01
The signal is weak.
Electrical activity measured at the scalp is small and mixed with noise. Useful systems must begin by knowing whether the signal is valid at all.
A deliberate choice. A measurable signal. A system that knows when to stop.
THE CURRENT PROTOTYPE
The current AXLminds prototype records non-invasive EEG wirelessly across four channels at a nominal 250 Hz. A five-minute worn recording completed on 30 August; further worn baselines were recorded in September. These are acquisition results, not proof of usable neural signal. The rate comes from device configuration and does not independently verify ADC cadence.
The next question is whether deliberate attention to two visual targets can be distinguished reliably from each other and from idle. September readiness checks halted before task trials, and later baselines showed substantial 50 Hz interference. Signal processing, visual stimulation, calibration tools, and guarded command software are implemented and tested offline. Reliable live command control is not yet validated.
An experimental EEGNet workflow supports offline analysis and observation-only predictions. Its wiring checks used a model trained on synthetic data; no task-trained model or held-out EEGNet accuracy result is established. A separate accelerometer-based slide prototype explores head movement as input. It is not EEG control, and physical movement performance remains unverified.
01 / PROBLEM
Useful signals are difficult to acquire honestly.
02
The environment changes it.
Movement, muscles, electrode contact, nearby electronics, and changing conditions can all alter a recording.
03
Every person is different.
Brain signals vary across people and across sessions. Calibration and measured uncertainty are part of the system, not optional polish.
04
The laboratory is not everyday life.
Many demonstrations depend on controlled conditions. The field still needs acquisition that is reliable, comfortable, and practical outside them.
02 / APPROACH
A narrow path from signal to action.
- 01
WEARABLE EEG
Non-invasive electrodes record scalp electrical signals, including brain activity, muscle activity, and environmental noise.
- 02
ACQUISITION
Hardware and software preserve the raw signal, timing, and channel context.
- 03
VALIDATION
Signal quality and movement checks reject data that should not produce an answer.
- 04
INTERPRETATION
Signal processing compares a narrow set of intended choices and keeps uncertainty explicit.
- 05
COMPUTER ACTION
The intended endpoint is a small, logged command. Live control stays gated until participant-specific evidence passes validation.
03 / PROTOTYPE
Narrow and measurable by design.
BUILT / EVIDENCE RECORDED
- Four-channel wireless EEG acquisition at 250 Hz nominal, with a completed five-minute worn recording
- Local raw EEG and movement-sensor recordings with timing, configuration, and integrity checks
- Attention decoding, visual stimulation, calibration, and guarded cursor software tested offline
- A guided confirmation protocol with frozen analysis and separate new recordings
- Experimental EEGNet replay and observation-only wiring checks using a synthetic-trained model
- Offline readiness and denoising comparison tools that do not change live control gates
NEXT VALIDATION GATES
- Usable physiological signal quality in the current electrode placement
- Line interference, device timing, and raw-recording completeness without treating quieter traces as better decoding
- Deliberate visual attention: left, right, and neutral on unseen trials
- Whether eyes, muscles, movement, or task cues explain a result
- Repeatability after refitting, idle rejection, and eventual live command performance
LONG-TERM GOAL
- Reliable operation beyond controlled experiments
- More natural and comfortable interfaces
- Broader real-time research into cognition, memory, communication, and control
Live command accuracy, idle false activations, decision latency, and long-duration stability remain unvalidated. Software tests and public-data replay do not replace participant evidence.
04 / FUTURE
Research directions, not product claims.
CONTROL
Translate deliberate brain activity into useful action with less friction.
COMMUNICATION
Reduce the distance between an intended idea and its expression.
MEMORY
Explore how technology might preserve and retrieve what people otherwise lose.
COGNITION
Research interfaces that can work alongside how people think.