A Bayesian approach to time delay estimation

Estimation of time delay between two sensors or estimation of multipath delay at single sensor is carried out by calculating the a posteriori pdf of the time delay from a given prior distribution. When the time delay is fixed over the observation interval, the a posteriori pdf is obtained from a fixed set of Kalman filters, each of which is matched to a different delay. When the time delay is Markov, an expanding set of Kalman filters is required. For this case, an approximation commonly used for target tracking in multitarget environments is used to obtain an algorithm with stable memory requirements. Monte Carlo simulation results are presented to determine algorithm performance.