A 200 ms causal filter reduced output changes without a clear error gain in ten SEEG listeners with a 400 ms reader delay

We tested a fixed causal output filter for a public SEEG reader of heard speech amplitude. The sample had ten listeners and 38 later test passages from OpenNeuro ds004703 version 1.1.0. Each primary reader used eight fixed contact pairs and 70 to 110 Hz power. Its output was available 400 ms after the sound target window. A 200 ms filter reduced mean output roughness by 73.06%. Mean error skill changed by -0.020 percentage points. The descriptive 95% person bootstrap interval was -1.705 to +2.024 points. At rapid sound changes, mean skill fell by 5.826 points; eight of ten listeners had worse scores. Smooth output did not imply better amplitude estimates in this test. The package supplies fixed reader outputs, compact neural features, targets, source versions and hashes, code, three figures, a streaming filter tool, and repeat instructions. This is an exploratory test on reused passages. It does not establish live BCI performance or a general population effect.

stereoelectroencephalography · causal output filter · speech amplitude · high gamma power · brain-computer interface · signal processing · exploratory study