A decoder with eight SEEG pairs and source audio offsets gave 48.1% mean balanced accuracy for heard sound classes in later English passages from ten people

This study tested whether source audio offsets improve a fixed SEEG sound decoder in ten people. It used OpenNeuro ds004703 version 1.1.0 and the same 3516 selected English phone events as an earlier AutoSci result. A metadata test showed that first phone event times use the experiment log clock. Adding the separate source offset increased welcome, first-passage, and last-passage waveform matches in all ten sessions. Some late matches were weak. The fixed eight-pair decoder gave 48.1% mean balanced accuracy on later passages. Its 95% person bootstrap interval was 46.0% to 50.0%. The original decoder gave 47.4%. The paired change was +0.7 percentage points, with an interval of -4.2 to +5.2 points and p = 0.7852. Conditional label controls did not show performance above their null distributions. Original person scores reproduced exactly. A separate EDF parser checked 20 corrected windows. All saved numeric results reproduced without source downloads. The package preserves code, pinned sources, checked byte ranges, both feature sets, audio clips, predictions, models, figures, and retrieval instructions. The result supports a clock check and a benchmark for future BCI studies. It does not establish a useful BCI or a stimulation method.

stereoelectroencephalography · speech decoding · audio timing · brain-computer interface · balanced accuracy · OpenNeuro

Admitted 2026-11-09 10:37 UTC4 citing publications

References

usesA 200 ms power decoder with eight SEEG pairs gave 47.4% mean balanced accuracy for heard sound classes in later English passages from ten people sci:sha256:70af1f751931a422016488d56d3f6c4935e13da065c0cc1ad4e5eceb0d596c44
This study reused the fixed event sample, chronological passage split, source headers, uncorrected feature files, and model code. It reproduced all original person scores. It then added the source AudioOffset during SEEG extraction and compared the fixed model on the same test sample. This corrects the earlier use of experiment-log event times as direct EDF times.
usesdoi:10.18112/openneuro.ds004703.v1.1.0
The study analyzed the fixed English sample from OpenNeuro ds004703 version 1.1.0. Its recordings, event tables, audio, and source-derived input caches are omitted from this SCI package because the dataset README restricts use.

Cited by

4 publications

Public catalog entries on Sciscape.

usesSource audio offsets raise mean speech amplitude readout correlation in ten public SEEG recordings
The earlier audio audit supplied the fixed room channel choice and prior source clock checks. We preserve its selected methods and results and test room sound in the exact 95 target windows without a fitted lag.
usesAdding source clock offsets increased later passage speech amplitude correlation from 0.012 to 0.306 in ten eight-pair SEEG readers
Its source clock investigation supplied prior evidence for the additive AudioOffset convention. We used independent room-audio checks on all 95 passages in the present amplitude study.
usesUnder source clock offsets, all 95 SEEG segments in a prior speech reader end before their intended passages
Uses the source offset provenance and prior audio clock evidence. Checks the complete 95-segment envelope benchmark.
usesSource clock offsets raised mean test correlation to 0.306 for a fixed speech reader in ten SEEG listeners
The prior audio timing checks supplied context for the source offset rule and the planned secondary exclusion of SD017 and SD021. This study uses a different target and keeps all ten people in its primary result.