Classifying Muscle Performance of Junior Endurance and Power Athletes Using Machine Learning

This paper aimed to characterise the muscle performance of endurance and power athletes using machine learning approach. In this regard, electromyography (EMG) features were extracted from the vastus lateralis muscle and used to feed the support vector machine (SVM) classifier. The performance of various EMG features was evaluated based on their classification accuracy, sensitivity, specificity, and F-score. The accuracy was the highest for the feature set selected using the feature selection approach compared to the single feature. Specifically, the performance of sequential backward selection (SBS) was superior to the sequential forward selection (SFS) approach. Meanwhile, based on the SVM classification result, the radial basis function (RBF) kernel performed better than the other investigated kernel types, such as linear, polynomial, and sigmoid kernel. This muscle characterisation of endurance and power athletes may be useful as a muscle-monitoring tool for future talent identification and talent development, particularly in young athletes.