Abstract: To pick up the specific EEG signals from the strong background noise is the key function of brain-computer interface (BCI) system. Based on the characteristics of the specific thinking-evoked EEG signals (TES), an approach is proposed to determine their distribution and pick up their waveform from strong noise. The EEG signals are decomposed by the way of discrete wavelet transform. Then, an analysis is made by combining the wavelet singularity detection with wavelet statistic analysis to determine in which dimension of wavelet transform the specific thinking-evoked signals are located. Thus, the specific thinking-evoked EEG signals can be reconstructed without noise. This approach provides an effective approach to eliminate the noise embedded in EEG signals, which is especially suitable to pick up the specific evoked EEG signals.