Time Window Optimization for Riemannian Geometry-based Motor Imagery EEG Classification DOI

Fanbo Zhuo,

Bo Lv, Fengzhen Tang

et al.

Published: July 15, 2024

The existing Riemannian geometry-based approaches for brain computer interface (BCI) employ fixed time windows. However, the inherent variability and dynamic changes among subjects necessitate robust adaptive solutions window optimization. Recognizing current limitations of classifiers, we propose a selection confidence metric (TWSCM) based on geometry. This operates manifold symmetric positive definite (SPD) matrices, providing theoretically grounded computationally efficient approach optimization process is unsupervised, which able to deal with online scenario without training labels. Experimental results BCI competition IV dataset IIa demonstrate that classification performance significantly improved most subjects. average over six by 7.52%. simulated experiment shows enhanced in comparison baseline experiments Additionally, an in-depth analysis TWSCM provides insights into variations Overall, this paper introduces first method within geometric framework, presenting effective interpretable optimizing windows motor imagery classification, novel promising perspective EEG signal analysis.

Language: Английский

A novel hybrid decoding neural network for EEG signal representation DOI

Youshuo Ji,

Fu Li, Boxun Fu

et al.

Pattern Recognition, Journal Year: 2024, Volume and Issue: 155, P. 110726 - 110726

Published: June 27, 2024

Language: Английский

Citations

5
STIT-Net- A Wavelet based Convolutional Transformer Model for Motor Imagery EEG Signal Classification in the Sensorimotor Bands DOI

S Chrisilla,

R. Shantha Selva Kumari

Clinical EEG and Neuroscience, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 29, 2025

Motor Imagery (MI) electroencephalographic (EEG) signal classification is a pioneer research branch essential for mobility rehabilitation. This paper proposes an end-to-end hybrid deep network “Spatio Temporal Inception Transformer Network (STIT-Net)” model MI classification. Discrete Wavelet Transform (DWT) used to derive the alpha (8–13) Hz and beta (13–30) EEG sub bands which are dominant during motor tasks enhance performance of proposed work. STIT-Net employs spatial temporal convolutions capture dependencies information inception block with three parallel extracts multi-level features. Then transformer encoder self-attention mechanism highlights similar task. The improves Physionet imagery dataset average accuracy 93.52% 95.70% binary class in respectively, 85.26% 87.34% class, four 81.95% 82.66% were obtained band respective based signals better compared results available literature. methodology further evaluated on other datasets, both subject-independent cross-subject conditions, assess model.

Language: Английский

Citations

0
Investigating the impact of feature extraction methods on prediction accuracy of neurological recovery levels in comatose patients post-cardiac arrest DOI
Şenol Çelik,

Semiha Sude Özgüzel,

İsmail Cantürk

et al.

Computer Methods in Biomechanics & Biomedical Engineering, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 16

Published: March 10, 2025

Cardiac arrest can cause irreversible Post-Cardiac Arrest Brain Injury (PCABI), but predicting PCABI with certainty remains challenging. This study aims to improve prognostication by neurological recovery using EEG data from the 'I-CARE: International Research Consortium Database.' Data were preprocessed an FIR Equiripple Bandpass Filter, and three feature extraction methods applied. Decision Tree, KNN, SVM, Ensemble Learning algorithms evaluated F1-Score, Accuracy, ROC-AUC. The highest accuracy, 0.89, was achieved Hamming-windowed streamline Tree after selection.

Language: Английский

Citations

0
ECA-ATCNet: Efficient EEG Decoding with Spike Integrated Transformer Conversion for BCI Applications DOI Creative Commons

Xuhang Li,

Qianzi Shen,

Haitao Wang

et al.

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(4), P. 1894 - 1894

Published: Feb. 12, 2025

The Brain–Computer Interface (BCI) has applications in smart homes and healthcare by converting EEG signals into control commands. However, traditional signal decoding methods are affected individual differences, although deep learning techniques have made significant breakthroughs, challenges such as high energy consumption the processing of raw data remain. This paper introduces Efficient Channel Attention Temporal Convolutional Network (ECA-ATCNet) to enhance feature applying Convolution (ECA-conv) across spatial spectral dimensions. model outperforms state-of-the-art both within-subject between-subject classification tasks on MI-EEG datasets (BCI-2a PhysioNet), achieving accuracies 87.89% 71.88%, respectively. Additionally, proposed Spike Integrated Transformer Conversion (SIT-conversion) method, based Spiking–Softmax, converts Transformer’s self-attention mechanism Spiking Neural Networks (SNNs) just 12 time steps. accuracy loss converted ECA-ATCNet is only 0.6% 0.73%, while its reduced 52.84% 53.52%. SIT-conversion enables ultra-low-latency, near-lossless ANN-to-SNN conversion, with SNNs similar their ANN counterparts image datasets. Inference 18.18% 45.13%. method offers a novel approach for low-power, portable BCI contributes advancement energy-efficient SNN algorithms.

