Model order determination is essentially important for modal parameters identification. A new index, named as model order indicator, was proposed to determine the proper model order. This method utilizes the impulse response function to construct the Hankel matrix. Then singular value decomposition is used to compute the singular values. The index is defined as the relative change ratio of the decreasing singular values and the maximum corresponds to the model order. Different cases, including various noise levels with different data length, are investigated. And the effectiveness of the proposed method is verified with measured signals from model test.