Recently, many studies have demonstrated the use of teleseismic P wave coda autocorrelation for imaging lithosphere structures. However, the reliability of the extracted reflections remains uncertain and a means of evaluation is lacking. In this paper, we propose a velocity analysis method that conveniently solves this problem in place of a synthetic experiment. This method considers the average velocity used for the horizontal slowness correction as an unknown quantity, and then uses the continuously varying average velocity for the horizontal slowness correction. Finally, this method obtains a stacked result that varies with the average velocity and the vertical two-way travel time to produce a va−t0 diagram. This method is similar to the velocity analysis method used in exploration geophysics. In this diagram, reliable reflections correspond to focused energy clusters, while noise lacks this feature. Therefore, this method helps determine which reflections are reliable, while also finding the appropriate parameters for data processing. Synthetic data tests were performed to demonstrate the validity of this method, as well as a test of field data for station BOSA, which illustrates the successful application of the method in the case of a sharp and flat Moho discontinuity. Finally, we applied the method to the NCISP-6 dense array, and observed obvious energy clusters representing reflections from the Moho discontinuity in the results of most stations. The depth and shape of the Moho discontinuity determined by this test is consistent with receiver function results, which verifies the robustness of this method in relatively complex applications.