Language: Английский

Citations

0
Graph neural network based on brain inspired forward-forward mechanism for motor imagery classification in brain-computer interfaces DOI Creative Commons
Qiwei Xue, Yuntao Song, Huapeng Wu

et al.

Frontiers in Neuroscience, Journal Year: 2024, Volume and Issue: 18

Published: March 28, 2024

Introduction Within the development of brain-computer interface (BCI) systems, it is crucial to consider impact brain network dynamics and neural signal transmission mechanisms on electroencephalogram-based motor imagery (MI-EEG) tasks. However, conventional deep learning (DL) methods cannot reflect topological relationship among electrodes, thereby hindering effective decoding activity. Methods Inspired by concept neuronal forward-forward (F-F) mechanism, a novel DL framework based Graph Neural Network combined mechanism (F-FGCN) presented. F-FGCN aims enhance EEG performance applying functional relationships propagation mechanism. The fusion process involves converting multi-channel into sequence signals constructing grounded Pearson correlation coeffcient, effectively representing associations between channels. Our model initially pre-trains Convolutional (GCN), fine-tunes output layer obtain feature vector. Moreover, F-F used for advanced extraction classification. Results discussion Achievement assessed PhysioNet dataset four-class categorization, compared with various classical state-of-the-art models. learned features substantially amplify downstream classifiers, achieving highest accuracy 96.11% 82.37% at subject group levels, respectively. Experimental results affirm potency FFGCN in enhancing performance, thus paving way BCI applications.

Language: Английский

Citations

2
An effective classification approach for EEG-based motor imagery tasks combined with attention mechanisms DOI
Jixiang Li, Wuxiang Shi, Yurong Li

et al.

Cognitive Neurodynamics, Journal Year: 2024, Volume and Issue: 18(5), P. 2689 - 2707

Published: May 3, 2024

Language: Английский

Citations

2
Unsupervised and semi-supervised domain adaptation networks considering both global knowledge and prototype-based local class information for Motor Imagery Classification DOI
Dongxue Zhang, Huiying Li, Jingmeng Xie

et al.

Neural Networks, Journal Year: 2024, Volume and Issue: 179, P. 106497 - 106497

Published: July 1, 2024

Language: Английский

Citations

2
Investigating critical brain area for EEG-based binocular color fusion and rivalry with EEGNet DOI Creative Commons

Zhineng Lv,

Xiang Liu,

Mengshi Dai

et al.

Frontiers in Neuroscience, Journal Year: 2024, Volume and Issue: 18

Published: Feb. 27, 2024

Introduction Binocular color fusion and rivalry are two specific phenomena in binocular vision, which could be used as experimental tools to study how the brain processes conflicting information. There is a lack of objective evaluation indexes distinguish or for dichoptic color. Methods This paper introduced EEGNet construct an EEG-based model classification. We developed EEG dataset from 10 subjects. Results By dividing data five different areas train corresponding models, results showed that: (1) area represented by back had large difference on signals, accuracy reached highest 81.98%, more channels decreased performance; (2) there was effect inter-subject variability, recognition still very challenge across subjects; (3) statistics relatively stationary at time same individual, highly reproducible individual. Discussion The critical meaningful developing computer interfaces (BCIs) based color-related visual evoked potential (CVEP).

Language: Английский

Citations

1
EEG classification with limited data: A deep clustering approach DOI
Mohsen Tabejamaat,

Hoda Mohammadzade,

Farhood Negin

et al.

Pattern Recognition, Journal Year: 2024, Volume and Issue: 157, P. 110934 - 110934

Published: Aug. 30, 2024

Language: Английский

Citations

1
Motor Imagery EEG signals classification using a Transformer-GCN approach DOI

A. Hamidi,

Kourosh Kiani

Applied Soft Computing, Journal Year: 2024, Volume and Issue: unknown, P. 112686 - 112686

Published: Dec. 1, 2024

Language: Английский

Citations

1