Manifold Learning-Based Common Spatial Pattern for EEG Signal Classification.

Area of Science:

• * Neuroscience
• * Machine Learning
• * Signal Processing

Background:

• * Riemannian manifolds offer powerful tools for electroencephalogram (EEG) signal classification.
• * High computational costs of Riemannian metrics hinder applications with high-dimensional data.
• * Existing methods struggle with efficiency and performance in complex EEG analysis.

Purpose of the Study:

• * To develop an efficient ensemble method (MLCSP-TSE-MLP) for EEG signal classification.
• * To reduce computational complexity while maintaining or improving classification accuracy.
• * To address the challenges posed by high-dimensional EEG features in Riemannian-based approaches.

Main Methods:

• * Proposed MLCSP-TSE-MLP ensemble classifier.
• * Riemannian graph embedding for low-dimensional manifold learning (MLCSP).
• * Tangent space mapping using Euclidean mean for computational efficiency (TSE).
• * Multilayer Perceptron (MLP) for final classification.

Main Results:

• * MLCSP-TSE-MLP demonstrated superior classification performance across three datasets.
• * Achieved significant improvements in training speed and reduced test time compared to traditional Riemannian methods.
• * The MLCSP-TSE module effectively enhances discrimination and reduces computational load.

Conclusions:

• * MLCSP-TSE-MLP is an effective and efficient method for high-dimensional EEG data classification.
• * The proposed approach offers a powerful tool for practical applications in neuroscience and beyond.
• * This method overcomes the computational limitations of traditional Riemannian techniques.