Feature extraction and classification for EEG signals using wavelet transform and machine learning techniques.

Area of Science:

• Neuroscience
• Signal Processing
• Machine Learning

Background:

• Electroencephalogram (EEG) signal analysis is crucial for understanding brain activity.
• Classifying EEG signals during cognitive tasks presents a significant challenge.
• Feature extraction is a key step in improving EEG classification accuracy.

Purpose of the Study:

• To develop and validate a discrete wavelet transform (DWT)-based feature extraction scheme for EEG signal classification.
• To differentiate EEG signals recorded during a complex cognitive task (Raven's test) from those recorded during rest (eyes open).

Main Methods:

• Applied DWT to EEG signals to extract relative wavelet energy features from approximation (A4) and detailed (D4) coefficients.
• Utilized four classifiers (Support Vector Machine, Multi-layer Perceptron, K-Nearest Neighbor) to classify the extracted features.
• Evaluated classifier performance using accuracy, sensitivity, specificity, and precision.

Main Results:

• Achieved classification accuracy exceeding 98% with Support Vector Machine, Multi-layer Perceptron, and K-Nearest Neighbor classifiers.
• The A4 and D4 coefficients, corresponding to frequency ranges of 0.53-3.06 Hz and 3.06-6.12 Hz respectively, were effective for classification.
• Demonstrated high performance in distinguishing cognitive task EEG from rest EEG.

Conclusions:

• The proposed DWT-based feature extraction method is effective for classifying EEG signals.
• This approach shows significant potential for accurately identifying EEG patterns associated with complex cognitive tasks.