In this paper a method to estimate time delays between two periodic signals using Extended Kalman Filters (EKF) is presented. Fractional Derivatives were used as an inspiration in the underlying EKF system model of the time delay to improve the approximation of the time delay transfer function by a truncated Taylor polynomial. This method results to reduce estimation offsets. The approach is based on the assumption that, apart from some noise and the time delay to be estimated, there is no difference between the two signals. Simulations confirm that this method works well for Gaussian bell curve-like signals with a period that is one order of magnitude greater than the time delay.