Development of a wearable EEG system for everyday use: overcoming motion artifacts, electrode positioning, and on-device computing challenges
Abstract:
Neurotechnology, with its convergence with artificial intelligence (AI), is anticipated to bring profound societal impacts beyond the medical realm, with Electroencephalography (EEG) based brain-computer interfaces (BCIs) positioned as key enablers in this field. However, wearable EEG devices face challenges such as susceptibility to motion artifacts, lengthy setup time, the need for assistance, and high power consumption, which restrict their usability outside clinical settings.
In this thesis, we address several of the critical limitations of wearable EEG technology, proposing and demonstrating solutions to make EEG systems truly wearable across varied real-world applications. First, we developed WalkingWizard, a prototype dry-electrode EEG embedded within a baseball cap, designed for use under motion. This flexible PCB-based device contains an embedded layer in the PCB to record motion artifact, enabling high signal-to-noise ratio (SNR) EEG sensing even while in motion. Steady-State Visually Evoked Potential (SSVEP) trials validated its performance, achieving high classification accuracy even at walking speeds of up to 5.0 km/h. We extended this with WalkingWizard Integrated, demonstrating further applications including SSVEP in VR environments, real-time emotional state monitoring, and the study of guided meditation effects on relaxation.
To address electrode placement challenges, we introduce PositionWizard, a self-guided EEG positioning framework leveraging EEG artifacts: motion artifact induced by bone-conduction headphone and jaw-clench artifacts, to guide precise electrode placement without external assistance. Trials confirmed PositionWizard's ability to reduce electrode placement offset and achieve highly reproducible SSVEP classification, enhancing system consistency.
Finally, we present EdgeWizard, a low-power, low-latency EEG system that integrates WalkingWizard technology with PACE, a CGRA-based SoC. Our feasibility study demonstrated the effectiveness of this architecture in reducing power consumption while improving real-time processing capabilities. Together, these developments provide a comprehensive pathway towards more accessible, versatile, and user-friendly wearable EEG systems.