Bartz, Daniel.

Proceedings - 2014 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2014. IEEE Computer Society, 2014. 6858523.

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

Hohne, J, Blankertz, B, Muller, K & Bartz, D 2014, Mean shrinkage improves the classification of ERP signals by exploiting additional label information. in Proceedings - 2014 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2014., 6858523, IEEE Computer Society, 4th International Workshop on Pattern Recognition in Neuroimaging, PRNI 2014, Tubingen, Germany, 14/6/4. https://doi.org/10.1109/PRNI.2014.6858523
Hohne J, Blankertz B, Muller K, Bartz D. Mean shrinkage improves the classification of ERP signals by exploiting additional label information. In Proceedings - 2014 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2014. IEEE Computer Society. 2014. 6858523 https://doi.org/10.1109/PRNI.2014.6858523
Hohne, Johannes ; Blankertz, Benjamin ; Muller, Klaus ; Bartz, Daniel. / Mean shrinkage improves the classification of ERP signals by exploiting additional label information. Proceedings - 2014 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2014. IEEE Computer Society, 2014.
@inproceedings{ae246d64756a44958be115567c1ee334,
title = "Mean shrinkage improves the classification of ERP signals by exploiting additional label information",
abstract = "Linear discriminant analysis (LDA) is the most commonly used classification method for single trial data in a brain-computer interface (BCI) framework. The popularity of LDA arises from its robustness, simplicity and high accuracy. However, the standard LDA approach is not capable to exploit sublabel information (such as stimulus identity), which is accessible in data from event related potentials (ERPs): it assumes that the evoked potentials are independent of the stimulus identity and dependent only on the users' attentional state. We question this assumption and investigate several methods which extract subclass-specific features from ERP data. Moreover, we propose a novel classification approach which exploits subclass-specific features using mean shrinkage. Based on a reanalysis of two BCI data sets, we show that our novel approach outperforms the standard LDA approach, while being computationally highly efficient.",
author = "Johannes Hohne and Benjamin Blankertz and Klaus Muller and Daniel Bartz",
year = "2014",
month = "1",
day = "1",
doi = "10.1109/PRNI.2014.6858523",
language = "English",
isbn = "9781479941506",
booktitle = "Proceedings - 2014 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2014",
publisher = "IEEE Computer Society",

}

TY - GEN

T1 - Mean shrinkage improves the classification of ERP signals by exploiting additional label information

AU - Hohne, Johannes

AU - Blankertz, Benjamin

AU - Muller, Klaus

AU - Bartz, Daniel

PY - 2014/1/1

Y1 - 2014/1/1

N2 - Linear discriminant analysis (LDA) is the most commonly used classification method for single trial data in a brain-computer interface (BCI) framework. The popularity of LDA arises from its robustness, simplicity and high accuracy. However, the standard LDA approach is not capable to exploit sublabel information (such as stimulus identity), which is accessible in data from event related potentials (ERPs): it assumes that the evoked potentials are independent of the stimulus identity and dependent only on the users' attentional state. We question this assumption and investigate several methods which extract subclass-specific features from ERP data. Moreover, we propose a novel classification approach which exploits subclass-specific features using mean shrinkage. Based on a reanalysis of two BCI data sets, we show that our novel approach outperforms the standard LDA approach, while being computationally highly efficient.

AB - Linear discriminant analysis (LDA) is the most commonly used classification method for single trial data in a brain-computer interface (BCI) framework. The popularity of LDA arises from its robustness, simplicity and high accuracy. However, the standard LDA approach is not capable to exploit sublabel information (such as stimulus identity), which is accessible in data from event related potentials (ERPs): it assumes that the evoked potentials are independent of the stimulus identity and dependent only on the users' attentional state. We question this assumption and investigate several methods which extract subclass-specific features from ERP data. Moreover, we propose a novel classification approach which exploits subclass-specific features using mean shrinkage. Based on a reanalysis of two BCI data sets, we show that our novel approach outperforms the standard LDA approach, while being computationally highly efficient.

UR - http://www.scopus.com/inward/record.url?scp=84905727960&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84905727960&partnerID=8YFLogxK

U2 - 10.1109/PRNI.2014.6858523

DO - 10.1109/PRNI.2014.6858523

M3 - Conference contribution

SN - 9781479941506

BT - Proceedings - 2014 International Workshop on Pattern Recognition in Neuroimaging, PRNI 2014

PB - IEEE Computer Society

ER -