To monitor the performance of a Dynamic Tuned Gyro (DTG) on-line under closed-loop working conditions, a closed-loop identification approach was explored for the DTG system. Firstly, the plant model was simplified and discretized to obtain the identification model set and model order based on the complex model structure of DTG closed-loop system. Then, the Instrumental Variable (IV) method was used to the closed-loop identification of the DTG. Aiming at the deficiency of the traditional Ⅳ method, a BJIV method was proposed. The method was applied to BJ model so that the plant model and noise model were not restricted by the identification model. Furthermore, simulations were used to analyze the consistency and asymptotic distribution of the BJIV method. Finally, single and continuous identification experiments were conducted on a DTG closed-loop system. Obtained results indicate that the estimations of BJIV method are unbiased and consistent with different noise levels, and its asymptotic variance is near-optimal. The experiment results on closed-loop identification show that the identification fitting degree is more than 90%. After continuous experiments for two hours, DTG closed-loop system is stable and the identification results are reliable.