Muscle endurance time estimation during isometric training using electromyogram and supervised learning.

Constant-force isometric muscle training is useful for increasing the maximal strength , rehabilitation and work-fatigue assessment. Earlier studies have shown that muscle fatigue characteristics can be used for evaluating muscle endurance limit. To predict muscle endurance time during isometric task using frequency spectrum characteristics of surface electromyography signals along with analysis of frequency spectrum shape and scale during fatigue accumulation. Thirteen subjects performed isometric lateral raise at 60% MVC of deltoid (lateral) till endurance limit. Time windowed sEMG frequency spectrum was modelled using 2-parameter distributions namely Gamma and Weibull for spectrum analysis and endurance prediction. Gamma distribution provided better spectrum fitting (P < 0.001) than Weibull distribution. Spectrum Distribution demonstrated no change in shape but shifted towards lower frequency with increase of magnitude at characteristic mode frequency. Support Vector Regression based algorithm was developed for endurance time estimation using features derived from fitted frequency spectrum. Time taken till endurance limit for acquired dataset 38.53 ± 17.33 s (Mean ± Standard Deviation) was predicted with error of 0.029 ± 4.19 s . R-square: 0.956, training and test sets RMSE was calculated as 3.96 and 4.29 s respectively. The application of the algorithm suggested that model required 70% of sEMG signal from maximum time of endurance for high prediction accuracy. Endurance Limit prediction algorithm was developed for quantification of endurance time for optimizing isometric training and rehabilitation. Our method could help personalize and change conventional training method of same weight and duration for all subjects with optimized training parameters, based upon individual sEMG activity.

Copyright © 2019 Elsevier Ltd. All rights reserved.

Declaration of Competing Interest None declared.