Abstract: The operational wind profiler radar network of the China Meteorological Administration cannot reliably distinguish precipitation and turbulence signals in PSD (power spectral density) data during rainfall. This limitation causes significant deviations, or even errors in the retrieved atmospheric vertical and horizontal wind speed products. Based on operation requirements, this article designs processing methods and core algorithms for multiple technical aspects of power spectrum data, and establishes a comprehensive algorithm processing flow. The results have been tested and proven to be effective. Among them, the key technology for noise level estimation employs Savitzky-Golay filtering instead of moving average for preprocessing spectral lines, and designs a noise level estimation method based on sliding Z-test after power spectrum sorting. Simulation results show that the performance is significantly better than that of existing algorithms. Additionally, the power spectrum signal recognition and detection method outlined in the article can precisely capture the bimodal signals indicative of precipitation through the application of various statistical algorithms and physical verification processes. Following Gaussian fitting, the mean velocity is calculated, enabling differentiation between vertical motion caused by atmospheric turbulence and the falling velocity of precipitation. Validation using Beijing radar observations on 30 July 2023, shows that the proposed algorithm accurately distinguishes atmospheric motion from precipitation fall velocity under all conditions, aligning with the fifth-generation ECMWF reanalysis data. In contrast, current operational products misinterpret precipitation fall velocity as atmospheric motion, resulting in substantial retrieval errors.