We consider the problem of estimating the frequency offset between transmit
ter and receiver in flat-fading communication channels using a training seq
uence. In [1] and [2], two frequency estimators were proposed to solve the
aforementioned problem. They rely on weighted linear regression for the pha
ses of the correlation sequence [1] or their differences [2]. The method in
[1] needs phase unwrapping and both methods require knowledge of the corre
lation of the fading process. In this letter, two simple methods which do n
ot require any of these assumptions are proposed. The first is the unweight
ed version of the method of [2]. The second is drawn from the array process
ing framework. It is based on a nonlinear least-squares approach that assum
es an unstructured correlation of the fading process. The performance of th
e proposed methods is as good as that of the methods in [1] and [2], but th
e robustness of our methods to modeling errors is much enhanced.