In order to solve the difficult problem of tight sandstone reservoir prediction due to the small difference of wave impedance, this paper presents a prediction method for channel characterization and“sweet spot”based on the spectral decomposition and fusion technology. In this method，the nonlinear mapping relationship between the target sand body and the seismic spectrum was established to form a mixed data volume with direct frequency information，and a facies control model of macroscopic geological significance was constructed. In this way，the high-resolution inversion was performed to determine the distribution of“sweet spots”in tight sandstone reservoirs. By taking the tight sandstone in the Jurassic Qianyi Member of Tongnanba anticline as the research object，the early channel distribution of Qianyi Member was accurately described by spectral decomposition and fusion technology. The high-resolution inversion results based on mixed data volumes reveal that the porosity of channel sand bodies in Qianyi Member is less than 6%，and the porosity at the anticline is high（4%-6%）. The horizontal distribution of sand bodies in the channel is stable，where the thickness of the brittle sand bodies（with gamma-ray（GR）less than 65 API）changes rapidly and is generally more than 20 m at the anticline. The reservoir prediction results indicate that the channel sand bodies in Qianyi Member of the anticline belt feature large thickness，good brittleness，high porosity，and shallow burial depth，which are not only the“sweet spots”development areas in Qianyi Member but also favorable target areas for tight reservoir stimulation in the future.