Fig. 5: Percent of variance accounted for and inter-stride consistency of experimental reconstructed data in neonates and the other groups.

a Average (z-transformed, across subjects) percent of variance accounted for (VAF) by the reconstructed EMGs in infants (left panels) and adults (right panels) using different number of modules and different decomposition algorithms and methods. In the left columns using principal component analysis (PCA) algorithm, from top to bottom: spatial decomposition, temporal decomposition, and PCA on single stride. In the right columns using non-negative matrix factorization (NNMF) algorithm, from top to bottom: spatial decomposition, temporal decomposition and space-by-time decomposition. Shaded areas denote 95% confidence intervals across subjects. Note that VAF seemed to be higher in neonates and to decrease with age in infants, though it was not sufficient to evaluate the dimensionality of the data sets. b Average (z-transformed, across subjects) inter-stride consistency measures across subjects for the modular decomposition (NNMF) of bilateral EMGs using different number of modules in infants (left column) and adults (right column). From top to bottom: inter-stride consistency of basic activation patterns (calculated as cosα, where α is the angle between two basic activation pattern vectors) derived from spatial decomposition, inter-stride consistency of synergies (cosα) derived from temporal decomposition, diagonality of the activation coefficient matrices derived from space-by-time decomposition. Shaded areas denote 95% confidence intervals across subjects. c Histograms, for each group of subjects, of the number of modules identified using the maximum of slope of the different consistency methods (from top to bottom as in a). Bar height denotes the percentage of subjects whose maximum slope of consistency measures is located in the corresponding number of modules. d Correlation between the number of modules identified from the maximum slope of diagonality and the age in days for each group of participants, excluding (left panel) and including (right panel) adults. Each data point represents the average (±SD) value across subjects for each group. Dashed lines, linear regressions. The sample size is the same for all panels (n = 13 for neonates, n = 9 for infants g1, n = 13 for infants g2, n = 21 for infants g3, n = 49 for infants g4, n = 29 for toddlers, n = 8 for preschoolers, and n = 15 for adults).