With high accuracy, high spatial-temporal resolution, large scale coverage, coherent Doppler lidar has been widely applied in the detection of wind shear, aircraft vortex, wind power generation, atmosphere turbulence and so on. For lidar signal processing, the key issue is how to extract weak Doppler frequency shift in the weak backscatter signal. Based on the atmospheric slices model, the simulated echo signal of coherent Doppler lidar was processed by different time-frequency methods. Simulation results show that the adaptive optimal-kernel time frequency representation outperforms the others, having the advantages of lower computation cost, suppressing cross terms efficiently and higher resolution in both time and frequency domains. Then the adaptive optimal-kernel time frequency representation was applied to the field experiment data derived from a 1.5 μm Coherent Doppler lidar in Hefei, Anhui Province in March, 2017. The retrieved wind velocity results were compared with that derived from the fast Fourier transform algorithm. Experimental results show that the range resolution is 1.2 meter within 3 kilometers, and maintains the continuity of wind speed retrieved form weak signal using a 50-points window in the far field over 3 kilometers. Furthermore it can track the wind details better and enhance the detection range to 6 kilometers as the temporal resolution is set to 1 second.