An anchor and difference code reduced mean six-bit 70 to 110 Hz power error in seven SEEG rest records with fixed limits

We compared two fixed-rate voltage codes on public SEEG rest data from seven people. Each record supplied eight contacts and seven adjacent differences. Both codes used early calibration and later test data. At six bits per value and fixed limits of six calibration standard deviations, an anchor and difference code reduced mean absolute 70 to 110 Hz power error from 9.84 to 2.11 dB. Error was lower in all seven records at the same 48-bit payload per sample. Mean contact waveform error rose from 8.74 to 10.24 percent. Mean bipolar waveform error fell from 25.34 to 6.27 percent. The descriptive 95 percent person-resample interval for the paired power-error difference was -9.89 to -5.46 dB. Other bit counts and voltage limits were secondary tests. Full source hashes, fresh sample reads, independent packet and Fourier checks, and a complete replay passed. The package keeps selected voltages, source records, code, a proof, packets, results, and four figures. This is an offline signal test. It does not test task decoding, hardware energy, or stimulation effects.

SEEG · signal compression · difference coding · high-gamma power · brain-computer interface · fixed-rate quantization