Recalibration burden

The recurring cost of restoring a decoder to usable accuracy after the neural signals it reads have changed. Because intracortical, ECoG, and EEG signals drift from one day — sometimes one hour — to the next, a decoder trained today typically degrades by tomorrow, and classic systems recover by running a fresh supervised calibration block: the user makes cued movements or attempts, labels are collected, and decoder parameters are refit. The recalibration burden is the sum of the calibration blocks, supervised labeling, clinician time, and lost usable time this demands. Reducing or eliminating it is the organizing goal of the entire drift-robust decoding field.

The burden is not a minor nuisance but a product-defining constraint. For a person using a home device without a technician, a decoder that needs a ten-minute cued calibration every morning is a very different assistive technology from one that simply works on wake. Progress here is measured less by peak accuracy than by how long a decoder holds usable accuracy with no supervised recalibration — hours, days, weeks, ideally the working life of the implant.