In order to study the relationship between the amplitude of the early-time-signal of GPR and the soil moisture content, a physical model was established and seven groups of soils with different water contents were prepared. 300, 400, and 900 MHz GPR antennas were used for detection. Three typical periods of the early-time-signals of GPR (first signal onset, first positive half cycle and first half cycle) were selected to calculate volume moisture. Firstly, Hilbert transform was used to calculate the envelop amplitude, then the reciprocal of their average of the three corresponding periods were calculated respectively, finally the relationship between the reciprocal of the average envelope amplitude and the results of moisture measured by the drying method were fitted. All fitting results showed linear correlation and the correlation coefficients were 0.883, 0.926 and 0.867, respectively. But the reciprocal of the amplitude envelope in the first positive half cycle had the best fitting effect with soil volume moisture, and it’s average relative error of moisture content was 0.483%, the average relative standard deviation between volume moisture calculated by the soil dielectric constant test and the Topp formula and the measured volume moisture was 1.069%. In conclusion, it is more accurate to predict the moisture of soil by studying the reciprocal of the radar wave average envelope amplitude in the first positive half cycle with the AEA method.