The bearing vibration acceleration signal is one of the important indicators for evaluating the performance of the deep groove ball (DGB) bearing. The bearing impact signal has a significant relationship with the surface defects of the bearing. An accurate bearing impact vibration signal model can provide a basis for judging the size of the bearing defect. Most of the current bearing signal models divide the process of the rolling element moving through the defect into four events: entering the stress relief process of the defect, impacting the bottom edge of the defect, impacting the exit edge of the defect, and exiting the defect. The signal model composed of these four events cannot explain the signal errors that occur during the stress relief stage, and the calculation of the length of the event signal is not accurate. Therefore, the defective ball bearing signal model considering four events cannot accurately estimate the size of the bearing defect. To solve this problem, this paper proposes an improved impact vibration signal model considering multiple events for predicting the size of the bearing defect. This signal model considers the geometric shape of the surface defect and the interaction between the ball and raceway during the motion of the rolling element passing through the raceway defect. The improved signal model divides the stress relief process into two events according to the influence of the interaction between the rolling element and the inner and outer raceways on the phase and amplitude of the signal model when the rolling element enters the defect. Besides, this signal model also studies the influence of the interactions among the rolling element, the healthy inner raceway, and the defective outer raceway when the rolling element enters the defect to the time when it impacts the exit edge of the defect. The results show that the interactions between the rolling element and the inner and outer raceways have a significant influence on its vibration during this process. This improved signal model has higher accuracy in predicting the signal amplitude and signal length of the main events in the bearing’s motion through the defect compared to the signal model in the literature. This research provides a theoretical basis for estimating the size of the DGB bearing defect, bearing failure diagnosis, and dynamic analysis.