Abstract: Embolic detection can be used for early diagnosis of cerebrovascular disease. The Doppler effect of ultrasound is a non-invasive means for the detection of emboli. In this paper, three signal processing methods, i.e., traditional spectral analysis, wavelet transform and Renyi information analysis, are used to analyze embolic Doppler signals. With the extracted characteristic parameters, a classification system using sensitive parameters is set up based on a BP neural network. The system can be used to classify emboli signals, interfering noise, and normal blood flow signals. From experiments of 300 simulated cases and 163 clinical cases of Doppler signals, it has been found that the embolic detection accuracy of the proposed method is higher than that using the traditional method. It is expected that automatic detection of emboli can be realized based on the proposed method.