Real-time estimation of moving time delay

A two-step algorithm for the estimation of a rapidly varying time delay between two stochastic signals is described. In the first step the maximum-likelihood (ML) estimate is computed over an observation interval small enough to consider the delay to be approximately constant. Due to the short averaging interval the probability of ambiguous peaks is greatly increased. Therefore in the second step, a non-linear, adaptive postfiltering algorithm is presented that effectively suppresses ´outliers´ of the ML-estimator.