A new evolutionary preprocessing approach for classification of mental arithmetic based EEG signals.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Brain-computer interface (BCI) systems translate electroencephalogram (EEG) signals into commands for assistive devices.
• Current BCI systems face challenges in accuracy due to noisy EEG signals.
• Improved EEG signal processing is crucial for effective BCI applications.

Purpose of the Study:

• To enhance the accuracy of BCI systems by developing a novel evolutionary approach for EEG signal preprocessing.
• To introduce a fusion-based preprocessing method inspired by chromosomal crossover.
• To evaluate the efficacy of the proposed method in classifying mental arithmetic (MA) based EEG signals.

Main Methods:

• Applied a novel fusion-based preprocessing method to an open-access EEG dataset from 29 subjects.
• Extracted features using an autoregressive model.
• Classified EEG signals using a k-nearest neighbor classifier.
• Compared the proposed method against the common average reference (CAR) method and no preprocessing.

Main Results:

• Achieved classification accuracy (CA) ranging from 67.57% to 99.70% for binary MA-based EEG signals.
• Attained an average CA of 88.71%, with 93.10% of subjects showing performance improvement.
• Demonstrated a 3.91% and 2.75% higher CA compared to CAR and no preprocessing, respectively.

Conclusions:

• The proposed fusion-based evolutionary preprocessing approach significantly improves EEG signal classification accuracy in BCI systems.
• This method shows great potential for enhancing the performance of BCIs, particularly for tasks involving mental arithmetic.
• The evolutionary strategy offers a promising direction for developing more robust and accurate BCI applications.