There are three major feature extraction domain of surface EMG (sEMG), time domain, frequency domain and nonlinear domain. Recently, Empirical Mode Decomposition (EMD) is developed. EMD is based on Hilbert-Huang transform (HHT) algorithm. This is mainly used in non-linear and non-steady-state signal analysis.. In addition, Ensemble Empirical Mode Decomposition (EEMD) is improvement of EMD by reducing mode mixing effects on EMD. This project investigates the feature performance derived by EMD and EEMD, for three feature extraction domain on sEMG. That is RMS amplitude, median frequency, and %RET / %DET derived from Recurrence Quantification Analysis. Pilot study showed that Median frequency derived from EEMD has lowest slope variation than the other method, such as Fourier transform, wavelet and EMD.