Time-Domain Versus Frequency-Embedded EEG Sequences for Sensorimotor BCI Using 1D-CNN

This study proposed a motor imagery (MI) classification pipeline featuring a 1−dimensional convolutional neural network (1D-CNN) with different time/frequency feature representation techniques. The objective was to classify right hand (RH) versus right foot (RF) MI tasks in both intra- an...

Full description

Saved in:
Bibliographic Details
Main Authors: Simanto Saha, Mathias Baumert, Alistair Mcewan
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/11113278/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849236301243482112
author Simanto Saha
Mathias Baumert
Alistair Mcewan
author_facet Simanto Saha
Mathias Baumert
Alistair Mcewan
author_sort Simanto Saha
collection DOAJ
description This study proposed a motor imagery (MI) classification pipeline featuring a 1&#x2212;dimensional convolutional neural network (1D-CNN) with different time/frequency feature representation techniques. The objective was to classify right hand (RH) versus right foot (RF) MI tasks in both intra- and inter-subject (pairwise and pooled) BCI settings using a 1D-CNN architecture trained on time-domain bandpass filtered electroencephalography (EEG) signals, frequency-embedded power spectral density (PSD) and cross-power spectral density (CPSD) sequences. The EEG signals were bandpass filtered with 4 Hz and 32 Hz cut-off frequencies, and PSD/CPSD sequences were estimated in the same frequency range. Thus, the number of input channels for 1D-CNN was N, N or <inline-formula> <tex-math notation="LaTeX">$N\times N$ </tex-math></inline-formula> for EEG signals, PSD or CPSD sequences. We used dataset IVa from BCI Competition III in 5&#x2212;fold cross-validation settings to evaluate intra-subject, inter-subject (pairwise) and inter-subject (pooled) BCI classification accuracies. We compared the performance of the proposed methods with classification algorithms featuring common spatial patterns (CSP) for benchmarking. The best overall classification accuracies (%) for intra-subject, inter-subject (pairwise) and inter-subject (pooled) BCIs were <inline-formula> <tex-math notation="LaTeX">$86.57\pm 11.69$ </tex-math></inline-formula>, <inline-formula> <tex-math notation="LaTeX">$70.80\pm 9.21$ </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">$76.61\pm 12.37$ </tex-math></inline-formula> using 1D&#x2212;CNN with time-domain EEG signals. The average classification accuracies using 1D-CNN with frequency-embedded PSD sequences were <inline-formula> <tex-math notation="LaTeX">$82.57\pm 10.20$ </tex-math></inline-formula>, <inline-formula> <tex-math notation="LaTeX">$69.32\pm 7.46$ </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">$71.32\pm 7.96$ </tex-math></inline-formula> for intra-subject, inter-subject (pairwise) and inter-subject (pooled) BCIs. The proposed time/frequency feature representation techniques with 1D-CNN outperformed CSP-based algorithms (p-value &#x003C;0.05). The comparative results suggest the utility of the proposed methods for MI classification, especially for a fully zero-training inter-subject BCI.
format Article
id doaj-art-d942286b75854a1ba3ac652d70df1e21
institution Kabale University
issn 2169-3536
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-d942286b75854a1ba3ac652d70df1e212025-08-20T04:02:18ZengIEEEIEEE Access2169-35362025-01-011313791913793010.1109/ACCESS.2025.359595311113278Time-Domain Versus Frequency-Embedded EEG Sequences for Sensorimotor BCI Using 1D-CNNSimanto Saha0https://orcid.org/0000-0001-5473-3662Mathias Baumert1https://orcid.org/0000-0003-2984-2167Alistair Mcewan2https://orcid.org/0000-0001-7597-6372School of Biomedical Engineering, The University of Sydney, Sydney, AustraliaSchool of Electrical and Mechanical Engineering, The University of Adelaide, Adelaide, AustraliaSchool of Biomedical Engineering, The University