By using the brightness temperature value of the microwave radiometer, the corresponding reflectivity factor of the cloud radar and the Lband radiosonde data in the CMA Meteorological Observation Centre, the BP (back propagation) neural network is used as the inversion tool, and the atmospheric humidity profile is retrieved. The weather condition is divided into clear day, low cloud, middle cloud, high cloud as four cases. The inversions of the humidity profiles without reflectivity factor are compared, and the root mean square errors by the two inversion methods in each altitude layer are analyzed. The comparison results show that the correlation coefficient of the inversed humidity profile with the reflectivity factor is 0.862, the root mean square error is 14.9%. On the contrary, the correlation coefficient of the inversed humidity profile without the reflectivity factor is 0.763, and the root mean square error is 19.2%.