Currently,velocity spectrum analysis methods widely used for ground-penetrating radar(GPR) signals mostly construct velocity spectra by superposing the amplitude energy of coherent signals to estimate electromagnetic wave velocities in subsurface media.In the case of multiple peaks and troughs in signal wavelets,velocity spectra constructed using these amplitude-based methods display multiple energy clusters,adversely affecting the identification,picking,and velocity estimation of subsequent energy peaks.Hence,this study proposed a method for estimating electromagnetic wave velocities in subsurface media based on common-offset GPR signal envelope and 3D velocity spectrum analysis.By scanning the signal envelope of hyperbolic diffracted waves in the GPR profile,the proposed method constructed the 3D velocity spectra of superimposed energy varying with zero-offset two-way travel time,test velocity,and measuring point position.Moreover,it extracted the slices of 2D velocity spectra according to the positions of hyperbolic vertices in the 3D velocity spectra.On this basis,the test velocities corresponding to the energy peaks in the slices of 2D velocity spectra were picked as the electromagnetic wave velocities in subsurface media.The numerical test results show that compared to the amplitude-based methods,the signal envelope-based 3D velocity spectrum analysis method obtained velocity spectra characterized by fewer continuous energy clusters,more concentrated energy,and minor velocity estimation errors,thus more effectively constructing the velocity model with inverse-time migration.