Full characterization and removal of motion artifacts from scalp EEG recordings

Abstract

Non-invasive scalp electroencephalogram (EEG) measurements allow for the development of biomedical devices that can be controlled via Brain-Machine Interface (BMI) systems. There are various applications of such systems for scientific, diagnostic, therapeutic, or restorative purposes. However, EEG recordings are often considered as prone to physiological and non-physiological artifacts of different types and frequency characteristics. Motion related non-physiological artifacts can be considered as one of the major contaminants of EEG recordings. Motion artifacts manifest themselves especially for mobile EEG recordings (i.e., Mobi applications) due to the movement of one or several EEG sensors, which may be time-locked with the actual motion that the subjects execute. Artifacts with these characteristics can hinder the true performance of BMI applications, especially when real-time mobile applications are considered (i.e., wearable robotic systems and exoskeletons). Although there several published research efforts to investigate the motion artifacts, there is currently no consensus on the exact characteristics and suitable real-time and/or offline removal methodologies of such artifacts.