An Automatic Subject Specific Intrinsic Mode Function Selection for Enhancing Two-Class EEG based Motor Imagery-Brain Computer Interface

Pramod Gaur, R. B. Pachori, H. Wang, Girijesh Prasad

Research output: Contribution to journalArticle

Abstract

The electroencephalogram (EEG) signals tend to have poor time-frequency localization when analysis techniques involve a fixed set of basis functions such as in short-time Fourier transform (STFT) and wavelet transform (WT). These signals also exhibit highly non-stationary characteristics and suffer from low signal-to-noise ratio (SNR). As a result, there is often poor task detection accuracy and high error rates in designed braincomputer interfacing (BCI) systems. In this paper, a novel preprocessing method is proposed to automatically reconstruct the EEG signal by selecting the intrinsic mode functions (IMFs) based on a median frequency measure. Multivariate empirical mode decomposition (MEMD) is used to decompose the EEG signals into a set of IMFs. The reconstructed EEG signal has high SNR and contains only information correlated to a specific motor imagery task. The common spatial pattern (CSP) is used to extract features from the reconstructed EEG signals. The linear discriminant analysis (LDA) and support vector machine (SVM) have been utilized in order to classify the features into left hand motor imagery (LHMI) and right hand motor imagery (RHMI) tasks. Our experimental results on the BCI competition IV dataset 2A show that the proposed method with fifteen channels outperforms bandpass filtering with twenty-two channels (> 1%) and by > 9 % (p = 0:0078) with raw EEG signals, > 13% (p = 0:0039) with empirical mode decomposition (EMD) based filtering and > 17 % (p = 0:0039) with discrete wavelet transform (DWT) based filtering.
LanguageEnglish
Number of pages10
JournalIEEE Sensors Journal
Early online date23 Apr 2019
DOIs
Publication statusE-pub ahead of print - 23 Apr 2019

Fingerprint

electroencephalography
Brain computer interface
Electroencephalography
imagery
brain
Signal to noise ratio
wavelet analysis
signal to noise ratios
Decomposition
decomposition
Discrete wavelet transforms
Discriminant analysis
Wavelet transforms
preprocessing
Support vector machines
Fourier transforms

Keywords

  • BCI
  • filtering
  • common spatial pattern ,
  • linear discriminant analysis
  • MEMD

Cite this

@article{b6813394683b4aef975c741db7402714,
title = "An Automatic Subject Specific Intrinsic Mode Function Selection for Enhancing Two-Class EEG based Motor Imagery-Brain Computer Interface",
abstract = "The electroencephalogram (EEG) signals tend to have poor time-frequency localization when analysis techniques involve a fixed set of basis functions such as in short-time Fourier transform (STFT) and wavelet transform (WT). These signals also exhibit highly non-stationary characteristics and suffer from low signal-to-noise ratio (SNR). As a result, there is often poor task detection accuracy and high error rates in designed braincomputer interfacing (BCI) systems. In this paper, a novel preprocessing method is proposed to automatically reconstruct the EEG signal by selecting the intrinsic mode functions (IMFs) based on a median frequency measure. Multivariate empirical mode decomposition (MEMD) is used to decompose the EEG signals into a set of IMFs. The reconstructed EEG signal has high SNR and contains only information correlated to a specific motor imagery task. The common spatial pattern (CSP) is used to extract features from the reconstructed EEG signals. The linear discriminant analysis (LDA) and support vector machine (SVM) have been utilized in order to classify the features into left hand motor imagery (LHMI) and right hand motor imagery (RHMI) tasks. Our experimental results on the BCI competition IV dataset 2A show that the proposed method with fifteen channels outperforms bandpass filtering with twenty-two channels (> 1{\%}) and by > 9 {\%} (p = 0:0078) with raw EEG signals, > 13{\%} (p = 0:0039) with empirical mode decomposition (EMD) based filtering and > 17 {\%} (p = 0:0039) with discrete wavelet transform (DWT) based filtering.",
keywords = "BCI, filtering, common spatial pattern ,, linear discriminant analysis, MEMD",
author = "Pramod Gaur and Pachori, {R. B.} and H. Wang and Girijesh Prasad",
year = "2019",
month = "4",
day = "23",
doi = "10.1109/JSEN.2019.2912790",
language = "English",

