A Biologically Inspired Approach to Frequency Domain Feature Extraction for EEG Classification

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Electroencephalogram (EEG) signal classification is crucial for mental decoding in brain-computer interfaces (BCI).
• Existing methods often require multi-channel EEG or complex feature extraction.
• Improving single-channel EEG classification performance is essential for practical BCI applications.

Purpose of the Study:

• To introduce and evaluate a novel feature extraction approach for single-channel EEG signal classification.
• To enhance the performance of brain-computer interfaces (BCI) in decoding various mental tasks.
• To identify optimal mental task pairs for binary classification in BCI systems.

Main Methods:

• A biologically inspired feature extraction method based on frequency domain analysis of power spectral densities (PSDs) was developed.
• The method was applied to a dataset of six subjects performing five distinct mental tasks (resting state, mental arithmetic, left hand imagination, right hand imagination, letter "A" imagination).
• Linear Discriminant Analysis (LDA) and Support Vector Machines (SVM) were used for pairwise and multiclass EEG signal classification.

Main Results:

• The proposed method achieved high classification accuracies: 83.06% for binary classification and 91.85% for multiclassification.
• Performance was comparable to state-of-the-art methods while utilizing only single-channel EEG and incurring low computational costs.
• The mental arithmetic versus letter imagination task pair demonstrated the highest binary classification accuracy (90.29%).

Conclusions:

• The developed frequency domain feature extraction method effectively improves single-channel EEG classification for BCI applications.
• The findings suggest that mental arithmetic and letter imagination tasks are highly discriminative for binary BCI.
• This research contributes to the advancement of efficient single-channel BCI systems and task selection for specific applications.