Machine Learning Classification of Non-Specifically Trained Muscle between Endurance and Power Athletes

The variations in muscular contraction between endurance and power athletes have usually been evaluated from lower limb muscles. The aim of this study is to integrate the application of machine learning in automatically classifying the muscle performance recorded from upper limb muscle. Muscle contraction of bicep brachii was recorded based on the surface electromyography (sEMG) analysis. The evaluation of muscle performance consists of three main processing parts, i.e., pre-processing, feature extraction, and classification. EMG features were extracted from three types of domains: time domain (TD), frequency domain (FD), and time-frequency domain (TFD). For classification purposes, a Support Vector Machine (SVM) classifier was used, and the classification performance was analysed based on the classification accuracy. The best classification performance was observed from the feature set selected from sequential backward selection (SBS). This finding shows that it is possible to differentiate muscle performance from non-specifically trained muscle, which might be further related to the intrinsic properties of different groups of athletes.