Window normalization preserved phase commands under constant positive gain and reduced clean accuracy in seven SEEG rest replays
We tested input gain changes in 28 selected bipolar SEEG pairs from seven people in OpenNeuro ds006065 version 1.0.0. Normalization of each past input window preserved every phase command under constant positive gain, with fixed models, scales, and gate thresholds. When gain was halved or doubled, the original reader had a mean absolute command rate change of 122.67%. The normalized reader had a change of 0%. On clean input, accepted-command accuracy fell from 69.54% to 66.82% against an offline phase reference. The decrease occurred in all seven people. A proof and direct voltage replays show gain and constant offset invariance. Abrupt gain steps break the proof within the history window. The reference uses future voltage. No stimulation or live interface benefit was tested. The package preserves selected voltages, code, figures, command arrays, source hashes, fresh source audits, and reproduction instructions.
SEEG · brain-computer interfaces · phase forecasting · window normalization · gain invariance · command accuracy
References
4- usesConfidence bins reduced error arc scores for 50 ms theta phase forecasts with an offline reference in seven SEEG rest recordings sci:sha256:33b71fcdcca582e278237f7c6ac04c201ee338c58b607503eba2a92616c0941b
This publication supplied the original frozen ridge phase models and their fit and tuning lineage. We used those fixed models through the cited voltage package and did not fit a new model.
- usesForecast vector gates gave higher phase accuracy than amplitude gates after equal early retention in seven SEEG rest replays with an offline reference sci:sha256:fbf0ce2e5b37014408502ae7aeeaa42b8787b86978ef0beb0506f488e249c88c
This publication supplied the selected 590 s voltage arrays, frozen phase models, source manifests, and phase command rule. The new analysis reproduced every clean candidate time and vector gate flag before testing input gain and window normalization.
- usesEarly frequency selection reduced mean 50 ms offline theta phase error by 3.55 degrees in seven SEEG rest recordings sci:sha256:5c3ae075960de0774e0fcc176ee07a5e3c9a353d86238f1e49023a23aa4d2bfa
Its preserved SEEG pair selection code ranks early theta power and excludes shared contacts. We used that code to reproduce pair selection from all seven fresh full source records.
- usesdoi:10.18112/openneuro.ds006065.v1.0.0
OpenNeuro ds006065 version 1.0.0 supplied the seven SEEG rest records analyzed here; the package preserves the source hashes, metadata, selected voltages, and retrieval audit.
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesFuture input replacement gave 51.12 percent phase hits at accepted 50 ms commands in seven SEEG rest records
Use its seven public voltage subsets, 28 selected SEEG pairs, frozen raw reader models, source metadata, and full source audit. Reproduce all 204043 clean candidates, 49780 accepted commands, and original gate thresholds.