A comparison between different Multiclass Common Spatial Pattern approaches for identification of motor imagery tasks - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Poster Communications Year : 2014

A comparison between different Multiclass Common Spatial Pattern approaches for identification of motor imagery tasks

Cecilia Lindig-León
  • Function : Author
  • PersonId : 962167
Laurent Bougrain

Abstract

Common Spatial Patterns (CSP) is a feature extraction method suited for two-class problems. However, there are some alternatives to apply it for multiclass tasks by using a group of ensemble classifiers that divide the problem into different binary classification tasks, from which the final decision is inferred as the combination of their responses. Nevertheless, there is another approach to extend CSP for multiple classes to a one-step-method by approximating the joint diagonalization of their covariance matrices. In this study, in order to identify whether CSP by Joint Approximate Diagonalization (JAD) represents an outperforming alternative to the standard mulMclass CSP alternatives, four different methods were applied on dataset 2a used in BCI competition IV.

Domains

Automatic
Fichier principal
Vignette du fichier
lindig-BCI14poster-CSPcomparison.pdf (323.71 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-01092272 , version 1 (08-12-2014)

Identifiers

  • HAL Id : hal-01092272 , version 1

Cite

Cecilia Lindig-León, Laurent Bougrain. A comparison between different Multiclass Common Spatial Pattern approaches for identification of motor imagery tasks. 6th International Brain-computer interface conference 2014, Sep 2014, Graz, Austria. 2014. ⟨hal-01092272⟩
396 View
176 Download

Share

Gmail Facebook Twitter LinkedIn More