A Neural Learning-Based Approach for EOG Artifact Removal from EEG

Abstract

In this paper, we proposed a novel methodology that combines a long short-term memory (LSTM)-based neural network with independent component analysis (ICA) to address the challenge of removing electrooculogram (EOG) artifacts from contaminated electroencephalogram (EEG) signals. Our approach achieved two primary objectives: 1) to estimate the horizontal and vertical EOG signals from the contaminated EEG data, and 2) to utilize ICA to eliminate the estimated EOG signals from the EEG, thereby generating an artifact-free EEG signal. To evaluate the effectiveness of our proposed method, we conducted experiments on a publicly available dataset consisting of recordings from 27 participants. We utilized established metrics such as mean squared error, mean absolute error, and mean error to assess the quality of our artifact removal technique. Furthermore, we compared the performance of our approach with two state-of-the-art deep learning-based methods reported in the literature, illustrating the superior performance of our proposed methodology.