<div>The observations for continuous-time dynamical models becomes widely available at a manner of extremely high frequency. In this paper, we consider the estimation of quadratic variation of semi-martingales using the discretized observations. Assuming the observations contain both multiple records and additive measurement error, we propose a double averaging procedure to estimate the quadratic variation of semi-martingales, the related asymptotic distribution of the estimator is established. Moreover, we show that the proposed double-averaging estimator achieves the same asymptotic efficiency as the case of no presence of multiple records. Simulation studies support the theoretical results, and we illustrate the estimator by a real data analysis. This is joint work with Bingyi Jing and Xinbing Kong.</div>