Doppler Ultrasound Blood Flow Signal (DUBFS) is a non-stationary signal that is widely used in the study of the clinical diagnosis of cardiovascular diseases. According to the theory of pulse diagnosis of Traditional Chinese Medicine, in this paper, a feature extraction method based on Hilbert_Huang transform is proposed in order to investigate the relationship between the DUBFS of wrist radial artery and the pathological changes of certain organs. The extracted features have applied in classification experiments on 4 groups of data, which are healthy persons, gastritis patients, cholecystitis patients and nephritis patients, respectively. Experimental results with high recognition rates demonstrate that Hilbert_Huang transform is an effective method of time-frequency analysis for DUBFS, and extracted features by proposed method have promising discriminating ability between the healthy people and 3 kinds of patients.