This paper proposes a system identification methodology by using eigensystem realization algorithm (ERA) with a doping signal such that noise effect can be attenuated. The doping signal is obtained by integrating optimization technique with singular value decomposition (SVD) technique. Theoretical analysis is performed to interpret the necessity of using the optimum signal and to explain the relationship between SVD and optimum signal. Simulation results show that the proposed ERA with a novel optimum signal can achieve higher accurate results as compared to that without doping the optimum signal.