Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Spatial-Temporal Mamba Network for EEG-based Motor Imagery Classification

Xiaoxiao Yang, Ziyu Jia

TL;DR

Experimental results on BCI Competition IV 2a and 2b datasets demonstrate STMambaNet's superiority over existing models, establishing it as a powerful tool for advancing MI-based BCIs and improving real-world BCI systems.

Abstract

Motor imagery (MI) classification is key for brain-computer interfaces (BCIs). Until recent years, numerous models had been proposed, ranging from classical algorithms like Common Spatial Pattern (CSP) to deep learning models such as convolutional neural networks (CNNs) and transformers. However, these models have shown limitations in areas such as generalizability, contextuality and scalability when it comes to effectively extracting the complex spatial-temporal information inherent in electroencephalography (EEG) signals. To address these limitations, we introduce Spatial-Temporal Mamba Network (STMambaNet), an innovative model leveraging the Mamba state space architecture, which excels in processing extended sequences with linear scalability. By incorporating spatial and temporal Mamba encoders, STMambaNet effectively captures the intricate dynamics in both space and time, significantly enhancing the decoding performance of EEG signals for MI classification. Experimental results on BCI Competition IV 2a and 2b datasets demonstrate STMambaNet's superiority over existing models, establishing it as a powerful tool for advancing MI-based BCIs and improving real-world BCI systems.

Spatial-Temporal Mamba Network for EEG-based Motor Imagery Classification

TL;DR

Experimental results on BCI Competition IV 2a and 2b datasets demonstrate STMambaNet's superiority over existing models, establishing it as a powerful tool for advancing MI-based BCIs and improving real-world BCI systems.

Abstract

Motor imagery (MI) classification is key for brain-computer interfaces (BCIs). Until recent years, numerous models had been proposed, ranging from classical algorithms like Common Spatial Pattern (CSP) to deep learning models such as convolutional neural networks (CNNs) and transformers. However, these models have shown limitations in areas such as generalizability, contextuality and scalability when it comes to effectively extracting the complex spatial-temporal information inherent in electroencephalography (EEG) signals. To address these limitations, we introduce Spatial-Temporal Mamba Network (STMambaNet), an innovative model leveraging the Mamba state space architecture, which excels in processing extended sequences with linear scalability. By incorporating spatial and temporal Mamba encoders, STMambaNet effectively captures the intricate dynamics in both space and time, significantly enhancing the decoding performance of EEG signals for MI classification. Experimental results on BCI Competition IV 2a and 2b datasets demonstrate STMambaNet's superiority over existing models, establishing it as a powerful tool for advancing MI-based BCIs and improving real-world BCI systems.
Paper Structure (22 sections, 12 equations, 3 figures, 2 tables)

This paper contains 22 sections, 12 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Methodology
  4. Embedding Component
  5. Mamba Encoders
  6. Mamba Rationale.
  7. Encoder Architecture.
  8. Spatial-Temporal Decoding.
  9. Classifier
  10. Experiment and Analysis
  11. Datasets
  12. BCI Competition IV 2a.
  13. BCI Competition IV 2b.
  14. Experimental Setup and Baseline Models
  15. ConvNets schirrmeisterDeepLearningConvolutional2018.
  16. ...and 7 more sections

Figures (3)

  • Figure 1: The provided image shows the overall architecture of STMambaNet, with each component enclosed in light grey boxes for better visual organization.
  • Figure 2: The structure of Mamba encoder, comprising two residual layers. The variance-pooled features are processed first by the Mamba encoder, followed by the average-pooled features.
  • Figure 3: Ablation study results for STMambaNet on the BCI Competition IV 2a dataset: (a) the subject-wise accuracy across nine subjects (S01 to S09) for different model configurations, including the full STMambaNet model, Temporal Mamba, Spatial Mamba, and a baseline model with no Mamba components; (b) the average accuracy for each model. The full STMambaNet model achieves the highest average accuracy, as marked by a gray downward-pointing triangle.