Doppler spectrum is the Ka-band cloud radar original data. Based on the consistency difference between cloud-precipitation signal and the “ghost echo” along with other noise signals in different Doppler spectral modes, this article uses the difference between short-pulse and long-pulse modes to determine the cloud signal boundary for the first time. The “ghost echo” is removed and the cloud signal boundary is adopted to calculate the noise level, forming a new denoising method. In addition, the sensitivity of cloud signal boundary changing with the difference threshold T is analyzed. Based on the small particle tracer method, the left cloud signal boundary after denoising is used to calculate the vertical air velocity. Two rainfall (snow) cases in Beijing are analyzed and compared with the in-situ aircraft observations of the vertical air velocity. It is verified that the new denoising method can effectively remove the “ghost echo”. Although a certain percentage deviation exists between the average vertical air velocity obtained by cloud radar and aircraft, the direction and magnitude correspond well in general. Thus, this method and the retrieved results are reasonable.