Real inputs.
Visible calculations.
Analyze an EEG recording. Inspect a live device stream. Turn saved task sessions into a personal practice plan. Each engine keeps its inputs, quality checks and outputs visible.
From samples to spectrum.
Bring a recording or your own compatible acquisition bridge. Processing runs on your device. No signal is generated when no data source is connected.
No data source. Import an EEG recording or connect a device bridge.
Your signal belongs here.
The waveform, spectrum and measurements appear only after samples pass the input checks.
- Source mode
- —
- Analyzed duration
- —
- Accepted 4 s epochs
- —
- Rejected 4 s epochs
- —
| Band | Frequency | Power µV² | Share of 1–45 Hz |
|---|
Band power is a signal measurement, not an intelligence, concentration, relaxation or brain-health score. Basic screening can miss artifacts.
Compare alpha power with your own reference
Keep the same channel, posture, eyes condition and recording setup. Calibration collects 15 accepted four-second epochs. Rejected epochs do not count.
Connect a device to begin.
No feedback until a reference and a later accepted epoch are available.
This display reflects relative alpha power only. Higher is not necessarily better. It does not prescribe a mental state, stimulation or treatment, and benefit has not been established.
Export contains channel measurements, accepted/rejected epoch reasons, settings and calibration when present. Nothing is saved automatically. Changing source or clearing data removes the current analysis from this page.
What the engine calculates
Four-second epochs are screened for flat signal, repeated samples, large amplitude changes and strong 50/60 Hz contamination. Each spectrum uses Welch’s method: two-second periodic Hann windows, 50% overlap, segment mean removal and one-sided power density.
Recorded-file band summaries average accepted epochs only. Live feedback compares a later accepted epoch with the median alpha share from 60 seconds of accepted calibration data. Feedback stops on stale data, gaps, disconnection or when you leave this tab.
Screening thresholds are disclosed engineering rules. They have not been validated as a medical signal-quality detector. Read the full protocol and limitations.
Your results.
A reasoned next step.
Choose the skill you want to practice. Four specialized software agents check your records, evaluate a personal prediction model, build your next session and review comparable progress. Every output has a visible source.
A plan starts with your evidence.
Complete and save a task, then run the agents. An empty account gets a starting protocol. Model predictions appear only after enough comparable history exists and the learned model beats its baseline.
Start with Focus GridTechnology you can inspect. These are bounded, on-demand software agents implemented in Noetryx's application. The Model agent fits learned numerical weights. Signal, Practice and Review use disclosed rules and statistics. No external chatbot or large language model is involved.
Your latest run is saved with its source revision and cleared when the underlying records change. No brain activity, medical diagnosis or proven training effect is inferred.
Explore the technology and capability status