Using Discriminative Lasso to Detect a Graph Fourier Transform (GFT) Subspace for robust decoding in Motor Imagery...

Area of Science:

• Neuroscience
• Signal Processing
• Machine Learning

Background:

• Motor imagery (MI) brain-computer interfaces (BCIs) decode user intentions from brain activity.
• Current decoding schemes for MI-BCIs face challenges in accurately interpreting complex EEG signals.
• Graph-based signal processing offers a novel approach to analyze functional relationships in neural data.

Purpose of the Study:

• To introduce a novel decoding scheme for motor imagery brain-computer interfaces (BCIs) based on the graph Fourier transform (GFT) concept.
• To leverage functional covariations in EEG signals, represented as graphs, for improved BCI performance.
• To define an information-rich GFT subspace using discriminative Lasso (dLasso) for enhanced feature extraction.

Main Methods:

• EEG data from motor imagery tasks were treated as signals on a sensor array graph.
• A graph representing functional covariations during specific imagined movements was constructed from training data.
• Graph-guided decomposition and dLasso were employed to define a discriminative GFT subspace.
• Multichannel EEG signals were transformed into features using matrix operations for classification.

Main Results:

• The proposed GFT-based decoding scheme was evaluated on two independent datasets.
• The novel method demonstrated favorable performance compared to existing popular BCI decoding alternatives.
• The approach effectively translates complex EEG patterns into discriminative features for movement intention classification.

Conclusions:

• The GFT-based decoding scheme offers a promising advancement for motor imagery brain-computer interfaces.
• Representing EEG as graph signals and utilizing GFT provides a powerful framework for BCI signal processing.
• This method enhances the accuracy and efficiency of decoding brain activity for BCI applications.