of Sydney, Sydney, AustraliaThis study proposed a motor imagery (MI) classification pipeline featuring a 1&#x2212;dimensional convolutional neural network (1D-CNN) with different time/frequency feature representation techniques. The objective was to classify right hand (RH) versus right foot (RF) MI tasks in both intra- and inter-subject (pairwise and pooled) BCI settings using a 1D-CNN architecture trained on time-domain bandpass filtered electroencephalography (EEG) signals, frequency-embedded power spectral density (PSD) and cross-power spectral density (CPSD) sequences. The EEG signals were bandpass filtered with 4 Hz and 32 Hz cut-off frequencies, and PSD/CPSD sequences were estimated in the same frequency range. Thus, the number of input channels for 1D-CNN was N, N or <inline-formula> <tex-math notation="LaTeX">$N\times N$ </tex-math></inline-formula> for EEG signals, PSD or CPSD sequences. We used dataset IVa from BCI Competition III in 5&#x2212;fold cross-validation settings to evaluate intra-subject, inter-subject (pairwise) and inter-subject (pooled) BCI classification accuracies. We compared the performance of the proposed methods with classification algorithms featuring common spatial patterns (CSP) for benchmarking. The best overall classification accuracies (%) for intra-subject, inter-subject (pairwise) and inter-subject (pooled) BCIs were <inline-formula> <tex-math notation="LaTeX">$86.57\pm 11.69$ </tex-math></inline-formula>, <inline-formula> <tex-math notation="LaTeX">$70.80\pm 9.21$ </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">$76.61\pm 12.37$ </tex-math></inline-formula> using 1D&#x2212;CNN with time-domain EEG signals. The average classification accuracies using 1D-CNN with frequency-embedded PSD sequences were <inline-formula> <tex-math notation="LaTeX">$82.57\pm 10.20$ </tex-math></inline-formula>, <inline-formula> <tex-math notation="LaTeX">$69.32\pm 7.46$ </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">$71.32\pm 7.96$ </tex-math></inline-formula> for intra-subject, inter-subject (pairwise) and inter-subject (pooled) BCIs. The proposed time/frequency feature representation techniques with 1D-CNN outperformed CSP-based algorithms (p-value &#x003C;0.05). The comparative results suggest the utility of the proposed methods for MI classification, especially for a fully zero-training inter-subject BCI.https://ieeexplore.ieee.org/document/11113278/Brain–computer interfaceelectroencephalographymotor imagerypower spectral density1-dimensional convolutional neural network
spellingShingle Simanto Saha
Mathias Baumert
Alistair Mcewan
Time-Domain Versus Frequency-Embedded EEG Sequences for Sensorimotor BCI Using 1D-CNN
IEEE Access
Brain–computer interface
electroencephalography
motor imagery
power spectral density
1-dimensional convolutional neural network
title Time-Domain Versus Frequency-Embedded EEG Sequences for Sensorimotor BCI Using 1D-CNN
title_full Time-Domain Versus Frequency-Embedded EEG Sequences for Sensorimotor BCI Using 1D-CNN
title_fullStr Time-Domain Versus Frequency-Embedded EEG Sequences for Sensorimotor BCI Using 1D-CNN
title_full_unstemmed Time-Domain Versus Frequency-Embedded EEG Sequences for Sensorimotor BCI Using 1D-CNN
title_short Time-Domain Versus Frequency-Embedded EEG Sequences for Sensorimotor BCI Using 1D-CNN
title_sort time domain versus frequency embedded eeg sequences for sensorimotor bci using 1d cnn
topic Brain–computer interface
electroencephalography
motor imagery
power spectral density
1-dimensional convolutional neural network
url https://ieeexplore.ieee.org/document/11113278/
work_keys_str_mv AT simantosaha timedomainversusfrequencyembeddedeegsequencesforsensorimotorbciusing1dcnn
AT mathiasbaumert timedomainversusfrequencyembeddedeegsequencesforsensorimotorbciusing1dcnn
AT alistairmcewan timedomainversusfrequencyembeddedeegsequencesforsensorimotorbciusing1dcnn