We’re crafting our complete ecosystem from the ground up—from the recording interface, through ML processing, to applications like artistic creation. Soon, we’ll venture into language recognition, drawing on groundbreaking theories to solve the inverse EEG challenge. Begin with brain-born art, evolve to thoughts woven into dialogue!
While the horizons we’ve charted brim with audacious promise, we recognize they’re too vast to conquer in a single stride. Instead, we’re channeling our talents into EEG-fueled digital artistry—a pathway not only to self-sufficiency but to nurturing resources for our grander pursuits. We’ll immerse ourselves here at the outset, traversing the full arc of development before venturing into linguistic realms.
Our blueprint pulses with innovation at every layer:
Electrode Arrays Reimagined for Unrivaled Reliability
We’re forging bespoke arrays that surpass today’s pinnacles—infusing them with steadfast precision to capture the brain’s subtlest symphonies.
Achieving reliable BCI starts with capturing clean, robust EEG signals. Our work centers on designing innovative scalp electrode arrays that maximize spatial resolution and comfort while minimizing artifacts. Once captured, these high-quality signals undergo advanced Fast Fourier Transform (FFT) analysis to extract meaningful brainwave features for real-time decoding. This two-step approach—superior signal detection, followed by powerful spectral analysis—is foundational to both our creative and language-intention goals. We actively invite collaboration from engineers, neuroscientists, and hardware specialists to help advance electrode design, artifact resistance, and open-source analysis pipelines.
- Innovative electrode arrays for cleaner, more comfortable EEG acquisition.
- Artifact-resistant detection for robust, real-world use.
- Advanced FFT and signal processing pipelines for extracting actionable brainwave features.
- Open-source hardware and spectral analysis tools for community-driven innovation.
Our Technical Tricks
The skin can be prepped using time-honored EEG essentials to achieve equilibrated impedances hovering just above 500 Ω, yet shy of 1 kΩ. These thresholds, seldom spotlighted in scholarly discourse, are eminently attainable as standard practice.
Electrode surfaces must steer clear of unalloyed metals, whose galvanic potentials sow discord in recordings—particularly at the languid low frequencies that echo our brain’s primordial cues for motion intent. Today, these signals elude steadfast capture, besieged by artifacts from undue polarization. The age-old chlorination of silver electrodes offers a bulwark, outshining mere metallic sheaths for stabilizing those deeper drifts. Enhanced waterproofing, too, will extend electrode endurance, while localized shielding against household electromagnetic hums will sharpen fidelity further.
Dense arrays in caps, though handy for easing technician toil in setup, falter when chasing rock-steady, low impedances—and they often court pressure-induced throbs. Paste-laden cup electrodes, by contrast, prove far gentler and steadfast for protracted sessions. We envision crafting a swifter-placement alternative with matching dependability; with secured funding, we’ll pioneer robotic deployment to upend the prep phase of EEG entirely.
To elevate EEG signal fidelity, we’re pioneering a wider spread of several key innovations:
Custom Electrodes
We start with age-old Ag/AgCl cup-electrodes, which we chlorinate ourselves after slight modification and waterproof re-assembly in silicone strips at specific inter-electrode distances. Please contribute by sending us your broken Ag or AgCl-electrodes to recycle! CONTRIBUTE!.
We intend to explore the performances of our designs scientifically, and will share the data openly. Successes will be developed into manufacturing lines available for sale.
To develop higher-density arrays, smaller electrodes will be created and assembled in a similar manner. We hope to be able to reach about 0.8 cm inter-electrode distances without losing our envisaged impedances of just below 1 kΩ, arranged in non-standard triangular arrays.
This is also an improvement in patient comfort as we will not give up the advantage we have with cup-electrodes.
Chlorination is essential to reach into the low-voltage, low-frequency waveforms known to indicate that planned movement is about to occur, up to half a second before it actually happens, giving assistance a natural feel and promoting cerebral integration of such assistance.
Adaptive Electrode Placement
Electrodes align equidistant in flexible silicone, also providing a waterproof environment to Ag not touching scalp, and its connections to cables. It can easily be prepared for, and placed in parted sections of hair for stability.
These strips can be concentrated in regions known for contribution to painting strokes and selection modes, which may even vary between individuals, allowing for enhanced adaptation to each person’s unique distributions of significant activities.
These distributions will form a separate line of scientific investigation. For simple painting there is expected to be less inter-individual variance in distribution than for speech, for example, but the techniques for ML-driven decisions herein should be able to be carried over in the future when we venture into the speech-detection arena.
Other than with indwelling/intracranial electrodes, we have little limitations in the size(s) of recording areas we can cover.
