Frequent BCI recalibration due to changing neural activity mark a significant hurdle in long-term BCI engagement. Researchers at stanford now demonstrate a clever approach enabling unsupervised decoder adaptation by infering task labels and reinforcing the neural decoder
#neuroskyence
www.nature.com
A hidden Markov model that uses probabilistic retrospective inference allows for up to one month of unsupervised recalibration in an online cursor-based intracortical brain–machine interface.
