In the process of signal processing in the laser Doppler velocity measurement system, the fast Fourier transform is usually used to extract the Doppler frequency shift and calculate the motion speed of the measured object. Due to the influence of spectrum leakage and barrier effect, the measurement accuracy can be reduced. In this paper, a hybrid convolution window function with a four-term Nuttall window and a five-term Rife Vincent window combined with an improved six spectral line interpolation algorithm is proposed to reduce the effects of spectrum leakage and barrier effects, so as to improve the accuracy of laser Doppler signal processing. A dual-beam backscattering differential laser Doppler velocimetry platform is built, and the minimum relative error measured with this hybrid convolution window with the addition of Gaussian white noise with R_SN=-10 dB is 0.0027% by simulation and real measurements, whereas the minimum relative errors measured using the two single window functions are 0.0103% and 0.0461%, respectively. Thus, the effectiveness of this method is verified.