Fatigue is a phenomena associated with reduction of maximum force in muscles while performing functional activities. Assessment of fatigue is important in the field of clinical conditions, rehabilitation and sports. In this work, an attempt has been made to discriminate fatigue and non-fatigue conditions using cyclostationary analysis. The surface electromyography (sEMG) signals are recorded from triceps brachii muscles of 15 adult normal subjects during dynamic contractions. These signals are subjected to cyclostationary analysis using spectral correlation density function (SCD). Fast Fourier transform accumulation algorithm is adopted to estimate the SCD. To distinguish fatigue and non-fatigue conditions, two features namely sum and entropy are extracted from SCD coefficients. The results show that magnitude and number of peaks in SCD spectrum are higher in fatigue conditions. Moreover, the extracted features are distinct in muscle contraction zones. This method appears to be useful in analyzing the sEMG signals during muscle fatigue condition. © 2014 IEEE.