Noise suppression is a key technique in retrieving weak magnetic signal from human heart. In order to improve the signal-to-noise ratio (SNR) of magnetocardiography (MCG) signals, signal average method is widely used. However, due to the existence of local environmental interference, averaging the whole section data would unavoidable bring distortion to the average signal. In this paper, the authors proposed a kind of selective average method by using the template matching. The results showed that this method can effectively eliminate the data section with interference such as low frequency fluctuation, pulse burr etc. Thus averaging MCG signal with high SNR and fidelity will be obtained.