Relative indentation is the input signal estimating contact force model parameters, so the signal is required to have a higher precision to ensure the accuracy of the estimated contact force model parameters. However, in the impact experiment, the vibration displacements in multiple directions are often coupled in the relative indentation, resulting in a large error of the measured relative indentation. This paper presents an experimental correction method for the relative indentation. Firstly, the relative indentation is decoupled by the established model of the spatial position of the hammerhead relative to the sample to reduce the errors caused by the rotation of the pendulum boom and the vibration of the base. A pendulum impact test device is established to verify the correction method of relative indentation. The results show that the maximum relative error between the contact force estimated by using the corrected relative indentation as the input signal and the measured contact force is less than 3%. The estimated contact force is in good agreement with the measured value, and the correlation coefficient is above 0.92. It shows that the experimental correction of the relative indentation has achieved good results, which verifies the accuracy of the correction method.