Cluster decomposing and multi-objective optimization based-ensemble learning framework for motor imagery-based...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Motor imagery (MI) is crucial for electroencephalogram (EEG)-based brain-computer interfaces (BCIs).
• EEG signals present non-stationary, non-linear characteristics and are susceptible to artifacts, posing classification challenges.
• Existing classifiers often struggle with generalization in MI-based BCIs.

Purpose of the Study:

• To develop a robust and generalizable classifier for MI-based BCIs.
• To address the challenges of EEG signal non-stationarity and artifact sensitivity.
• To improve the overall performance of MI-based BCI systems.

Main Methods:

• Proposed a cluster decomposing based ensemble learning framework (CDECL) for EEG classification.
• EEG data was decomposed into sub-datasets using clustering for diverse classifier training.
• Ensemble learning was optimized using a stochastic fractal based binary multi-objective fruit fly optimization algorithm.

Main Results:

• The CDECL framework effectively constructed a diverse ensemble classifier.
• Superior classification performance was demonstrated on two public EEG datasets (BCI Competition IV datasets IIb and IIa).
• The proposed method outperformed several competing classification techniques.

Conclusions:

• The CDECL framework offers a powerful approach for EEG classification in MI-based BCIs.
• This method shows significant promise for enhancing the performance and reliability of BCI systems.
• The findings contribute to advancing the field of brain-computer interfaces.