EEG Signals Feature Extraction Based on DWT and EMD Combined with Approximate Entropy

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Accurate classification of motor imagery is crucial for advancing brain-computer interface (BCI) performance.
• Existing methods face challenges in effectively processing electroencephalogram (EEG) signals for motor imagery recognition.

Purpose of the Study:

• To propose a new feature extraction technique to improve the classification accuracy of motor imagery in EEG signals.
• To address limitations in signal processing, specifically wide frequency band coverage during empirical mode decomposition (EMD).

Main Methods:

• Electroencephalogram (EEG) signals were decomposed using discrete wavelet transform (DWT) into narrow band signals.
• Sub-band signals were further processed with empirical mode decomposition (EMD) to obtain intrinsic mode functions (IMFs).
• Approximate entropy was calculated on reconstructed signals using selected IMFs, serving as feature vectors for support vector machine (SVM) classification.

Main Results:

• The proposed method effectively extracts features from EEG signals for motor imagery classification.
• The combined DWT-EMD approach with approximate entropy demonstrated improved classification accuracy compared to standard methods.
• The technique successfully managed wide frequency band coverage issues inherent in EMD.

Conclusions:

• The novel feature extraction method significantly enhances the classification accuracy of motor imagery in BCI applications.
• This approach offers a more robust way to process complex EEG signals, paving the way for more reliable BCIs.
• The integration of DWT, EMD, and approximate entropy presents a promising direction for future BCI research and development.