Evolutionary computation extracts a super sEMG feature to classify localized muscle fatigue during dynamic contractions

Abstract

This study developed a new muscle fatigue feature based on sEMG signals. The evolved feature is combining 11 traditional muscle fatigue sEMG parameters to optimally classify the sEMG signals. The myoelectric signals were recorded from 13 subjects performing biceps brachii contractions until fatigue. By utilizing the 11 features and a combination of randomly selected mathematical operators a Genetic Algorithm (GA)evolved a novel composite feature. Davies Bouldin Index (DBI) was used by the GA during the seeding and evolution process in its fitness function to measure the separation of the combined feature. Classification results show an average of 75.4% correct classification and a significant improvement (P <; 0.01) of 11.94% when compared with the averages of eight standard sEMG features that are used in current muscle fatigue studies.