In this paper, signal energy, information entropy and joint entropy were analyzed to introduce general selective method of wavelet basis function, and these methods was applied in slow wave diagnosis of electrogastrogram(EGG). In order to verify the rationality of evaluation criteria, EGG signals were processed and filtered with different wavelet basis function in the experiments. Experimental results show that db9 wavelet could achieve the most energy of filtered signal, by the EE/ET composite index, db5 wavelet function could achieve the best result. Meyer wavelet could realize the most similar index. Power spectrum of EGG was used to prove the effectiveness of selective method of wavelet basis function, and effective selection criteria were also provided to select wavelet basis function in wavelet transform.