Abstract: Empirical mode decomposition (EMD) was applied to deal with non-stationary signal, and can describe the characteristics of frequency. But EMD yields has its own interpretation of combinations of pure tones, and mode mixing is one of the problems. We give a solution involving an improved masking signal. The improved masking signal method is based on masking signal method, to change the way choosing the masking signal, we get the IMF1 without mode mixing. The method also allows EMD to be used to separate components that are similar in frequency that would be inseparable with standard EMD techniques. The experiment shows that the approach based on a improved masking signal to improve EMD is practical and effective.