Abstract: A dynamic programming algorithm based on the global optimization is presented to extract the maximum frequency waveform of the Doppler ultrasound blood flow signal. This approach was tested using simulated and clinical Doppler ultrasound signals together with other three methods. The relative error of maximum frequency waveform estimation using this method is 20 times smaller than the one using the traditional methods at most. The much better estimation precision under the strong noise shows that the dynamic programming extraction method has better noiseproof performance. It is shown that this proposed method can be used to extract the maximum frequency waveform precisely from the strong background noise and can be applied further to clinical vascular diseases diagnosis.