Detrended Fluctuation Analysis (DFA), a novel nonlinear analysis, is a useful tool to study and understand the properties and complexity of surface Electromyography (sEMG) signal. Many noises that contaminate sEMG signals in real applications display trends that become difficult to analyze sEMG signal. The different types of trend fitting of DFA algorithm are used to eliminate these problems. In this study, the performance of DFA algorithm for sEMG-based control is presented. Moreover, the six types of trend, namely linear, quadratic, cubic, fourth order, fifth order, and sixth order polynomial functions are evaluated. The experimental results show that the scaling exponents of linear trend in various hand movements have the significant different values and small experimental variation. Hence, linear trend is a suitable trend fitting for sEMG signal. However, the interesting result is when we considered the scaling exponent of each pair of EMG hand movements; the appropriate trend is changed. Therefore, the selection of optimal trend fitting will improve the effectiveness in analysis of sEMG signals and become the useful tool to extract feature in sEMG-based control. Moreover, the DFA relate to the fractal analysis. The better performance of DFA algorithm over the other fractal parameters in sEMG-based control is shown.