Adaptive GCN and Bi-GRU-Based Dual Branch for Motor Imagery EEG Decoding

Sensors, Journal Year: 2025, Volume and Issue: 25(4), P. 1147 - 1147

Published: Feb. 13, 2025

Decoding motor imagery electroencephalography (MI-EEG) signals presents significant challenges due to the difficulty in capturing complex functional connectivity between channels and temporal dependencies of EEG across different periods. These are exacerbated by low spatial resolution high signal redundancy inherent signals, which traditional linear models struggle address. To overcome these issues, we propose a novel dual-branch framework that integrates an adaptive graph convolutional network (Adaptive GCN) bidirectional gated recurrent units (Bi-GRUs) enhance decoding performance MI-EEG effectively modeling both channel correlations dependencies. The Chebyshev Type II filter decomposes into multiple sub-bands giving model frequency domain insights. Adaptive GCN, specifically designed for context, captures more than conventional GCN models, enabling accurate spatial-spectral feature extraction. Furthermore, combining Bi-GRU Multi-Head Attention (MHA) time segments extract deep time-spectral features. Finally, fusion is performed generate final prediction results. Experimental results demonstrate our method achieves average classification accuracy 80.38% on BCI-IV Dataset 2a 87.49% BCI-I 3a, outperforming other state-of-the-art approaches. This approach lays foundation future exploration personalized brain-computer interface (BCI) systems.

Language: Английский

---

Augmenting Brain-Computer Interfaces with ART: An Artifact Removal Transformer for Reconstructing Multichannel EEG Signals

NeuroImage, Journal Year: 2025, Volume and Issue: unknown, P. 121123 - 121123

Published: March 1, 2025

Artifact removal in electroencephalography (EEG) is a longstanding challenge that significantly impacts neuroscientific analysis and brain-computer interface (BCI) performance. Tackling this problem demands advanced algorithms, extensive noisy-clean training data, thorough evaluation strategies. This study presents the Removal Transformer (ART), an innovative EEG denoising model employing transformer architecture to adeptly capture transient millisecond-scale dynamics characteristic of signals. Our approach offers holistic, end-to-end solution simultaneously addresses multiple artifact types multichannel data. We enhanced generation data pairs using independent component analysis, thus fortifying scenarios critical for effective supervised learning. performed comprehensive validations wide range open datasets from various BCI applications, metrics like mean squared error signal-to-noise ratio, as well sophisticated techniques such source localization classification. evaluations confirm ART surpasses other deep-learning-based methods, setting new benchmark signal processing. advancement not only boosts accuracy reliability but also promises catalyze further innovations field, facilitating brain naturalistic environments.

Language: Английский

---

Enhancing Motor Imagery EEG Classification with a Riemannian Geometry-Based Spatial Filtering (RSF) Method

Neural Networks, Journal Year: 2025, Volume and Issue: unknown, P. 107511 - 107511

Published: April 1, 2025

Language: Английский

---

OS-SSVEP: One-shot SSVEP classification

Neural Networks, Journal Year: 2024, Volume and Issue: 180, P. 106734 - 106734

Published: Sept. 25, 2024

Language: Английский

---