Global Navigation Satellite System-Reflectometry (GNSS-R) in soil moisture retrieval, the traditional method mainly includes linear regression and exponential regression methods. To address the defects of conventional methods such as poor prediction accuracy and sizeable computational effort, the Elman neural network with dynamic learning features is introduced. An Elman neural network-based soil moisture retrieval method is proposed to establish a multi-parameter retrieval model. Finally, the model is trained to validate the feasibility of this model. The results indicate that the soil moisture values estimated by the GNSS-R soil moisture retrieval method based on the Elman neural network have minor errors with the actual measured soil moisture values. Based on this model, the coefficient of determination (R 2 ) is 0.8988, and the Root Mean Square Error (RMSE) of soil moisture is 0.0207. When compared to the traditional linear regression model, the soil moisture values predicted by this method are more accurate and closer to the measured soil moisture values, demonstrating the method's validity and reliability.