Multiclass EEG signal classification utilizing Rényi min-entropy-based feature selection from wavelet packet...

Area of Science:

• Neuroscience
• Signal Processing
• Machine Learning

Background:

• Electro-encephalogram (EEG) signal analysis is crucial for brain-computer interface (BCI) development.
• Wavelet Packet Transformation (WPT) is commonly used for feature extraction from EEG data.
• Effective feature selection is vital for accurate classification, especially in multi-class BCI problems.

Purpose of the Study:

• To propose and evaluate a novel feature selection method based on Rényi min-entropy for multi-class BCI.
• To compare the performance of the proposed method against conventional techniques like Shannon entropy and mutual information.
• To enhance the efficiency and accuracy of EEG signal classification in BCI systems.

Main Methods:

• Utilized the BCI competition-IV dataset featuring 4-class motor imagery EEG signals.
• Preprocessed EEG data and extracted features using Wavelet Packet Transformation (WPT).
• Applied Rényi min-entropy, Shannon entropy, and mutual information for feature selection.
• Classified the selected features using various machine learning algorithms.

Main Results:

• The proposed Rényi min-entropy-based feature selection method demonstrated superior performance compared to conventional methods.
• The approach achieved higher classification accuracy for multi-class motor imagery EEG signals.
• The study confirmed the effectiveness of Rényi min-entropy in selecting discriminative features for BCI.

Conclusions:

• The Rényi min-entropy-based algorithm offers a more efficient and accurate feature selection strategy for multi-class BCI.
• This novel method advances the field of BCI by improving EEG signal classification capabilities.
• The findings suggest a promising direction for developing advanced BCI applications.