A flexible analytic wavelet transform based approach for motor-imagery tasks classification in BCI applications.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Motor Imagery (MI) based Brain-Computer Interfaces (BCIs) offer a vital communication channel for individuals with disabilities.
• Electroencephalogram (EEG) is a non-invasive method suitable for capturing MI tasks in BCI systems.
• Reliable assessment of MI tasks is crucial for effective BCI system performance.

Purpose of the Study:

• To propose a novel approach for classifying distinct MI tasks using EEG signals.
• To evaluate the efficacy of Flexible Analytic Wavelet Transform (FAWT) for feature extraction in MI-BCI.
• To identify the most effective classifier for distinguishing between right-hand (RH) and right-foot (RF) MI tasks.

Main Methods:

• EEG signals were analyzed using the Flexible Analytic Wavelet Transform (FAWT) to decompose them into sub-bands.
• Temporal moment-based features were extracted from these sub-bands.
• Feature normalization was applied, and FAWT-based features were used with multiple classifiers, including an ensemble Subspace k-Nearest Neighbour (kNN) classifier.

Main Results:

• The ensemble Subspace kNN classifier demonstrated robust performance in distinguishing RH and RF MI tasks.
• The fourth sub-band achieved the highest accuracy (99.33%), sensitivity (99%), specificity (99.6%), F1-Score (0.9925), and kappa value (0.9865).
• Significant results were also observed for other sub-bands using the subspace kNN classifier.

Conclusions:

• FAWT-based features are effective for classifying EEG signals in RH and RF MI tasks.
• The study highlights the effectiveness of ensemble classifiers for MI-task classification.
• The proposed method outperforms existing state-of-the-art techniques, showing potential for BCI applications like controlling external devices.