In view of GIIRS, a hyperspectral vertical detector equipped on the FY-4A geostationary satellite, a channel selection experiment for the retrieval of temperature is carried out. Using line-by-line radiative transfer model, according to the information content, channels are selected by step-by-step iterative method, which can be used to describe the effective information of various atmospheric parameters in the infrared hyperspectral data. Finally, sensitive channels using for retrieval of temperature profile are obtained，56 channels in global and 58 in FY-4A detection range. The selected channels have small correlation with each other,and they have high information content, which can be used as the representative channel set for temperature profile retrieval research and data assimilation.