The topic of this paper is an improved PolyMAX for a system with close modes or heavy damping. Due to the phenomenon of modal interference induced by close modes or heavy damping, the effectiveness of system identification may be therefore degraded. According to the theory of mechanical vibration, the response data function can be expressed in rational fraction form through the curve fitting technique, and the modal identification can be implemented from parametric estimation from rational fractional coefficients. However, we cannot acquire the mode shape information because the conventional common denominator model only indicates the frequency response function of a single-degree-of-freedom system. In this paper, we propose the matrix-fractional coefficients model constructed by the frequency response functions of a multiple-degree-of-freedom system to perform modal estimation. In addition, to get rid of the phenomenon of omitted modes from the distortion from modal interference among the vibration modes of a system, we introduce a system model with higher-order matrix-fractional coefficients in the proposed method. The vibration modes of systems and fictitious modes caused by the numerical computation can be effectively separated through the different-order constructed stabilization diagram. Modal identification can be implemented by solving the eigenproblem of companion matrix yielded from least square estimation. Numerical simulation of a full model of sedan, confirms the validity and robustness of the proposed parametric-estimation method for a system with modal interference.