Mobile EEG artifact correction on limited hardware using artifact subspace recon- struction

TL;DR

This paper presents an implementation of artifact subspace reconstruction (ASR) for mobile EEG devices with limited hardware, enabling real-time artifact correction for portable EEG applications outside laboratory settings.

Contribution

It introduces a portable, online ASR implementation optimized for limited hardware like single-board computers, expanding real-time EEG artifact correction capabilities.

Findings

Successful translation and deployment of ASR on limited hardware

Validation with publicly available EEG datasets

Facilitates real-time EEG analysis outside labs

Abstract

Biological data like electroencephalography (EEG) are typically contaminated by unwanted signals, called artifacts. Therefore, many applications dealing with biological data with low signal-to-noise ratio require robust artifact correction. For some applications like brain-computer-interfaces (BCI), the artifact correction needs to be real-time capable. Artifact subspace reconstruction (ASR) is a statistical method for artifact reduction in EEG. However, in its current implementation, ASR cannot be used in mobile data recordings using limited hardware easily. In this report, we add to the growing field of portable, online signal processing methods by describing an implementation of ASR for limited hardware like single-board computers. We describe the architecture, the process of translating and compiling a Matlab codebase for a research platform, and a set of validation tests using publicly available data sets. The implementation of ASR on limited, portable hardware facilitates the online interpretation of EEG signals acquired outside of the laboratory environment.