Classification of imaginary movements in ECoG with a hybrid approach based on multi-dimensional Hilbert-SVM solution

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-computer interfaces (BCIs) offer potential for assistive technologies.
• Electrocorticographic (ECoG) signals provide a rich source of neural data for BCIs.
• Accurate classification of motor imagery is crucial for effective BCI control.

Purpose of the Study:

• To investigate the classification of ECoG signals for motor imagery.
• To identify reliable features for distinguishing between different imagined movements.
• To develop and evaluate a novel BCI approach using entropy and machine learning.

Main Methods:

• Channel selection using Tsallis entropy in the Hilbert domain.
• Nonlinear classification of motor imagery using Support Vector Machines (SVMs).
• Integration of Hilbert-based and statistical/entropy measurements with SVMs.

Main Results:

• Accurate classification of imagined left small-finger and tongue movements.
• Achieved 95% accuracy on the training set (264/278).
• Achieved 73% accuracy on the testing set (73/100).

Conclusions:

• Invariant phase transition features are reliable for classifying ECoG signals.
• The proposed method combining entropy and SVMs is effective for ECoG-based BCIs.
• This approach supports the use of classification techniques in BCI development.