Noise Management
We believe the old concept of “garbage in—garbage out” and will minimise artefacts generated and recorded in the first place.
Increased comfort should be only the starting point. Improved, balanced, low impedances are associated with more stable electrodes, resulting in less artefacts of environmental origin. To enhance this further, metal gauze screens will be placed in front of monitors and other sources of high electromagnetic fields. And, the electrodes will be covered with soft, isolating material under a layer of conductive material providing further local isolation of the electrodes and leads from electromagnetic fields.
Further, old techniques of physically reducing artefacts to a minimum before digitisation will also be used to a maximum to ensure the widest possible frequency window available for ML to identify significant brain activity down to sub-μV levels. Sufficient signal fidelity enabled in such a way may open scalp-based speech recognition, a field recognised as a wide-application technology with limitless application and supporting further acceleration of technological development of mankind. That is in addition to rehabilitation of individuals who lost motor speech ability. It may even benefit people with other speech deficits eventually.
Application of real-time artifact suppression algorithms directly into the acquisition hardware to minimise motion and environmental noise recording and elevate CMRR is a direct side-effect of these old techniques to suppress biological and physical artefacts.
High-Density Configurations
We intend to develop ultra-dense electrode grids to capture fine-grained spatial patterns of brain activity, crucial for decoding complex intentions. Traditionally denser arrays are formed by adding into the square grid of the 10-10 System, but denser patterns can be obtained by triangular patterns without more risk for salt bridges. Even that can be placed in straight hair partings, thus not increasing prepping and placement complexity, or adversely affecting electrode stability and contact fidelity.
Ergonomics at the Heart of Preparation
Mindful of the human touch, our designs embrace fluid prepping techniques: accelerating setup while restoring crystalline fidelity to high-density configurations, turning ritual into revelation.
Improved prepping technique will also facilitate robotic placement of high-density arrays in future. This is highly dependent on securing large amounts of surplus funding at a later stage.
Paving the Way for Robotic Grace
With tomorrow in sight, we’re priming our systems for seamless robotic orchestration—elevating accuracy across countless placements and liberating technicians from the grind, so focus can soar to loftier pursuits.
ML Architectures: The Bold Core of Detection and Insight
As highlighted in our electrode contact strategies, superior signal fidelity forms the bedrock for advancing AI deeper into the pipeline.
Yet, our ambitions extend further: We’re engineering machine learning (ML) and software architectures grounded in enhanced physical precision, capable of pinpointing the origins of electrical activity within brain tissue at resolutions of at least 0.5 mm. Attaining such acuity may necessitate high-resolution MRI scans—an added layer of investment, though one we deem indispensable.
Beyond merely mapping the beta and gamma waves long linked to pre-movement intentions, we aim to broaden the entire frequency landscape. This includes heightened sensitivity near the mains power supply frequency, extending down to the faintest extremes of conventional EEG ranges. We recognize potent low-frequency oscillations emerging up to two seconds prior to action, and we’re dedicated to exploring their roles—not just in refining localization, but in forging more robust ML frameworks. Such insights could unlock the exhilarating domain of intent-driven language decoding.
Crafting machine learning frameworks to unerringly sense electrical whispers and map their intricate dances of origin and interplay? This is our project’s most daring crescendo—yet utterly within reach, a testament to our shared ingenuity.
We’re honing our EEG recording parameters for exquisite precision, minimizing the “clutter” that clouds AI inputs—pristine data paves the way for truly captivating creations.
We hold that much of the diminished clarity in EEG recordings stems from rampant artifacts that obscure the subtlest neural whispers. Clinical amplifiers routinely discern waves as faint as 0.5 μV, while signals dipping to 1 μV often demand detection in forensic contexts to affirm electrocortical silence in cases of suspected brain death. This realm of ultra-fine resolution is no stranger to the field—merely a formidable craft to wield with unwavering consistency.
Moreover, our initial EEG captures will eschew the elaborate post-recording software sorcery now commonplace for scrubbing artifacts from raw traces. By deferring such interventions until signal purity peaks, we stand poised to elevate today’s pinnacles of resolution—even for scalp-derived models decoding language (as text) from brain rhythms.
An Intuitive Canvas for Brain-Born Expression
We’re birthing a user-whispering art platform, bolstered by AI companionship, empowering those whose hands rest idle—by choice or circumstance—to paint directly with neural strokes. Rooted in open-source foundations, it tempers costs and invites global embrace, democratizing this profound craft.
And oh, how we yearn for this to bloom as the vital bridge: a foundation from which language itself unfurls, unlocking cascades of expression and utility in the chapters yet to come.