The analysis of the surface myoelectric signal is a valuable and the signal is recorded while a muscle is performing a sustained contraction tool for assessing the progression of localized fatigue. Mean and median frequencies of an electromyogram power spectrum are often used as indicators of muscular fatigue during static contraction. To evaluate muscle fatigue during cyclic dynamic contraction, according to the features of surface myoelectric signal during dynamic contraction, an instantaneous frequency detection technique based on cross Wigner-Ville distribution is proposed. The estimation error of the algorithm by means of a simulated stochastic process with statistical properties similar to the those of surface myoelectric signal. The theoretical and experiment validates that the instantaneous frequency suites well to tracking spectral changes due to muscle fatigue with very low SNR.