Abstract: Geological conditions of mine are much more complex,the heterogeneity and anisotropy of coal and rock,the coal seam structure and association with rock strata,and the environment discrepancies of shot or receiver caused by the roadway conditions,all of this will affect the formation and propagation of channel wave,and cause the actual propagation characteristics of channel wave different from “rock coal rock” theoretical model,the accuracy and difficulty of the calculated coal seam thickness by the ISS has increased. In order to improve the prediction accuracy of coal seam thickness in coal mining face,a method based on the theory of channel wave dispersion and frequency division processing technology is proposed,the relationship between the group velocity and the coal seam thickness under different frequencies is established,the influence of imaging frequency band on the imaging accuracy is analyzed by using the narrow band filtering method,and the principle of frequency band optimization for CT is proposed. It can improve the inversion accuracy of coal seam thickness from group velocity. This method is applied to the coal seam thickness detection of 2408 working face in Jiaoping mining area,Tongchuan,Shaanxi province. The average thickness of coal seam is 7．6 m. A comparative analysis of CT frequency band in the range of 50-80 Hz is developed. The classification relationship between coal seam thickness samples and group velocity of channel wave is defined. 87 samples before mining are used to evaluate the impact of each imaging frequency band on the group velocity and the coal seam thickness. The correlation between the group velocity of the optimized frequency band CT and the coal seam thickness samples before and after mining is respectively 81. 6% and 85. 1% . The error between the inversion thickness of coal seam from optimum frequency band CT and 181 actual coal seam thickness data before and after mining is analyzed. The absolute value of prediction error of coal seam thickness is less than 5% and 10% ,reaching 48. 62% and 85. 64% respectively. The results show that the accuracy of coal seam thickness predicted by the ISS is related to the frequency band of CT. The higher main frequency and the narrower band of frequency,the higher accuracy of CT for calculating the coal seam thickness,but the smaller the maximum thickness can be inversed. Therefore,the main imaging frequency should consider that CT has good effect on most areas of the coal seam in the study area,use the frequency corresponding to the average coal seam thickness on the curve which has the maximum change rate of channel wave group velocity,and the best band width is the minimum frequency range with higher correlation between the group velocity and the coal seam thickness samples.