Abstract: In the tunnel construction at limestone area,the unfavorable geological conditions ahead the tunnel face is the most important factor to restrict the safe and efficient construction of the tunnel. In addition,the adverse geological bodies such as karsts,faults and fracture zones have strong concealment. Thus,how to improve the interpretation level of tunnel short-range advanced geological prediction is particularly important. In order to improve the quantitative in- terpretation of ground penetrating radar (GPR) signals characteristics,a forward modeling based on time domain finite difference method and an experimental study of indoor physical model were carried out. Based on the principle of wavelet transform and singularity detection,a radar wavelet basis with a high similarity to GPR emission wavelet was constructed and added to the wavelet analysis toolbox. Using the radar wavelet basis as a foundation,a new time-energy density analysis based on wavelet transform method (TEDAWT) was proposed. The TEDAWT method was applied to the quantitative identification of the cavity fillings in the GPR forward modeling and physical experiment,and com- pared with the recognition results of the waveform analysis method,Db4 wavelet transform method and radar wavelet transform method. The results show that the waveform analysis method can effectively identify the size of cavity fillings, but reflection coefficient and other prior knowledge are required to determine the interfacial reflection of cavity fillings which is located in a crest or trough. The Db4 wavelet transform modulus maximum method is easy to obtain the singu- lar points of GPR signals,but the extraction position of singular points is related to the time-frequency localization characteristics of different wavelet bases. The relative error of the recognition results using Db4 wavelet transform meth- od is about 15% . The recognition effect of the radar wavelet transform modulus maxima method and the TEDAWT method are both better,while the TEDAWT method has higher resolution,and does not need to select the optimal scale. When the size of the cavity fillings is larger than the wavelength of the electromagnetic wave,the relative error of the TEDAWT method is less than 5% .