Eight SEEG contacts reduced mean log speech amplitude test error by 12.1% beyond cue timing in ten recordings with corrected clocks
We tested whether SEEG adds speech amplitude information beyond word cue timing. Ten people took part in a Dutch word reading task. The target was log audio RMS. Each reader used eight contacts. We tested 194 complete late word trials. With no added input delay, the mean error reduction was 12.1% (95% interval: 7.0% to 17.4%). Error fell in 9 of ten people. The interval is above zero. The whole-trial pairing control gave p = 0.0001. The source clocks had backward steps. A correction used current and past timestamps only. The package contains code, compact derived data, models, test predictions, figures, source hashes, and retrieval instructions. This task used overt speech and known cues. It does not establish silent speech reading or the neural origin of the signal.
stereo-EEG · speech production · speech amplitude · cue timing · high gamma · ridge regression · trial pairing · brain-computer interface
References
4- usesSEEG power reduced speech amplitude error on later passages in four of ten listeners with eight fixed pairs sci:sha256:9d7986c2893a49c9a3d06daa806e2fe9a866569d94a0968c0542dfed56513cc9
Used the earlier study as a method reference for trailing power windows, a small input budget, and testing on later data. This study uses new overt-speech data and adds a cue timing control. No earlier arrays, code, or fitted weights were reused.
- usesdoi:10.82901/nemar.nm000371.v1.0.0
The analysis used the pinned NEMAR nm000371 v1.0.0 release.
- usesdoi:10.17605/OSF.IO%2FNRGX6
The NEMAR release records this OSF project as the original source data record.
- doi:10.1038/s41597-022-01542-9
Cited by
7 publicationsPublic catalog entries on Sciscape.
- usesEight SEEG contacts gave 0.53% mean forecast gain beyond earlier audio at a 100 ms lead in ten cued speech records with corrected clocks
We reused the ten compact SEEG and audio feature arrays, source metadata, retrieval and extraction code, and corrected clock rule. They enabled new future-lead readers and earlier-audio controls. All reader weights were fit anew.
- usesEight SEEG contacts gave a mean 0.95% error reduction beyond recent audio for 100 ms speech forecasts in ten recordings
We reuse the all-contact feature arrays, source extraction code, pinned metadata, and raw retrieval recipe. These inputs enable new future voice-level tests with recent-audio controls. We fit new models and reconstruct all ten feature inputs from the raw source files. We do not reuse the source study's fitted models.
- Eight selected SEEG contacts lowered log speech amplitude error beyond cue and text controls in ten recordings with repaired time stamps
[Legacy relation: reproduces] All ten cue plus eight-contact readers reproduced the earlier selected contacts, penalties, and test mean squared errors. This replay checked the reference before the text control was added.
- usesEight SEEG contacts reduced mean error by 1.60 percent for sound windows that end 200 ms after released input in ten records with cue and past sound controls
We use the ten all-contact derived SEEG and sound feature files, pinned source metadata, source hashes, and raw retrieval and extraction code. We rebuilt all ten sources and verified exact input agreement. These inputs support a new fixed-lead forecast test with cue and past sound controls.
- usesPast SEEG improved mean 100 ms voice level forecasts beyond audio history in ten recordings with corrected clocks
This study uses the compact log power and audio level arrays, complete-word split, source metadata, extraction code, and public raw source hashes from this package. The same test words are reused. This study adds earlier shared cutoffs and audio history controls. A fresh raw rebuild reproduces all inherited arrays.
- usesJoint selection reduced contact correlation but gave no clear error gain for four SEEG contacts in ten speech recordings
We use all ten feature files, source metadata, extraction and retrieval code, and the eight-contact reference result. We test matched contact selection rules on the same later trials.
- usesPast SEEG improved mean 100 ms voice level forecasts beyond audio history in ten recordings with corrected clocks
This study uses the compact log power and audio level arrays, complete-word split, source metadata, extraction code, and public raw source hashes from this package. The same test words are reused. This study adds earlier shared cutoffs and audio history controls. A fresh raw rebuild reproduces all inherited arrays.