Tacit representation of muscle activities during coordination training: Muscle synergy analysis to visualize motor enhancement in virtual trajectory of multi-joint arm movement

The tacit representation of muscle coordination has been a major topic of research on motor control since Bernstein's pioneering work. To unravel the mechanisms underlying voluntary movements, we investigated the electromyography signals of six muscles in a non-dominant upper limb during fast spiral movements on a horizontal plane. We considered muscle synergy to be a coordination index that we defined as the balance among co-activations of agonist-antagonist muscle pairs; it is a composite unit related to adjusting the impedance across joints. The virtual trajectory is a time series and is a succession of equilibrium points at the endpoint; it can be represented by the weights for the muscle synergies. Muscle synergy analysis was performed for three healthy subjects before and after voluntary training for eight days. The results revealed that (1) the six muscle activities in a non-dominant upper limb during spiral tracing are explained by three muscle synergies representing the bases for the radial, argumental, and null movements, respectively, of a hand according to polar coordinates centered on the shoulder; (2) the three muscle synergy bases for movements hardly changed with voluntary training kinematics, whereas the kinematics assessment scores for all subjects greatly improved; and (3) the virtual trajectory drastically changed with motor enhancement, especially in the argument direction. When the subjects were asked to perform fast spiral tracing, the polished virtual trajectory formed a beautiful but slightly distorted spiral curve that rotated in the opposite direction of the kinematic trajectory. This may originate from dynamic compensation by the central nervous system. A central factor in motor skill acquisition must be learning a virtual trajectory by considering the dynamic effect of movement especially in the argument direction. Our results imply that virtual trajectories for movements can be learned with invariant bases using polar coordinates, i.e., muscle synergies.