A Comparison of Stationary and Cyclostationary TDOA Estimators

The classical technique for measuring the time-difference of arrival (TDOA) of a signal between two sensors has always been a filtered cross correlation. Recent papers, however, have suggested that cyclostationary correlation methods may achieve a much better TDOA estimate in the presence of strong interference for signals that are not truly stationary-such as digital communications signals. Of these, single-cycle cyclostationary methods, such as Gardner´s spectral coherence alignment (SPECCOA) algorithm, have been highlighted as robust methods of measuring the TDOA of a signal of interest regardless of the interference present in the environment. No clear comparison has been presented showing the relative benefits of cyclostationary TDOA methods over classical stationary methods and in particular over optimal stationary methods. This paper compares the best of these two classes of TDOA estimators together to determine their relative merits in a narrowband noise environment. This comparison reveals that single-cycle methods perform much worse than their traditional filtered correlation counterparts, while the full multicycle methods can achieve up to 40% lower mean square error than their stationary counterparts