Source-linked AI summary
A Dry-Contact Ear-EEG System With Continuous Electrode-Skin Impedance Mismatch Monitoring for Motion Artifact Cancellation Using DRL Stimulus
Lohan Atapattu, Sajitha Madugalle, Imasha Nethmal, Erandee Jayathilaka, Avishka Herath, Kithmin Wickremasinghe, Simon L. Kappel, Nilan Udayanga, Chamira U. S. Edussooriya
TL;DR
Motion-induced electrode-skin impedance mismatches degrade dry-contact Ear-EEG during natural movement, while conventional artifact-removal methods may be computationally costly or require multiple channels, large datasets, or offline processing. The paper uses a DRL-injected 1 kHz stimulus to monitor impedance mismatch and drive online adaptive cancellation, reducing artifact power while preserving observable alpha modulation.
Problem
Motion-induced electrode-skin impedance mismatches can severely degrade dry-contact Ear-EEG, while many existing artifact-removal methods are costly or unsuitable for low-power wearable systems.
Method
A DRL-injected 1 kHz sinusoidal stimulus is demodulated to derive an impedance-mismatch reference for an NLMS adaptive filter followed by Hampel residual-spike suppression.
Results
Artifact power reductions of 6.5, 12.6, and 9.0 dB were achieved during head nodding, electrode tapping, and jaw clenching, respectively, while alpha-band modulation remained observable.
Takeaways & Limitations
The results support the feasibility of continuous relative impedance-mismatch monitoring for motion artifact cancellation in wearable dry-contact Ear-EEG.
Takeaways & Limitations
Ground-truth validation is still needed, and future work includes larger cohorts and phantom-based electrical validation.
Abstract
from arXiv · showhide
Dry-contact ear-electroencephalography (Ear-EEG) enables wearable neural monitoring. However, motion induced electrode-skin impedance (ESI) mismatches between electrodes can severely degrade signal quality. To the best of our knowledge, this paper presents the first proof-of-concept dry-contact EarEEG system that uses a driven-right-leg (DRL) stimulus for continuous ESI mismatch monitoring, enabling online adaptive motion artifact cancellation. A 1 kHz sinusoidal stimulus is injected through the DRL electrode. The resulting response to the injected carrier is separated from the EEG using bandpass filtering and demodulation, and then used to extract the ESI mismatch information as the reference input for a normalized least-mean-square adaptive filter followed by a Hampel filtering stage. To evaluate artifact suppression and preservation of neural activity, alpha-band EEG activity was analyzed involving four healthy participants performing head nodding, electrode tapping, and jaw clenching. The system achieved artifact power reductions of 6.5, 12.6, and 9.0 dB (77.6%, 92.8%, and 86.4%, respectively) while alpha-band modulation remained clearly observable after processing. This demonstrates the feasibility of DRL-stimulusbased ESI mismatch monitoring for motion artifact cancellation in wearable dry-contact Ear-EEG.
I. INTRODUCTION
Dry-contact Ear-EEG supports wearable neural monitoring, but motion-induced electrode-skin impedance mismatches can severely degrade EEG quality. The proposed DRL-stimulus approach monitors mismatch continuously and supports online adaptive motion artifact cancellation.
- Motion artifacts from head movement, jaw motion, and contact-pressure changes overlap key EEG bands, so frequency filtering can remove neural information.
- Existing artifact-removal approaches may require high computation, multiple channels, large datasets, or offline processing, limiting use in low-power wearable Ear-EEG.
- ESI variations provide a motion-artifact reference because they directly reflect electrode-skin interface changes, whereas conventional monitoring can burden sensing electrodes and common-mode rejection.
- The system injects a low-amplitude 1 kHz DRL sinusoid, demodulates its carrier response into an ESI-mismatch-correlated reference, and applies online adaptive cancellation.
- The evaluation used alpha-band modulation and motion-artifact cancellation during head nodding, electrode tapping, and jaw clenching in four healthy participants.
II. SYSTEM OVERVIEW AND METHODOLOGY
The system combines one in-ear sensing electrode, a forehead reference, and a DRL circuit for EEG acquisition and ESI-reference measurement. A DRL-injected carrier separates the EEG and ESI-related signals by frequency.
- The architecture uses one in-ear sensing electrode referenced to an external forehead electrode at Fp1, together with a DRL electrode.
- The DRL feeds back an inverted common-mode estimate while injecting a 1 kHz sinusoid into the common-mode voltage.
- The injected frequency is selected outside the EEG frequency range of 0.5-40 Hz.
- When electrode-skin impedances mismatch, part of the common-mode signal appears with EEG; a 0-100 Hz LPF extracts EEG and a 900-1200 Hz BPF extracts ESI.
B. Analog Multiplication and I/Q Demodulation
The extracted ESI carrier is synchronously demodulated into baseband components that track motion-related impedance mismatch and provide a relative artifact reference.
- The extracted ESI signal is multiplied by in-phase and quadrature-phase stimulus signals, then low-pass filtered to remove double-frequency components.
- The resulting baseband components estimate amplitude and phase changes associated with motion-induced ESI mismatch.
- The extracted in-phase component serves as a relative ESI-mismatch-correlated reference rather than an absolute impedance estimate.
C. Hardware Implementation
The hardware uses commercially available components to acquire Ear-EEG and demodulated ESI signals, transmit data wirelessly, and support online processing.
- The acquisition circuit board integrates the system hardware, while a separate BLE module provides signal transmission.
