DSA-MFNet: Deep-shallow Attention based Multi-frame Fusion Network for EEG motor imagery classification

2021 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), Год журнала: 2023, Номер 16, С. 2021 - 2026

Опубликована: Дек. 5, 2023

One of the most prevalent brain-computer interface (BCI) paradigms is Electroencephalogram (EEG) motor imagery (MI). It has found extensive applications in numerous fields. While there have been significant strides achieving high MI classification performance, certain challenges still persist:Effective utilization time-varying spatial and temporal features from multi-channel brain signals remains elusive. Fully leveraging interactive information embedded within finite-length MI-EEG samples an open question. In this study, we introduce Deep-Shallow Attention-Based Multi-Frame Fusion Network (DSA-MFNet) tailored for EEG-based classification. The architecture DSA-MFNet encompasses both a Attention (DSA) module (MF) module. detail, DSA integrates deep-shallow convolution block, which extracts intricate deep spatial-temporal surface-level features. Subsequently, attention block emphasizes salient data. Meanwhile, MF delves into interactions amongst multiple frames data, shedding light on unique characteristics EEG signals.Our model sets new benchmark by outperforming leading techniques, accuracy 86.6% BCI Competition IV-2a dataset subject-dependent modes. For sake transparency to foster further research, will be making our code trained models available GitHub.

Язык: Английский

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

Опубликована: Июль 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.

Язык: Английский

Hierarchical Window Attention for Motor Imagery EEG Classification

Опубликована: Май 27, 2024

Язык: Английский