Abstract: Heart sound signal feature extraction and automatic identification have important clinical significance. In order to fully reflect the characteristics of the heart sound signal, in this paper, firstly we used DB6 wavelet to reduce noise of the heart sound signal, and then used the hilbert-huang transform (HHT) to extract the time-domain and frequency-domain characteristic values of heart sound signal, and then extracted band energy values through adaptive lifting wavelet packet, and finally classified and recognized heart sound by support vector machine. We experimented to 240 cases of abnormal heart sounds and normal heart sounds from clinical collection, the results show that recognition rate can reach 97.5%. It is clear that HHT and adaptive lifting wavelet packet are effective identification means for normal and abnormal heart sound signal.