- The analog front end includes input buffers, fully differential instrumentation amplifiers, a DRL circuit, and a fourth-order differential bandpass filter.
- Fully differential multipliers perform I/Q demodulation before both Ear-EEG and demodulated ESI signals are digitized.
- Power converters provide 5 V, 2.5 V, 3.3 V, and −5 V supply levels.
D. Online Processing Algorithm Implementation
The online pipeline filters EEG and impedance signals, uses the impedance-derived reference for adaptive artifact cancellation, and evaluates processing in four healthy participants across three motion conditions.
- Online Processing Algorithm Implementation: The filtered ESI reference drives an adaptive FIR filter, while the filtered EEG serves as the processed signal input.
- Online Processing Algorithm Implementation: NLMS updates accommodate large impedance-reference amplitude changes, with threshold-based step-size scaling to limit over-adaptation during stable EEG periods.The implemented filter has 64 taps, μ0 = 0.06, and ϵ = 10^-3.
- Online Processing Algorithm Implementation: The study collected Ear-EEG from four healthy participants using a custom earpiece with an Ag/AgCl sintered pellet electrode.The DRL and reference electrodes were placed at Fpz and Fp1, respectively.
- Online Processing Algorithm Implementation: Artifact trials used 300 s recordings divided into five eyes-open, eyes-closed, and intentional-artifact segments for motion-condition evaluation.The tested motions were head nodding, earpiece tapping, and jaw clenching.
III. RESULTS AND DISCUSSION
The results section evaluates sinusoidal stimulation, alpha modulation, and impedance-aware motion artifact cancellation in dry-contact Ear-EEG recordings.
- III. RESULTS AND DISCUSSION: The evaluation targeted Ear-EEG quality under sinusoidal impedance stimulation, alpha-band modulation, and the proposed digital motion artifact cancellation pipeline.
- III. RESULTS AND DISCUSSION: The study reports alpha-modulation results using individual and grand-average alpha modulation ratios.
A. Effect of Sinusoidal Stimulation
The study compared eyes-closed Ear-EEG recordings with and without a 1 kHz DRL stimulus using relative alpha power. The nearly unchanged alpha-band measure supports stimulation for impedance monitoring without noticeable alpha degradation.
- A. Effect of Sinusoidal Stimulation: The effect of 1 kHz DRL stimulation was evaluated by comparing eyes-closed recordings with and without stimulation using relative alpha power.
- A. Effect of Sinusoidal Stimulation: 0.06 dB was the difference between average RAP values of -2.15 dB without stimulation and -2.21 dB with stimulation.RAPdB uses alpha-band power relative to total EEG power after excluding the surrounding alpha band.
B. Alpha Wave Modulation
Alpha modulation was quantified from alpha-band power, with spectrograms and grand-average plots used to assess modulation across recordings. The recordings showed detectable alpha modulation across the four participants.
- B. Alpha Wave Modulation: AMR is calculated from mean alpha-band power during eyes-closed and eyes-open conditions.
- B. Alpha Wave Modulation: The alpha-modulation spectrogram uses 4 s segments with 3 s overlap, alongside a grand-average alpha power plot for 12 recordings.Eye blink artifacts are identified as EBA.
- B. Alpha Wave Modulation: The motion-artifact evaluation compares pre- and post-filtering spectrograms, time-domain signals, and relative alpha-band power across experimental segments.
C. Motion Artifact Cancellation
Motion artifact cancellation was evaluated across head nodding, electrode tapping, and jaw clenching using PSD-based artifact reduction and alpha-band measures. The results showed reduced low-frequency artifacts and preserved observable alpha-band modulation, with electrode tapping producing the strongest reduction.
- Evaluation: Artifact reduction was evaluated across head nodding, electrode tapping, and jaw clenching using 12 recordings per condition from four participants.Assessment used spectrogram inspection, average artifact band power reduction, and relative alpha band power improvement.
- Evaluation: RdB and R% were calculated from mean PSD values and averaged first within participants, then across four participant-level values.The metrics were computed separately for each recording before hierarchical averaging.
- Results: The Hampel stage replaced less than 1% of samples on average, while Fig. 4 showed reduced low-frequency artifacts and preserved observable alpha-band modulation.The sparse replacements indicate limited residual spike correction within the reported processing results.
- Results: Electrode tapping produced the strongest artifact reduction, whereas head nodding and jaw clenching also introduced EMG components not directly represented by the ESI reference.The reported interpretation attributes tapping performance to its direct effect on the electrode-skin interface and capture by the ESI reference.
IV. CONCLUSION AND FUTURE WORK
The study demonstrates a wireless dry-contact Ear-EEG system that continuously monitors relative ESI mismatch with a DRL stimulus for motion artifact cancellation. Reported reductions reached 77.6%, 92.8%, and 86.4% across three motion conditions while alpha-band modulation remained observable; future work targets broader validation and deployment.
- Conclusion: The wireless dry-contact Ear-EEG system uses a DRL-based sinusoidal stimulus for continuous relative ESI mismatch monitoring.
- Conclusion: Artifact power reductions were 77.6%, 92.8%, and 86.4% during head nodding, electrode tapping, and jaw clenching, respectively.
- Conclusion: Alpha-band modulation remained observable after processing, supporting impedance-aware processing for wearable Ear-EEG during motion.
- Future work: Future work will include larger cohorts, ground-truth and phantom-based electrical validation, a compact fully in-ear earpiece, and extension to other wearable biosensing modalities.