In this work, an attempt has been made to differentiate sEMG signals under muscle fatigue and non-fatigue conditions using multiscale features. Signals are recorded from biceps brachii muscle of 50 normal adults during repetitive dynamic contractions. After preprocessing, the signal is divided into six segments, out of which first and last segments are considered for this analysis. Fuzzy Approximate Entropy (fApEn) is computed for each subject in the time scales ranging from 1 to 10. Features such as median, low scale median and high scale median are extracted from Multiscale Fuzzy Approximate Entropy (MSfApEn) and used for further analysis. The results show an increase in amplitude of the sEMG signals under fatigue condition. The MSfApEn values are higher in the case of nonfatigue indicating more complexity. The features extracted for the series are effective in differentiating the two conditions. The statistical t test performed indicated high statistical significance (p-value <<0.001) It appears that this method of analysis can be used for clinical evaluation of muscles. © 2015 IEEE.