A comparison of algorithms for adaptive estimation of the time delay between sampled signals

Recent work has shown that the time delay between two sampled signals that are highly correlated can be determined with iterative algorithms. The original algorithm was based on the LMS steepest descent technique. However, the error function depends on the correlation function of the input signals and thus, in general, is not unimodal. Two new algorithms will be examined which are not based on steepest descent or gradient techniques. One technique is based on a genetic algorithm which recombines binary strings to generate new strings which converge to the delay value with minimum error. The other technique is a linear search which estimates the error at different delay values and then chooses the delay value with minimum estimated error. Computer simulations will be used to illustrate the performance of these three algorithms.