Abstract: The noise signal of ground penetrating radar(GPR)usually has the unsteady and nonlinear characteristics. To remove these noises and improve the accuracy of GPR image interpretation, denoising with HHT method is studied. Firstly, the basic theory of the HHT is described. Then the feasibility of HHT method is verified with numerical simulation of GPR data denoising. Finally, the method is used in the tunnel geological advanced prediction of GPR data processing. The research shows that the method can be used in GPR signal denoising and that the precision of GPR signal interpretation can be improved by contrasting three characteristic parameters images obtained by Hilbert transform with EMD composite images.