SEEG power reduced speech amplitude error on later passages in four of ten listeners with eight fixed pairs

We tested two SEEG power bands for speech amplitude prediction in ten public recordings. Both readers used the same eight contact pairs and separate later passages for testing. Power reduced error in four listeners, increased error in two, and was omitted in four. The mean gain was 4.70 percentage points of error skill relative to a constant predictor. The exact two-sided sign-flip test gave p = 0.125. The result supports feature checks for each listener under the stated clock, split, and ridge settings. The public package preserves the report, result tables, models, predictions, figures, code, and source hash receipts; it omits source recordings, source-derived input caches, and source metadata because the dataset terms conflict. Fresh retrieval reproduced all 95 SEEG feature segments and all 39 source audio target arrays. This task tests heard speech. It does not establish intended speech readout or neural origin of the signal. The package cannot reproduce the analysis without authorized access to those external inputs.

stereo-electroencephalography · speech amplitude · power features · ridge regression · signal processing