Motor Imagery Classification Based on Local Log Riemannian Distance Matrices Selected by Confusion Area Score

Jinhyo Shin, Wonzoo Chung

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

Abstract

In this paper, we propose a novel motor imagery (MI) classification method in electroencephalogram (EEG)-based brain-computer interface (BCI) using local log Riemannian distance matrices (LRDM). The proposed method selects optimal local LRDM based on confusion area score which is designed to minimize the overlap between the class-dependent distributions of Riemannian distance between local covariance matrices that are generated from adjacent (local) channels centered on each channel. A feature vector is formed by concatenating vectorized selected local LRDM and used as input to support vector machine (SVM) in order to classify motor imagery. The performance of the proposed method is evaluated using BCI Competition III dataset IVa and BCI Competition IV dataset I. The results confirm performance improvement of the proposed method compared to existing MI classification methods.

Original languageEnglish
Title of host publicationPattern Recognition - 6th Asian Conference, ACPR 2021, Revised Selected Papers
EditorsChristian Wallraven, Qingshan Liu, Hajime Nagahara
PublisherSpringer Science and Business Media Deutschland GmbH
Pages159-169
Number of pages11
ISBN (Print)9783031023743
DOIs
Publication statusPublished - 2022
Event6th Asian Conference on Pattern Recognition, ACPR 2021 - Virtual, Online
Duration: 2021 Nov 92021 Nov 12

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume13188 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference6th Asian Conference on Pattern Recognition, ACPR 2021
CityVirtual, Online
Period21/11/921/11/12

Keywords

  • Brain-computer interface (BCI)
  • Electroencephalogram (EEG)
  • Local feature
  • Log Riemannian distance matrix
  • Motor imagery (MI)

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint

Dive into the research topics of 'Motor Imagery Classification Based on Local Log Riemannian Distance Matrices Selected by Confusion Area Score'. Together they form a unique fingerprint.

Cite this