A Doppler ultrasound analysis method based on wavelet package transform was proposed for embolic detection. The embolic Doppler signal was firstly decomposed using the wavelet packet. Then the sensitive characteristics were calculated from each sub-band signal and used in the emboli classification. This method was applied to analyze 300 cases simulated and 163 cases clinical Doppler signals. The error ratio of embolic detection using this method was 13 percents lower than that using the traditional spectrogram analysis method. It was shown that this method overcame the limit between time and frequency resolution in the short time Fourier transform, improved the accuracy of embolic detection greatly and extracted more reliable parameters for the clinical diagnosis.