Coherent Wind Lidar can obtain the atmospheric wind velocities through measuring the frequency shift of laser back from the atmospheric aerosol caused by Doppler effect. The scattered energy propagating back toward the receiver is very weak. So the frequency extraction from the weak signal belongs to the field of frequency estimation. The zero-padding Fast Fourier Transform(FFT) algorithm applied to the frequency extraction of the Coherent Wind Lidar has the advantage of algorithm simplicity, fast operation and high stability. Compared with the pulse-pair(PP) algorithm and the improved PP algorithm, zero-padding FFT algorithm was validated to have an obvious advantage in processing weak signal of Coherent Wind Lidar. In addition, the simulation with MATLAB shows that the measurable SNR of the weak signal can reach -26.6 dB by using the zero-padding FFT algorithm in the condition of 3 000 pulses non-coherent integration (128 samplings in one range gate). At last, the wind velocity test data gotten by the Coherent Wind Lidar was processed by improved PP algorithm and the zero-padding FFT algorithm, and through the processing results gotten by two algorithms, the advantage of zero-padding FFT algorithm is verified the advantage again.