The EEG signal is one of mini-voltage. In order to improve the performance of EEG and increase the measure efficiency, we must eliminate the noise in EEG. Wavelet threshold de-nosing techniques provide a new way to reduce noise in EEG. This paper proposes a wavelet new threshold de-noising function based on the study of the disadvantages of the hard threshold function and the soft threshold function. The simulation results show that the wavelet new threshold function is able to eliminate the noise effectively which can not only save the smooth effect of the soft threshold function but save the numerical value advantage of the hard threshold function.