FATIQ: A Deep-NET based regression model for force reduction coefficient and repetition prediction during fatigue for biceps and

Fatigue affects generation of muscle force and hence understanding of fatigue helps to understand limitations in performing a task. In the current study, we collected electromyogram and motion data of the healthy subjects' upper limbs during biceps and hammer curl exercises in two cyclic models with repeated flexion-extension and flexion-hold-extension-hold at various speeds. About 1211-time domain features and frequency domain features were extracted and reduced to 131 features and were used to train and validate a Deep NET-based regression model (FATIQ) developed inhouse for prediction of force reduction coefficient (FRC). The force reduction coefficient was used to estimate the repetitions of flexion-extension cycle and further used in the prediction of repetition to fatigue. Post regression the features, namely, psr, tm4, tm5 and wamp, were observed to be linearly distinguishable in FRC with 95% confidence level. The study is novel in understanding fatigue of biceps and triceps muscles in terms of force reduction coefficient and the number of repetitions to fatigue. The proposed model could predict the force reduction coefficient, the current repetition number, and the repetition to fatigue. Researchers can use this model to monitor a subject's performance in strength training and rehabilitation.