Hi ZR,
To perform classification of epileptic and non-epileptic EEG signals, refer the below steps:
- Load the EEG data
- Preprocess the data by filtering and normalizing the EEG signals to remove noise
- Apply wavelet transform to extract features from the wavelet coefficients. You can also take other statistical measures like mean, variance and entropy from different wavelet bands
- Once the above feature extraction is performed, a machine learning classification model like SVM, k-NN or neural networks can be created to classify the features into epileptic or non-epileptic categories.
- Assess the performance of your classifier using metrics like accuracy, precision, recall and F1-score.
To perform wavelet decomposition, experiments can be performed with the choice of wavelet function (e.g., ‘db4’, ‘sym4’). Other features such as entropy, energy and higher order statistics can also be used for a better classification score. For more information, refer the following resource:
- Time-frequency convolution network for EEG Data classification: https://www.mathworks.com/help/wavelet/ug/time-frequency-convolutional-network-for-eeg-data-classification.html