}

TY - JOUR

T1 - An Automatic Subject Specific Intrinsic Mode Function Selection for Enhancing Two-Class EEG based Motor Imagery-Brain Computer Interface

AU - Gaur, Pramod

AU - Pachori, R. B.

AU - Wang, H.

AU - Prasad, Girijesh

PY - 2019/4/23

Y1 - 2019/4/23

N2 - The electroencephalogram (EEG) signals tend to have poor time-frequency localization when analysis techniques involve a fixed set of basis functions such as in short-time Fourier transform (STFT) and wavelet transform (WT). These signals also exhibit highly non-stationary characteristics and suffer from low signal-to-noise ratio (SNR). As a result, there is often poor task detection accuracy and high error rates in designed braincomputer interfacing (BCI) systems. In this paper, a novel preprocessing method is proposed to automatically reconstruct the EEG signal by selecting the intrinsic mode functions (IMFs) based on a median frequency measure. Multivariate empirical mode decomposition (MEMD) is used to decompose the EEG signals into a set of IMFs. The reconstructed EEG signal has high SNR and contains only information correlated to a specific motor imagery task. The common spatial pattern (CSP) is used to extract features from the reconstructed EEG signals. The linear discriminant analysis (LDA) and support vector machine (SVM) have been utilized in order to classify the features into left hand motor imagery (LHMI) and right hand motor imagery (RHMI) tasks. Our experimental results on the BCI competition IV dataset 2A show that the proposed method with fifteen channels outperforms bandpass filtering with twenty-two channels (> 1%) and by > 9 % (p = 0:0078) with raw EEG signals, > 13% (p = 0:0039) with empirical mode decomposition (EMD) based filtering and > 17 % (p = 0:0039) with discrete wavelet transform (DWT) based filtering.

AB - The electroencephalogram (EEG) signals tend to have poor time-frequency localization when analysis techniques involve a fixed set of basis functions such as in short-time Fourier transform (STFT) and wavelet transform (WT). These signals also exhibit highly non-stationary characteristics and suffer from low signal-to-noise ratio (SNR). As a result, there is often poor task detection accuracy and high error rates in designed braincomputer interfacing (BCI) systems. In this paper, a novel preprocessing method is proposed to automatically reconstruct the EEG signal by selecting the intrinsic mode functions (IMFs) based on a median frequency measure. Multivariate empirical mode decomposition (MEMD) is used to decompose the EEG signals into a set of IMFs. The reconstructed EEG signal has high SNR and contains only information correlated to a specific motor imagery task. The common spatial pattern (CSP) is used to extract features from the reconstructed EEG signals. The linear discriminant analysis (LDA) and support vector machine (SVM) have been utilized in order to classify the features into left hand motor imagery (LHMI) and right hand motor imagery (RHMI) tasks. Our experimental results on the BCI competition IV dataset 2A show that the proposed method with fifteen channels outperforms bandpass filtering with twenty-two channels (> 1%) and by > 9 % (p = 0:0078) with raw EEG signals, > 13% (p = 0:0039) with empirical mode decomposition (EMD) based filtering and > 17 % (p = 0:0039) with discrete wavelet transform (DWT) based filtering.

KW - BCI

KW - filtering

KW - common spatial pattern ,

KW - linear discriminant analysis

KW - MEMD

UR - http://10.1109/JSEN.2019.2912790

UR - https://ieeexplore.ieee.org/document/8695803

U2 - 10.1109/JSEN.2019.2912790

DO - 10.1109/JSEN.2019.2912790

M3 - Article

ER -