Empirical Mode Decomposition and k-Nearest Neighbors Combination for Removing Ocular Artifact in EEG Signals

Electroencephalography (EEG) is a commonly used technique for assessing brain activity through electrodes placed on the scalp. EEG signals hold significant information often utilized for diagnosing neurological disorders and interpreting psychological symptoms. However, EEG signals are frequently interfered with ocular artifacts (OAs), which can lead to potential misinterpretations since OA frequently occurs during the recording process. Thus, OA is very crucial to be removed or minimized the OA effect. This study proposes a novel method for automatically removing OAs by combining Empirical Mode Decomposition (EMD) with classification methods such as k-NN, SVM, and k-means. Those methods were designed to reduce the OA effect and obtain the EEG signal without omitting the information outside of OA area. The reduction of ocular artifacts (OAs) involved several steps: preprocessing, feature extraction, classification, de noising, and signal reconstruction to achieve a clean EEG signal. The proposed methods were evaluated for their effectiveness in reducing OAs by using Root Mean Square Error (RMSE) and Signal-to-Noise Ratio (SNR). The results indicate that the EMD-kNN method performs best, achieving an RMSE of 0.0211 and an SNR of 0.1950. Hence, this study can be concluded that EMD-kNN can perform better noise cancellation and minimal error compared to EMD-SVM and EMD-K-Means. In the future, research could investigate the integration of these methods with real-time EEG monitoring systems.