Classifying Ocular and Muscle Artifacts in EEG Signals

For many decades, the electroencephalography has been the widespread technique for capturing brain electrical activity. However, a main limitation in this field is that the EEG signals are contaminated by many artifacts caused by physiological and nonphysiological sources. In this regard detecting such artifacts is an essential step toward a correct interpretation of the brain activity. In this paper, a new approach founded on EEG signal analyses is proposed to automatically identify artifacts. To further develop this research, we propose employing a set of machine learning methods to classify data obtained from EEG signals dataset with artifacts. In relation to this investigation, we suggested a fusion of spectral, statistical, and amplitude features. Eventually, an assessment step is carried out by several assessment metrics such as accuracy, recall/sensitivity, f-score, precision and specificity. The most relevant attained outcomes are the following: an accuracy of 90%, a recall of 90%, an f-score of 92%, a precision of 91%, and a specificity of 95%. This demonstrates that our proposed approach founded on EEG signals can be recommended for detecting artifacts.