Aiming at the problems of weak atmospheric echo signal and large noise interference detected by Doppler wind lidar, a semi-soft threshold algorithm based on wavelet transform is proposed. Firstly, the principle of semi-soft threshold algorithm based on wavelet transform is analyzed. The incoherent Doppler wind lidar system with Mach-Zehnder interferometer as frequency discriminator is built. Then, three algorithms of hard threshold, soft threshold and semi-soft threshold are used to test the echo signal detected by laser radar. The experimental results show that compared with other threshold methods, the signal-to-noise ratio(SNR) of semi-soft threshold is 4.933 5 dB and the root mean square error(RMSE) is 0.676 2. The denoising effect is better and the detection accuracy of atmospheric wind speed is effectively improved.