Fig. 6: State-dependent decoding model improves the performance of handwriting trajectory prediction.
From: Human motor cortex encodes complex handwriting through a sequence of stable neural states
a, Diagrammatic representation of the state-dependent decoding process during handwriting (DyEnsemble). Utilizing encoding models for each state established through TFC, DyEnsemble dynamically infers the state and adaptively assembles a state-specific decoder in real time based on the incoming neural signals. This approach allows for adaptive switching between decoding models along with state switches. b, The writing trajectory was decoded using state-dependent and state-independent neural decoders. c,d, Performance of handwriting decoding using RMSE and R2 between the ground truth trajectory and the decoded trajectory (offline evaluation with threefold cross-validation with no overlapping characters across folds in each session, n = 18; paired two-tailed t-test, t(17) = 31.20, P = 1.896 × 10−16, Cohen’s d = 2.199 (c); paired two-tailed t-test, t(17) = 31.37, P = 1.734 × 10−16, Cohen’s d = 2.495 (d)). All box plots depict the median (horizontal line inside the box), 25th and 75th percentiles (boxes), and minimum and maximum values (whiskers) Source data.
