In this paper, we propose an unbalance parameters identification method based on a time-domain response that combines a matrix equation and spectral correction technique. Numerical simulations and experiments are conducted to demonstrate the performance of the proposed method. A commonly used method for identifying unbalance parameters based on time-domain response is the equivalent load method, in which the modal expansion technique is used to extend the vibration response of a few of measurement points to the vibration response of all degrees of freedom of the model. However, this technique may result in incorrect identification. To solve the above problem, we first directly derive a matrix equation describing the relationship between the vibration response at the measurement point and the unbalanced forces based on the state space equations of the reduced-order model. This process avoids the application of modal expansion techniques and effectively improves the robustness of the identification results. The unbalanced force can then be obtained based on the vibration response by solving the matrix equation using the regularisation method. Finally, the unbalance parameters are obtained by processing the unbalanced force using the spectral correction technique. The simulation and experimental results show that, compared with the traditional equivalent load method, the proposed method can identify the unbalance parameters with higher accuracy, which proves its application prospect in balancing machines and field balancing.