Identification of Time-Varying and Time-Scalable Synergies From Continuous Electromyographic Patterns

Muscle synergies, which is the concept of modular activation of a set of muscles for producing complex motor behaviors, have been studied for a long time. Several definitions of muscle synergies have been proposed, and different algorithms have identified synergies in a large number of contexts. However, most of the studies so far used the dataset with a prior segmentation. This approach restricted the variety of movements that can be used for the muscle synergy analysis. We propose an extended version of the time-varying synergy algorithm to support continuous recordings of electromyographic signals and movements with different scales in time. We observed that the reconstruction accuracy with the new algorithm was comparable to the one of the original case scenario, whereas time-varying synergies algorithm had a poor performance when it was applied to movements with different scales in time. In addition, the similarity of parameters suggests that it is possible to identify a movement independent of movement frequency using time-varying and time-scalable synergies.