Mixed-Norm Based Broad Learning System for EEG Classification.

Area of Science:

• Neuroscience
• Machine Learning
• Signal Processing

Background:

• Designing robust classifiers with strong generalization is crucial for Brain-Computer Interface (BCI) research.
• Electroencephalography (EEG) signals are often contaminated by artifacts, posing challenges for accurate classification.
• Existing classifiers may struggle with the inherent noise in EEG data.

Purpose of the Study:

• To propose a novel classifier with enhanced generalization and noise resilience for BCI applications.
• To adapt the Broad Learning System (BLS) by incorporating a mixed-norm optimization model.
• To evaluate the performance of the proposed classifier on noisy EEG datasets.

Main Methods:

• A new classifier based on the Broad Learning System (BLS) architecture was developed.
• The l2-norm optimization in BLS was replaced with a mixed-norm based model.
• The Augmented Lagrange Multiplier (ALM) method was employed for efficient model optimization.
• Experiments were conducted on two publicly available EEG datasets.

Main Results:

• The proposed mixed-norm based classifier demonstrated strong generalization capabilities.
• The classifier exhibited robust performance across various noise environments by adjusting a mixed parameter.
• Experimental results confirmed the desirable performance of the new method on EEG signal classification.
• The proposed classifier outperformed the standard BLS and other existing methods in handling noisy EEG data.

Conclusions:

• The proposed mixed-norm based classifier offers a more reasonable choice for classifying noisy EEG signals in BCI.
• The flexible parameter setting allows for effective performance maintenance in diverse noise conditions.
• This approach advances the development of powerful and robust classifiers for BCI research.