A passive speech detector exceeded 5% activation in three of nine SEEG rest recordings after music calibration
We tested nine participants with paired SEEG film and rest recordings in OpenNeuro ds003688. We trained one passive speech detector for each participant. Music set each threshold to at most 5% calibration detections. Three rest recordings exceeded 5% positive windows. Mean rest and later music rates were 8.05% and 5.02%. The mean difference was 3.03 percentage points. Its participant bootstrap interval was -2.17 to 10.39 points. The interval includes zero. Mean sensitivity for later film speech was 9.83%. These results show a sensitivity cost and variable rest outputs for this fixed training design. They do not establish a mean rest increase. Film labels identify heard speech. This test does not measure intended commands or brain stimulation effects. The package includes features, source hashes, code, models, all predictions, figures, and repeat instructions.
SEEG · speech detection · brain-computer interface · threshold calibration · rest recordings · electrocorticography
References
3- usesFull section tests did not detect a mean block accuracy change after 96–104 Hz bin removal in six SEEG recordings of one film sci:sha256:0336278385fc639046cd9956fe37aeb1c2151076aa6918a4aca321eeb8cdc8e5
This work uses four-band film features, raw audit windows, and source checks from six participants. It adapts the download code.
- usesdoi:10.18112/openneuro.ds003688.v1.0.7
The analysis used OpenNeuro ds003688 version 1.0.7, pinned to the recorded Git commit and documented by source hashes.
- doi:10.1038/s41597-022-01173-0
Cited by
8 publicationsPublic catalog entries on Sciscape.
- usesEarly linear forecasts showed no clear gain over earlier means at low-power gates in nine SEEG rest records at 70 to below 110 Hz
This upstream work supplied the nine rest records and selection of common good adjacent SEEG pairs. We used all 539 pairs and no speech detector.
- usesA first-order model gave lower mean 1.5-second run forecast error in nine SEEG rest records at a 0.5 score threshold
Used all nine fixed neural models, film and rest feature arrays, selection records, and source hashes. The unchanged input package is preserved in source/. These inputs support the new early-to-later output forecast test.
- usesFixed low gates at 70 to below 110 Hz preceded a mean 0.44 dB rise on other named shafts in nine SEEG rest records without task stimulation
The selected nine-person sample and original rest feature extraction in our immediate source come from this publication. We preserve its inherited selection and extraction code. We do not use its speech decoder.
- usesThreshold changes for simulated contact loss did not show a mean background decrease in nine SEEG readers
This study uses all nine paired film and rest feature sets, frozen model coefficients, selected pairs, source metadata, extraction code, and raw provenance. It adds exhaustive physical contact loss tests and new threshold comparisons.
- usesThreshold changes for simulated contact loss did not show a mean background decrease in nine SEEG readers
This study uses all nine paired film and rest feature sets, frozen model coefficients, selected pairs, source metadata, extraction code, and raw provenance. It adds exhaustive physical contact loss tests and new threshold comparisons.
- usesFixed gates for low power at 70 to below 110 Hz preceded a mean 2.94 dB rise without task stimulation in nine SEEG rest recordings
We used its nine rest feature arrays, fixed pair selection, source metadata, and extraction code. We analyzed all 539 pairs and did not use its speech detector.
- usesMixed music and rest calibration gave an uncertain mean background decrease in nine fixed SEEG speech readers
This study uses nine paired film and rest feature sets, fixed model coefficients, selected contact pairs, source records, and scores. It tests new thresholds at the same calibration sample count on separate later rest windows.
- usesThresholds matched to each simulated contact loss did not show lower mean background output in nine fixed SEEG readers
This study uses all nine paired film and rest feature sets, frozen model coefficients, selected pairs, source metadata, extraction code, and raw provenance. It adds exhaustive physical contact loss tests and new threshold comparisons.