EEG channel selection based on sequential backward floating search for motor imagery classification.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-computer interfaces (BCIs) leverage electroencephalogram (EEG) signals for motor rehabilitation.
• Motor imagery (MI) tasks are crucial for BCI functionality.
• EEG channel selection is vital for enhancing MI classification accuracy and reducing data redundancy.

Purpose of the Study:

• To implement Sequential Backward Floating Search (SBFS) for optimal EEG channel selection in MI-BCIs.
• To propose a modified SBFS to reduce computational time for EEG channel selection.
• To evaluate the efficacy of the proposed method on public BCI datasets for left and right hand MI tasks.

Main Methods:

• Sequential Backward Floating Search (SBFS) algorithm applied for EEG channel selection.
• A modified SBFS approach utilizing symmetrical channel pairing to accelerate the selection process.
• Extensive experimental validation on four public BCI datasets.

Main Results:

• SBFS significantly improved MI classification accuracy compared to using all channels or conventional channels (C3, C4, Cz) (p < 0.001).
• The modified SBFS demonstrated superior performance over existing state-of-the-art channel selection methods.
• The proposed method effectively reduces redundant information by selecting informative EEG channels.

Conclusions:

• The modified SBFS offers an efficient and effective strategy for EEG channel selection in MI-BCIs.
• This approach enhances BCI performance, holding promise for improving motor function in individuals with disabilities.
• Optimized channel selection is critical for advancing BCI technology and applications.