Decision tree structure based classification of EEG signals recorded during two dimensional cursor movement imagery

Area of Science:

• Biomedical Engineering
• Neuroscience
• Computer Science

Background:

• Electroencephalography (EEG) signals for brain-computer interfaces (BCIs) are non-stationary and prone to noise and artifacts.
• Improving the speed and accuracy of BCI systems is crucial for effective human-computer communication.

Purpose of the Study:

• To propose a fast and accurate decision tree structure-based classification method for EEG data.
• To classify EEG data corresponding to up/down/right/left computer cursor movement imagery.

Main Methods:

• A novel decision tree structure-based classification approach was developed.
• EEG data from three healthy subjects (ages 24-29) were acquired across two sessions on different days.
• The method automatically generates subject-specific decision tree structures (DTS).

Main Results:

• The proposed method achieved classification accuracies of 55.92%, 57.90%, and 82.24% on test data from the three subjects.
• Demonstrated a 12.25% improvement over the best results from closely related studies.

Conclusions:

• The decision tree structure-based method offers a flexible and accurate approach for EEG classification in BCI systems.
• The method requires only a subject's training set and can be easily updated.
• Significant accuracy improvements were observed despite data collected across different sessions.