Skip to main content
. 2019 Jan 7;16:6. doi: 10.1186/s12984-018-0469-5

Fig. 2.

Fig. 2

Graphical representation of muscle synergy extraction process. a Processed myosignals of 22 muscles (M1 to M22 in the figure) were factorized into muscle synergies with a non-negative matrix factorization algorithm, for a range of 1–8 possible number of muscle synergies N. For each N in this range, the amount of variance in myosignals that these N muscle synergies could explain (R2) was determined using eq. 2. b Explained variances R2 and possible number of muscle synergies N plotted in an R2 versus N curve (blue curve in the figure). By repeatedly fitting straight lines through decreasing portions of the R2 versus N curve (grey dotted lines in the figure), it was determined which portion of the curve was essentially straight (red dotted line represents fitted line through this essentially straight part of the curve). The starting point of this essentially straight portion of the curve was established as the to-be-found number of muscle synergies N* (in the present example 4)