EEG feature descriptors and discriminant analysis under Riemannian Manifold perspective

Chuong H. Nguyen, Panagiotis Artemiadis

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

This paper presents a framework to classify motor imagery in the context of multi-class Brain Computer Interface based on electroencephalography (EEG). Covariance matrices are extracted as the EEG signal descriptors, and different dissimilarity metrics on the manifold of Symmetric Positive Definite (SPD) matrices are investigated to classify these covariance descriptors. Specifically, we compare the performance of the Log Euclidean distance, Stein divergence, Kullback-Leibler divergence and Von Neumann divergence. Furthermore, inspired from the conventional Common Spatial Pattern, discriminant analysis performed directly on the SPD manifold using different mentioned metrics are proposed to improve the classification accuracy. We also propose a new feature, namely Heterogeneous Orders Relevance Composition (HORC), by combining different relevance matrices, such as Covariance, Mutual Information or Kernel Matrix under the Tensor Framework and Multiple Kernel fusion. Multi-Class Multi-Kernel Relevance Vector Machine is adopted to provide a sparse classifier and Bayesian confidence prediction. Finally, we compare the performance of total 16 methods on the dataset IIa of the BCI Competition IV. The results shows that the mentioned dissimilarity metrics perform quite equally on the original manifold, whereas the proposed discrimination methods can improve the accuracy by 3-5% on the reduced dimension manifold.

Original languageEnglish (US)
JournalNeurocomputing
DOIs
StateAccepted/In press - 2017

Fingerprint

Discriminant Analysis
Discriminant analysis
Electroencephalography
Brain-Computer Interfaces
Bayes Theorem
Imagery (Psychotherapy)
Brain computer interface
Covariance matrix
Tensors
Classifiers
Fusion reactions
Chemical analysis
Datasets

Keywords

  • Brain Computer Interface (BCI)
  • Discriminant analysis
  • Electroencephalography (EEG)
  • Multi-class
  • Multi-kernel
  • Relevance Vector Machine
  • Riemannian Manifold

ASJC Scopus subject areas

  • Computer Science Applications
  • Cognitive Neuroscience
  • Artificial Intelligence

Cite this

EEG feature descriptors and discriminant analysis under Riemannian Manifold perspective. / Nguyen, Chuong H.; Artemiadis, Panagiotis.

In: Neurocomputing, 2017.

Research output: Contribution to journalArticle

@article{cc711091c6544da8acea9f79a7146af3,
title = "EEG feature descriptors and discriminant analysis under Riemannian Manifold perspective",
abstract = "This paper presents a framework to classify motor imagery in the context of multi-class Brain Computer Interface based on electroencephalography (EEG). Covariance matrices are extracted as the EEG signal descriptors, and different dissimilarity metrics on the manifold of Symmetric Positive Definite (SPD) matrices are investigated to classify these covariance descriptors. Specifically, we compare the performance of the Log Euclidean distance, Stein divergence, Kullback-Leibler divergence and Von Neumann divergence. Furthermore, inspired from the conventional Common Spatial Pattern, discriminant analysis performed directly on the SPD manifold using different mentioned metrics are proposed to improve the classification accuracy. We also propose a new feature, namely Heterogeneous Orders Relevance Composition (HORC), by combining different relevance matrices, such as Covariance, Mutual Information or Kernel Matrix under the Tensor Framework and Multiple Kernel fusion. Multi-Class Multi-Kernel Relevance Vector Machine is adopted to provide a sparse classifier and Bayesian confidence prediction. Finally, we compare the performance of total 16 methods on the dataset IIa of the BCI Competition IV. The results shows that the mentioned dissimilarity metrics perform quite equally on the original manifold, whereas the proposed discrimination methods can improve the accuracy by 3-5{\%} on the reduced dimension manifold.",
keywords = "Brain Computer Interface (BCI), Discriminant analysis, Electroencephalography (EEG), Multi-class, Multi-kernel, Relevance Vector Machine, Riemannian Manifold",
author = "Nguyen, {Chuong H.} and Panagiotis Artemiadis",
year = "2017",
doi = "10.1016/j.neucom.2017.10.013",
language = "English (US)",
journal = "Neurocomputing",
issn = "0925-2312",
publisher = "Elsevier",

}

TY - JOUR

T1 - EEG feature descriptors and discriminant analysis under Riemannian Manifold perspective

AU - Nguyen, Chuong H.

AU - Artemiadis, Panagiotis

PY - 2017

Y1 - 2017

N2 - This paper presents a framework to classify motor imagery in the context of multi-class Brain Computer Interface based on electroencephalography (EEG). Covariance matrices are extracted as the EEG signal descriptors, and different dissimilarity metrics on the manifold of Symmetric Positive Definite (SPD) matrices are investigated to classify these covariance descriptors. Specifically, we compare the performance of the Log Euclidean distance, Stein divergence, Kullback-Leibler divergence and Von Neumann divergence. Furthermore, inspired from the conventional Common Spatial Pattern, discriminant analysis performed directly on the SPD manifold using different mentioned metrics are proposed to improve the classification accuracy. We also propose a new feature, namely Heterogeneous Orders Relevance Composition (HORC), by combining different relevance matrices, such as Covariance, Mutual Information or Kernel Matrix under the Tensor Framework and Multiple Kernel fusion. Multi-Class Multi-Kernel Relevance Vector Machine is adopted to provide a sparse classifier and Bayesian confidence prediction. Finally, we compare the performance of total 16 methods on the dataset IIa of the BCI Competition IV. The results shows that the mentioned dissimilarity metrics perform quite equally on the original manifold, whereas the proposed discrimination methods can improve the accuracy by 3-5% on the reduced dimension manifold.

AB - This paper presents a framework to classify motor imagery in the context of multi-class Brain Computer Interface based on electroencephalography (EEG). Covariance matrices are extracted as the EEG signal descriptors, and different dissimilarity metrics on the manifold of Symmetric Positive Definite (SPD) matrices are investigated to classify these covariance descriptors. Specifically, we compare the performance of the Log Euclidean distance, Stein divergence, Kullback-Leibler divergence and Von Neumann divergence. Furthermore, inspired from the conventional Common Spatial Pattern, discriminant analysis performed directly on the SPD manifold using different mentioned metrics are proposed to improve the classification accuracy. We also propose a new feature, namely Heterogeneous Orders Relevance Composition (HORC), by combining different relevance matrices, such as Covariance, Mutual Information or Kernel Matrix under the Tensor Framework and Multiple Kernel fusion. Multi-Class Multi-Kernel Relevance Vector Machine is adopted to provide a sparse classifier and Bayesian confidence prediction. Finally, we compare the performance of total 16 methods on the dataset IIa of the BCI Competition IV. The results shows that the mentioned dissimilarity metrics perform quite equally on the original manifold, whereas the proposed discrimination methods can improve the accuracy by 3-5% on the reduced dimension manifold.

KW - Brain Computer Interface (BCI)

KW - Discriminant analysis

KW - Electroencephalography (EEG)

KW - Multi-class

KW - Multi-kernel

KW - Relevance Vector Machine

KW - Riemannian Manifold

UR - http://www.scopus.com/inward/record.url?scp=85033593249&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85033593249&partnerID=8YFLogxK

U2 - 10.1016/j.neucom.2017.10.013

DO - 10.1016/j.neucom.2017.10.013

M3 - Article

AN - SCOPUS:85033593249

JO - Neurocomputing

JF - Neurocomputing

SN - 0925-2312

ER -