Evaluation of Riemannian ASR on cEEGrid data: an artifact correction method for BCIs

Artifact Subspace Reconstruction (ASR) is an online artifact correction method for multichannel data recordings. We recently proposed an adaptation of the ASR method using Riemannian geometry (rASR) and found that rASR outperforms ASR in terms of artifact correction capabilities and speed. In our previous investigation we focused on event-related potentials and eye-blink artifacts, in this analysis we extend our recent work by evaluating the performance of rASR in preserving a well-studied oscillation (alpha) while correcting muscle artifacts captured in the EEG data recorded with sparse electrode coverage. We use EEG data recorded with behind-the-ear electrodes (cEEGrid) and find that rASR preserves the power of the alpha frequency band while correcting the muscle artifacts in the EEG data successfully.