Abstract: Based on the three calculated nonlinear parameters of the waves, i.e., maximum Lyapunov exponents, complexity and approximate entropy, which are extracted from the original electroencephalograph (EEG) signals with wavelet transform technique, a comprehensive criterion is proposed to assess the human-body fatigue levels. The three nonlinear parameters are available to quantitatively reflect human's physiological activities and, further, evaluate the fatigue level of human body. Computing and analyzing statistically the existing EEG data, a comprehensive assessment criterion is set up for fatigue levels in terms of the nonlinear parameters as above. For the known 18 sets of different EEG signal data measured, the assessment results are obtained correspondingly. Compare with actualities, the identification accuracy of fatigue got in this way tends to 100%. However, the accuracy to differ being a bit tired from medium tiredness or being extremely fatigued is relatively lower than expected.