The EEG acquisition caused by noninvasive acquisition, which make the characteristics of signal contains more high frequency signal and the signal is difficult to separation cleanly, design a new method by combine component analysis (ICA) method with an improved algorithm combined with wavelet transform, to achieve the separation of EEG signal noising. The highfrequency signal in the target signal is filtered by Wavelet transform. And the output signal is reconstructed as the input signal of the ICA algorithm; overcome the independent component analysis method of genetic algorithm cannot distinguish the noise. This paper incorporates wavelet transform with ICA method based on genetic algorithm, and applicate them to evoked potentional extraction, the signal after wavelet packet filtering is reconstructed into the input signal of ICA，which get a good effect.