Understanding sensorimotor neural dynamics during human locomotion is a core neuroscience objective, yet there are fundamental challenges to the collection of high-fidelity neuroelectric signals during motion. Mobile, real-world data recordings introduce significant electrical noise that prevents the collection of high-fidelity electrical brain activity during dynamic movements. Our aim was to investigate the effects of novel hardware and signal processing methods for collecting electroencephalography (EEG) brain signals, using an electrical testbed that mimicked the human head. A 60-channel high-density array of tripolar concentric ring electrodes and conventional disk electrodes were used on an electrical testbed to recover simulated brain activity in the presence of electrical muscle artifacts, and canonical correlation analysis was applied to both datasets to test additional noise removal.

An electrical head phantom device was constructed using ballistics gelatin material and anatomically relevant broadcast antenna to accurately model human brain activity and material properties of the human head during simulated locomotion. Simulated brainwave activity consisted of random, time-varying sinusoidal bursts with unique frequency content from within known electrocortical spectral bands. Electrical muscle activity was captured from the neck muscles of a walking human subject and broadcast into the head phantom.

Muscle activity was introduced at varying amplitudes to compare neural source signal recovery from tripolar concentric ring electrodes and conventional disk electrodes, before and after canonical correlation analysis for myoelectric noise removal. By measuring spectral power changes and peaks that matched the ground truth input signals, we identified improved myoelectric artifact removal in tripolar concentric ring electrode recordings compared to conventional electrodes, and additional improvements after canonical correlation analysis signal processing.