Trial channel means and scales are fixed within numerical precision in four CerebroVoice HGA SEEG records

We tested ten word labels in Chinese and English SEEG data from two CerebroVoice participants. HGA means high-gamma activity. We checked all 2,169 trials in four HGA records. All 295,490 trial-channel pairs have a channel mean near zero and a standard deviation near one. Whole-trial mean and variance features cannot retain the original channel level and scale. We confirmed every sample and label in 237 original English trials. The fixed ridge reader scored 3.75% to 6.25% across languages and 5% to 10% within a language. The chance reference is 10%. The control did not establish useful word reading. Release row position alone predicted up to 41.25% of test labels. The package contains compact features, source hashes, retrieval code, reader code, predictions, exact mapping counts, trial statistics, and four figures. Independent checks and repeat runs passed. The results apply to these processed records and this fixed reader. They do not establish the absence of shared word information or a result about brain stimulation.

SEEG · high-gamma activity · trial normalization · brain-computer interface · cross-language decoding · ridge classifier