Based on the research of working condition and speed variation of elevator on the grain combine harvester, the kinematics and physical mechanics of grain movement on the momentum sensor were analyzed, and then a model of the mass flow sensor named voltage/elevator rotating speed model was put forward to improve the estimation accuracy of grain yield, so that the relationship between grain mass flow and impact force was described deeply by the proposed model. In the harvesting experiments, there existed many vibration noises. In order to eliminate those noises and ensure the accuracy of yield prediction, the impact voltage signals needed to be further processed. Firstly, a regression difference method was used to reduce vibration interference. Secondly, a dual threshold denoising method and two kinds of interpolation methods were proposed to eliminate the singular value of the difference voltages. One of the interpolation methods was threshold replacement and the other was formervalue replacement. The formervalue replacement method could relatively keep the real yield data, and its estimation error was smaller. Furthermore, the rotating speed was normalized to eliminate the influence of dimension and simplify calculation. Consequently, a standardized processing method of impulse voltage was put forward. The verification results showed that the proposed voltage/elevator rotating speed had a high accuracy and the RMSE was 2.03%, which could really characterize the actual situation of grain movement.