Software
Decode first, offline
Phase 0 trains classical and deep models on the Nieto Inner Speech EEG set. Splits are subject-wise. TribeV2 is assessed, not assumed.
Phase 0
Software backbone is live · Inner Speech EEG
Software backbone is live
Inner speech, measured from the outside.
NORT is a research project for non-invasive EEG+sEMG acquisition and speech decoding. The long-term vision is communication for people who cannot speak. Today we publish methods, not medical claims.
Phase 0
Reproducible decoding on open Inner Speech EEG
Phase 1
Single-pod temple / ear wearable for EEG+sEMG
Open logs
Every method choice is written down, including misses
How the work is structured
Two tracks, one constraint: stay non-invasive until signal quality is proven on the bench.
Hardware
One temple pod
Phase 1 is a research instrument: 4–6 hydrogel contacts, EEG-class AFE, BLE timestamps, and a charging case. No stimulation. No charging while worn.
Boundary
Non-invasive, explicit
Invasive speech BCIs can work. NORT stays outside the skull. ALS communication is the north star, not a current product claim.
Latest from the lab
All posts →Aug 4, 2026 · Status
Iteration analysis after the classical ladder
What held up across CSP, FBCSP, Riemannian, and SVD runs — and what did not transfer.
Jul 21, 2026 · Software
EEGNet on Inner Speech derivatives
First deep baseline with the same subject-wise splits used by the classical stack.
Jul 20, 2026 · Methods
Classical ladder final report
CSP+LDA, FBCSP findings, Riemannian TS+LDA, and truncated SVD+SVM in one comparison.
Follow the work
Lab notes, method comparisons, and hardware gates — published as they happen. No press theater.
Open the blogBuild with us
Signal processing, hardware, ethics, or datasets. Tell us how you want to help. We reply to serious notes.
Start a conversation