Automated Epilepsy Detection System Based On Tertiary Wavelet Model(TWM) Techniques

A computerized method for classifying epilepsy from electroencephalogram (EEG) data using a Tertiary Wavelet Model (TWM) as a feature extractor is detailed. Before extracting features such as PFD, HFD, and Approximate Entropy, the feature set is standardized. Then, the initial information is divided into five sub bands—gamma, beta, alpha, theta, and delta—in the suggested method for EEG information. We then use the Kruskal-Walli’s test to identify the traits that vary substantially across the various groups. Afterwards, the notable characteristics are fed into well-known machine learning algorithms as XGBoost, DT, KNN, SVM, and RF. The outcomes show that the Random Forest classifier (with a standard Sen of $\mathbf{9 9. 7 5 7 5 \%}$, Spe of $\mathbf{9 9. 9 1 \%}$, and Acc of $\mathbf{9 9. 1} \%$ in melanoma detection) is the best classification. By classifying the signals into seizure and non-seizure categories, the Bonn University EEG Databases are utilized provides the basis for the system’s overall evaluation. According to the results shown in the picture below, the suggested model outperforms the prior approaches. In epilepsy, this demonstrates that TWM improves both the calculation speed and accuracy of the discriminant in seizure detection.