Enhancing P300 based character recognition performance using a combination of ensemble classifiers and a fuzzy fusion...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Brain-computer interfaces (BCIs) enable communication via brain signals, bypassing muscle activity.
• P300 spellers are common BCI applications known for simplicity and reliability.
• Improving P300 speller performance is crucial for practical usability.

Purpose of the Study:

• To propose a novel multi-feature subset fuzzy fusion (MSFF) framework for P300 spellers.
• To enhance the recognition of user spelling intention in brain-computer interfaces.
• To improve the practical usability of P300-based BCIs.

Main Methods:

• Feature selection using the Lasso algorithm and subsequent feature division.
• Construction of ensemble Linear Discriminant Analysis (LDA) classifiers.
• Fuzzy fusion of ensemble classifiers for intention recognition.

Main Results:

• Achieved 100% accuracy on BCI Competition II Dataset IIb after 4 epochs.
• Attained 96% accuracy on BCI Competition III dataset II.
• Reached 98.3% accuracy on the BNCI Horizon Dataset.
• Demonstrated effective utilization of temporal signal information.

Conclusions:

• The proposed MSFF method significantly improves P300-based BCI performance.
• MSFF offers superior or comparable results to existing machine learning algorithms.
• Enhanced classification performance increases the practical utility of P300 spellers.