Abstract: In experimental studies on wave propagation in soil, wave velocities calculated by the start-to-start method and the cross-correlation method often yield inconsistent results with significant errors, and no widely accepted standard has been established. To address this issue, a more objective and accurate method for wave velocity measurement is proposed by improving the cross-correlation method. This method adopts the first resonant frequency as the excitation frequency and a Gaussian-modulated sinusoidal pulse (GMSP) as the excitation signal, coupled with the cross-correlation technique to determine wave velocity. Since the shape of the GMSP closely matches that of the received signal, the accuracy of the cross-correlation method is significantly improved. Moreover, the applicable frequency range of this method is defined, within which the vibration modes of EE and BE generate only pure P-waves or S-waves, effectively eliminating near-field effects. The wave velocity of sand at different water contents and dry densities are measured, and the results show that, the proposed method eliminates near-field effects in signals measured, and consistent wave velocities are obtained using both the cross-correlation and first arrival methods. Under saturated conditions, the P-wave velocity in sand reaches 1700 m/s due to the high modulus of water. As the water content decreases, the velocity drops sharply to 350–400 m/s, followed by a slight rebound. The S-wave velocity is approximately 150 m/s under dry conditions, decreasing to around 100 m/s as water content increases, and further declines near saturation. These variations are closely related to adsorbed water and matric suction. The test results of different dry densities show that with the increase of dry density, the growth rate of S-wave velocity is faster than that of P wave.