Abstract: Based on the frequency dispersion theory, the channel wave can inverse the coal seam thickness distribution in mining face and delineate the thin coal area to a certain extent by using a certain frequency or band-pass filter frequency band. However, due to the complexity and variability of the coal seam thickness in mining face, there is a significant difference between the actual and theoretical dispersion characteristics of channel wave, which reduces the prediction accuracy of coal seam thickness. In order to further improve the effect of channel wave in predicting coal seam thickness, based on the theoretical relationship of group velocity, coal seam thickness and frequency, the sensitive frequency, narrow sensitive frequency band and wide sensitive frequency band of different coal seam thickness are calculated by second-order, third-order and fourth-order derivative, and using the relation between group velocity and frequency, each order derivative and its variation curve. The coal seam thickness and its range can be calculated for different sensitive frequencies. A coal seam thickness prediction method based on multi-frequency joint application of channel wave is proposed by using multiple sensitive frequency weighted reconstruction and sensitive frequency band. For the test mining face, the coal seam thickness is inverted by the group velocity of the 147 Hz sensitive frequency, the narrow sensitive frequency from 123 Hz to 172 Hz, and the wide sensitive frequency band from 103 Hz to 188 Hz corresponding to the disclosed coal seam average thickness in tunnelling. At the same time, considering the effective frequency band of channel wave and the range of mainly disclosed coal seam thickness in tunnelling, the group velocity of each coal seam thickness corresponding to the sensitive frequency is weighted and reconstructed with the proportion of each coal seam thickness as the weight factor, and the coal seam thickness in mining face is predicted by the multi-frequency weighted reconstruction group velocity. The distribution trend of coal seam thickness predicted by single sensitive frequency, narrow sensitive frequency band, wide sensitive frequency band and multi-frequency weighted reconstruction is basically consistent with the actual mining situation, and the thick coal seam area and thin coal seam area are well predicted. The prediction accuracy of the four methods can reach 76.15%, 83.68%, 81.17% and 86.62% respectively for the coal seam thickness with the deviation not more than 20% (0.50 m). The results show that the sensitive frequency and the sensitive frequency band will decrease, and the range of sensitive frequency band will also decrease, while coal seam thickness increases. With the increase of frequency, the predicted optimal coal seam thickness will become thinner, and the thickness range will also become smaller. Single sensitive frequency, narrow sensitive frequency band, wide sensitive frequency band and multi-frequency weighted reconstruction can effectively predict the thickness of coal seam in mining face. However, the prediction accuracy and precision of the sensitive frequency band and multi-frequency weighted reconstruction of multi-frequency joint application are improved compared with the single sensitive frequency. In particular, the multi-frequency weighted reconstruction has the best effect. Therefore, multi-frequency weighted reconstruction can be preferred, narrow sensitive frequency band and wide sensitive frequency band can be followed, and single sensitive frequency is the finally in predicting the coal seam thickness by channel wave.