Six calibration passages gave a mean 6.5% lower later squared error than two in nine SEEG sound readers at a 400 ms delay
We tested whether more calibration passages reduced later error in a SEEG reader of heard sound amplitude. The public source was OpenNeuro ds004703 version 1.1.0. Nine people had six selected early calibration passages and 100 later test passages. Each reader used eight selected bipolar pairs and a fixed 400 ms output delay. Pair selection and penalty choice used calibration data only. Six passages gave a mean 6.51% relative decrease in squared log amplitude error compared with the first two. Seven of nine people improved. The 95% interval with shared test sound weights was -1.76% to 12.05%. It includes zero. At fixed penalty 100, an average over two stage orders assigned 10.96 percentage points of error decrease to fitting and 0.10 to selection. The mean stage pattern also held with a constant penalty per training row. The test set has prior use, so this is an exploratory reanalysis. It does not test intended speech, a live interface, or stimulation. The package preserves compact data, pinned retrieval records, code, all 396 models, predictions, five figures, and replay instructions. Fresh source checks covered all 154 neural ranges. Independent numerical checks and a full replay passed.
stereoelectroencephalography · brain-computer interface · speech amplitude · calibration · ridge regression · neural decoding
References
5- usesFixed SEEG readers had higher mean sound amplitude correlation at 400 ms than at zero delay on 100 unused later passage records in nine people sci:sha256:3c8b962f91454c5d1b32469fb10a6ecb9534841b65f6b7ab946bc56a062e2126
We used the later-test record rule and corrected clock mapping. Its earlier test use limits independent inference from our new calibration comparison.
- usesSelected SEEG pair deviations reduced mean error in log sound amplitude on 100 later passages in nine people with a 400 ms output delay sci:sha256:a2f979743f761bc5638efa44fcc4e1fe5803608c1ad497cd8ed4a35d31ce82af
We used the compact neural features, audio targets, source records, and MIT loading code. We fitted new readers and checked inherited inputs against its manifest.
- usesTen offline SEEG readers gave mean correlations of 0.214 at zero added delay and 0.302 at 400 ms for heard sound amplitude sci:sha256:0363ee7f6d1912a2c732d3938bfa9b204378683c1ded14e573a003ec7c3ef2fc
We used the selected six-passage calibration sets, fixed delay convention, and ridge method. We changed calibration size and separated selection from fitting.
- usesdoi:10.18112/openneuro.ds004703.v1.1.0
The source study identifies OpenNeuro ds004703 version 1.1.0 as the dataset used for the recordings, sound targets, channel labels, and timing records. Source inputs are omitted from this package because the saved source terms conflict.
- supportsdoi:10.1038/s41467-024-44844-9
Cites the source experiment that describes the clinical intracranial recordings during conversational speech listening.
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesBoth donor rules chose a penalty of 1000 for nine SEEG sound readers with two calibration passages and a 400 ms delay
We used the compact public inputs, passage choices, and exact reader functions. The local two- and six-passage predictions reproduced this source result. We added person and exact sound exclusions to test donor penalty selection.
- usesBands from four calibration passages per person covered 85 of 100 later sound traces with 117 points in nine SEEG recordings
This study uses the compact SEEG power features, sound targets, source identities, and loading and ridge code from this package. Two early passages fit the reader. Four separate passages set uncertainty bands.