Abstract: At present, a variety of algorithms have been used to measure the flow velocity of backward-facing step turbulent media, but most of them have the problems of low repeatability and poor robustness. Therefore, a velocity measurement method of line array sensors based on wavenumber-frequency spectrum analysis is proposed. Firstly, the turbulence degree and Taylor freeze signal characteristics of the backward-facing step data are analyzed. Secondly, FDBM and MUSIC algorithms are used to convert sensors data from the x-t domain to the k-\omega domain, showing the characteristic "ridge line" related to the backward-facing step velocity. Then, the influence of different beam forming filtering algorithms and power spectrum block length on velocity "ridge" is explored. Finally, wavenumber-frequency spectrum is used to analyze the slope of ridge line to solve the velocity. Compared with the traditional cross-correlation method, the relative error is 2.83 \text% and the repeatability is 0.32 \text% . The results show that the wave-frequency spectrum method can effectively measure the backward-facing step flow velocity, has good repeatability and robustness.