Methodological and Computational Aspects of Extracting Extensive Muscle Synergies in Moderate-Intensity Locomotions
S. А. Moiseev, Aleksandr M. Pukhov, Е. А. Михайлова, Р. М. Городничев
Abstract
In motor synergy studies, modern approaches typically include the consideration of synergistic effects at one or, less often, two levels of the nervous system organization, as well as their interaction in the context of solving various motor tasks, including locomotion. However, even when considering movements that are similar in their biomechanical structure, there arise discrepancies in some aspects of studying muscle synergies, which is thought to be due to the use of different methodological and computational approaches that disregard the specificity of motor acts under consideration. The aim of the study was to find optimal parameters of electromyographic (EMG) signal preprocessing and to evaluate their influence on the results of extensive muscle synergy extraction during moderate-intensity locomotions. The study was carried out on 10 male sprinters during moderate-intensity treadmill running. Electromyograms of 16 superficial skeletal muscles of the trunk, upper and lower extremities were recorded. Muscle synergies were extracted by matrix factorization methods using various procedures of original signal preprocessing. It was found that extended muscle synergies extraction gives the best results through the application of a bandpass filter in the range of 20–450 Hz with additional 20-Hz filtering after electromyogram rectification. The number of extracted synergies from different phases of the running step cycle was greater when using principal component analysis (PCA) and smaller when using factor analysis (FA), while in terms of accounted variance, both methods return better results than when analyzing the whole running step cycle. The number of extracted synergies decreased proportionally to an original data amount decrease, while the quality of synergy extraction decreased as the number of analyzed muscles decreased. The results of original EMG signal reconstruction using data factorization methods were more effective when considering individual phases as opposed to the whole running step cycle. The indices of muscle synergy extraction were sensitive to the original set of EMG signals included in the analysis.