Motor-Imagery-Based Brain-Computer Interface Using Signal Derivation and Aggregation Functions.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Brain-computer interface (BCI) systems enable communication between the brain and external devices.
• Motor imagery (MI) is a key BCI approach, often utilizing electroencephalography (EEG) for noninvasive brain dynamics measurement.
• Current BCI performance is limited by challenges in EEG pattern recognition, including channel selection, signal-to-noise ratio, and information redundancy.

Purpose of the Study:

• To introduce a novel enhanced fusion framework for motor imagery-based BCI systems.
• To improve the accuracy and performance of existing BCI frameworks.
• To explore advanced signal processing and classifier fusion techniques for EEG data.

Main Methods:

• Implemented a time-invariant EEG signal differentiation preprocessing step.
• Incorporated the sensorimotor rhythm band as an additional frequency feature.
• Utilized up to six diverse classifiers and various aggregation functions (including Choquet/Sugeno integrals and overlap functions) for decision making.

Main Results:

• The enhanced fusion framework achieved 88.80% accuracy on a dataset of 20 volunteers.
• An optimized version of the system reached up to 90.76% accuracy.
• Choquet/Sugeno integrals and overlap functions demonstrated superior performance in classifier fusion.

Conclusions:

• The proposed enhanced fusion framework significantly advances MI-based BCI performance.
• Advanced signal processing and sophisticated classifier fusion methods are crucial for improving BCI accuracy.
• The study highlights the effectiveness of specific integral and overlap functions for robust BCI decision-making.