Automatic Classification and Artifacts Removal of EEG Signal Using Supervised and Unsupervised Model

The well-known neurological disorder (epilepsy) is recurrent seizures. It can profoundly affect a person's life and well-being, as well as society as a whole. Many different detection methods are employed to determine whether a person has epilepsy or not. One of the often-employed methods is the examination of electroencephalography (EEG) signals. The EEG is important in epilepsy for a variety of reasons. However, some of the EEG signal aberrations can behave as noise and reduce the predictability of the process. To find the optimal method for artifact removal, this project attempts to build supervised and unsupervised algorithms and compare their performances. For this purpose, a dataset consisting of EEG signals of both normal humans and people who are affected by epilepsy is collected This dataset is then preprocessed using two artifact removal techniques called the Independent Component Analysis (ICA) and Principal Component Analysis (PCA). The features that can be useful during the process of epilepsy detection are extracted using a feature extraction method called the Fourier Transform. The features are then classified using two different techniques. The feature classification techniques used in this study are the quadratic discriminant analysis and the Random Forest (RF) algorithm. The models are then trained and tested using the collected EEG signals. The results of the testing are then compared to find the best combination of the feature classification technique and the artifact removal technique. It is found that the combination of the RF algorithm and the ICA is the best algorithm with an accuracy of over 97%. The second-best combination is the RF algorithm and the PCA algorithm. This combination can be deployed as a backend processor of an epilepsy detection application. As it has an inbuilt, artifact removal it is expected to provide flawless predictions in the future.