A reader of past voltage reduced 7 Hz filter phase error in nine SEEG rest records with known initial state and 100 ms simulated gaps every 200 ms

A reader used 500 ms of past voltage to predict the input to a fixed 7 Hz filter during 100 ms simulated gaps. Its learned map had no state input. The test used nine public SEEG rest records, four fixed pairs per record, and one exact initial state. With gaps every 200 ms, mean recovery phase error was 36.29 degrees, compared with 44.99 degrees for held voltage. The paired decrease was 8.69 degrees. Its descriptive 95% record bootstrap interval was 5.77 to 11.98 degrees. All nine record means decreased. The difference from the earlier guarded reader was uncertain. A proof shows fixed filter decay of state differences when voltage histories are shared and independent of the estimated state. The proof permits varying observed intervals. All nine source extracts matched fresh public data. A separate process reproduced all 189 analysis outputs exactly. The package preserves data, code, weights, figures, provenance, and replay instructions. This is an exploratory study of numerical filter recovery. It did not test true neural phase, measured stimulation artifacts, or stimulation outcomes.

SEEG · signal processing · phase estimation · missing data · state-space filter · brain-computer interface · ridge regression