Decoding invariant spatiotemporal synergy patterns on muscle networks of lower-limb movements via surface electromyographic signals

Abstract

Muscle network, which enables sport enthusiasts to understand the insightful mechanism in lower-limb movement, optimizes cross-linked force-generation modes, enhances sports performance and reduces the risk of injury. To investigate muscle network, synergy patterns via the decompositions of surface electromyographic (sEMG) signal with strengths of linkage are rigorously analyzed. Although existing literatures cover functionalities of muscle network or synergy patterns separately, little evidence shows their collective mechanism. In this work, we deciphered the mechanism of synergy patterns on muscle network among lower-limb muscles. The experiments were conducted on twelve muscles from ten participants, with each one running at four pre-setup fixed speeds on the treadmill and sEMG recorded. Seven synergy patterns were extracted via non-negative matrix (NMF) decomposition, after calculating the mean value of interpretation variance (VAF), and the dynamic time warping (DTW) algorithm along with cosine similarity (CS) were applied for time-varying activation coefficients. Further, we recapitulated synergy patterns on multiple running gait cycles, obtained spatiotemporal invariant characteristics of muscle network from them, and decoded the force-generation modes through muscle network. Our research indicates that the weight similarity of synergy patterns reached 97.73 % on average for seven synergies under four different running speeds, meaning that alteration of speeds exerts little effects on synergy patterns on muscle network during lower-limb movements.