Search

Motion and muscle artifacts can undermine signal quality in electroencephalography (EEG) recordings during locomotion. We evaluated approaches for recovering ground-truth artificial brain signals from noisy EEG recordings. We built an electrical head phantom that broadcast four brain and four muscle sources. Head movements were generated by a robotic motion platform. We recorded 128-channel dual layer EEG and 8-channel neck electromyography (EMG) from the head phantom during motion. We evaluated ground-truth electrocortical source signal recovery from artifact contaminated data using Independent Component Analysis (ICA) to determine: (1) the number of isolated noise sensor recordings needed to capture and remove motion artifacts, (2) the ability of Artifact Subspace Reconstruction to remove motion and muscle artifacts at contrasting artifact detection thresholds, (3) the number of neck EMG sensor recordings needed to capture and remove muscle artifacts, and (4) the ability of Canonical Correlation Analysis to remove muscle artifacts. We also evaluated source signal recovery by combining the best practices identified in aims 1-4. By including isolated noise and EMG recordings in the ICA decomposition, we more effectively recovered ground-truth artificial brain signals. A reduced subset of 32-noise and 6-EMG channels showed equivalent performance compared to including the complete arrays. Artifact Subspace Reconstruction improved source separation, but this was contingent on muscle activity amplitude. Canonical Correlation Analysis also improved source separation. Merging noise and EMG recordings into the ICA decomposition, with Artifact Subspace Reconstruction and Canonical Correlation Analysis preprocessing, improved source signal recovery. This study expands on previous head phantom experiments by including neck muscle source activity and evaluating artificial electrocortical spectral power fluctuations synchronized with gait events.

Electroencephalography (EEG) headsets are the most commonly used sensing devices for brain-computer interface. In real-world applications, there are advantages to extrapolating data from one user session to another. However, these advantages are limited if the data arise from different hardware systems, which often vary between application spaces. Currently, this creates a need to recalibrate classifiers, which negatively affects people's interest in using such systems. In this paper, we employ active weighted adaptation regularization (AwAR), which integrates weighted adaptation regularization (wAR) and active learning, to expedite the calibration process. wAR makes use of labeled data from the previous headset and handles class-imbalance, and active learning selects the most informative samples from the new headset to label. Experiments on single-trial event-related potential classification show that AwAR can significantly increase the classification accuracy, given the same number of labeled samples from the new headset. In other words, AwAR can effectively reduce the number of labeled samples required from the new headset, given a desired classification accuracy, suggesting value in collating data for use in wide scale transfer-learning applications.