Single frequency estimation with non-uniform sampling

A new frame structure is proposed for pilot symbol assisted modulation (PSAM) systems which includes a small number of known pilots at the start of each frame. This facilitates the estimation of large frequency offsets without aliasing. Cramer Rao lower bounds and the exact ML estimates are derived for AWGN and Rayleigh fading channels. Finally, simulations show that the ML estimator has much better performance than existing reduced complexity estimators when applied to this sparse observation model.