Independent Vector Analysis for Feature Extraction in Motor Imagery Classification.

Area of Science:

• Neuroscience
• Signal Processing
• Machine Learning

Background:

• Brain-computer interfaces (BCI) rely on electroencephalogram (EEG) signals for motor imagery classification.
• Current methods often use single datasets, limiting performance in multi-source scenarios.
• Independent Component Analysis (ICA) is a related technique for signal separation.

Purpose of the Study:

• To propose and evaluate a novel feature extraction method using Independent Vector Analysis (IVA) for multi-dataset EEG motor imagery classification.
• To enhance the accuracy and robustness of BCI systems by leveraging statistical dependencies across datasets.
• To investigate the effectiveness of IVA-derived features with both traditional and deep learning classifiers.

Main Methods:

• Independent Vector Analysis (IVA) was applied to extract features from multiple EEG datasets.
• Extracted IVA components served as input for Support Vector Machines (SVM), K-Nearest Neighbors (KNN), EEGNet, and EEGInception classifiers.
• Motor imagery classification performance was evaluated using the selected classifiers.

Main Results:

• The proposed IVA-based feature extraction method demonstrated improved performance in classifying EEG motor imagery.
• The approach showed promising results in clustering patients within motor imagery-based BCI.
• An average classification accuracy of 86.7% was achieved using the IVA features.

Conclusions:

• Independent Vector Analysis (IVA) offers a powerful approach for feature extraction in multi-dataset EEG-based BCI.
• Leveraging inter-dataset dependencies through IVA enhances motor imagery classification accuracy.
• The proposed method shows potential for improving the clinical application of BCI technology.