Recent Study of Artifact Elimination in EEG Signals

Abstract

The progress of electroencephalography (EEG) has promoted this technology to be widely used in various fields such as computer science, medical engineering, and signal processing because of its non-invasiveness and low cost. However, the quality of EEG recordings may be degraded due to the introduction of artifacts, which has a non-negligible negative impact on subsequent operations. Artifacts are unwanted signals that vary in their source, such as muscular, ocular, and cardiac ones. Many studies have proposed methods to remove artifacts from EEG signals, from the classic ones including filtering, empirical mode decomposition, and blind source separation to the recent frequent introduction of neural networks. This work reviews the highly influential publications in the past 5 years to update the latest headways. Based on the survey of essays and papers, we notice hybrid methods with the greatest proportion. Besides, not only do we discover the development and advancement of the typical methods of signal processing mentioned, but we also find that machine learning techniques are becoming more prevailing and effective for artifact removal. Moreover, we observe that recent works have proposed integrated architectures and blockchain solutions to address the problem globally, regardless of the type of artifact.