Features for two more Dutch SEEG participants are ready, but the planned group test remains incomplete

This package adds checked log high-gamma and log-mel features for participants sub-05 and sub-06 in the SingleWordProductionDutch-iBIDS dataset. Their NWB files were read from the OSF archive by HTTP range requests, with zip CRC32 checks passing; the source URL, file hashes, and participant sidecars are recorded. The unchanged reference pipeline produced 70–170 Hz log high-gamma and 23-bin log-mel features at 10 ms frames for 60 contacts and 100 words in sub-05, and 127 contacts and 99 words in sub-06. The planned analysis did not finish within the 45-minute run limit, so this package reports no new statistical result. It preserves the preregistered analysis plan, code, feature files, and earlier results for sub-01 through sub-04. The four-participant report found that standard mean-bin Pearson r matched the envelope-only reconstruction within 0.002 in all eight participant-by-decoder cases; within-speech spectral-shape r exceeded all 20 circular-shift null values in all eight cases. Those earlier findings are not a group-level test: at n = 4 the one-sided exact Wilcoxon minimum p is 0.0625. The included sub-05 and sub-06 features enable the planned n = 6 test without another raw-data download, but that test remains to be run.

SEEG · speech production · spectrogram reconstruction · high-gamma · log-mel · spectral shape · envelope · feature extraction · Dutch word production