Performance of ML estimators for time of arrival estimation

The performance of a set of high-resolution algorithms for estimating the time of arrival of a band-limited signal received on a multipath channel is compared. The algorithms investigated use an increasing degree of known information about the channel and estimate the unknown elements using the maximum-likelihood criterion. In the first algorithm, no information on the amplitudes and phases of the multipath components is assumed. In the second algorithm, the statistics of the complex path gains are known to be jointly Gaussian and in the third case, the path gains are jointly Gaussian with a known correlation matrix. The three algorithms are compared using a simple two path channel model where the second path position is assumed known and the statistics of the error in measuring the delay of the first path are compared