This thesis deals with fine frequency tracking in orthogonal frequency division multiplexing (OFDM) systems. The usual practice for fine frequency tracking is to use multiple or an entire OFDM block of pilot or training data, either in the time domain or in the frequency domain to perform frequency offset estimation. Here we show that, instead of the elaborate approach of using multiple pilot samples, the problem can be greatly simplified by using only a single pilot sample from a selected time slot in the received time-domain OFDM block to still achieve reasonable tracking performance. It is proven that our single sample ML frequency offset estimator tends to be unbiased and its mean square error (MSE) will approach the Cramer-Rao bound (CRB) as SNR is increased.